adam_di_akimow/lift_indicator_suite
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

# 1. Базовая структура проекта

Browse source
This commit is contained in:
Dmitry Akimov 2026-03-05 16:42:22 +03:00
parent a318805e4e
commit 9331028b0a
13791 changed files with 2946121 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

118
README.md
View file

@ -1,2 +1,118 @@
# tft_manufacture_test
Рабочее окружение NXP IMXRT1052
## Структура проекта
Монорепозиторий для MIMXRT1052CVJ5B. Три независимых firmware-проекта с общей инфраструктурой сборки, тестирования и инструментарием.
---
## Структура репозитория
```bash
/
├── .devcontainer/ # VSCode Devcontainer — единое окружение для всех разработчиков
├── cmake/ # Общие CMake модули и toolchain files
│ ├── toolchain_arm.cmake # ARM cross-compilation toolchain
│ └── toolchain_host.cmake # Host GCC для unit-тестов
├── sdk/ # NXP MCUXpresso SDK — vendored, только нужные компоненты
│ ├── CMakeLists.txt # CMake-таргеты для каждого драйвера
│ ├── CMSIS/
│ ├── devices/
│ │ └── MIMXRT1052/
│ │ ├── drivers/ # fsl_flexcan, fsl_lpuart, fsl_usdhc и др.
│ │ ├── startup/
│ │ └── utilities/
│ └── components/
├── bsp/ # Board Support Package
│ └── board/
│ ├── board.mex # Исходник конфигурации для NXP Config Tools
│ ├── pin_mux.c/h # Сгенерировано из board.mex (Pins Tool)
│ ├── clock_config.c/h # Сгенерировано из board.mex (Clocks Tool)
│ └── board.c/h # Ручная инициализация специфики платы
├── lib/ # Библиотеки и зависимости
│ ├── CMakeLists.txt # Агрегатор — подключает нужные модули через опции
│ ├── freertos/ # vendored
│ ├── fatfs/ # vendored
│ ├── mbedtls/ # vendored (если используется)
│ ├── unity/ # vendored (test framework)
│ ├── fff/ # vendored (fake functions для тестов)
│ └── hal/ # submodule — аппаратно-независимые библиотеки
│ # (второй разработчик, активно развивается)
├── firmware/
│ ├── test/ # [Проект 1] Тестовая прошивка — входной контроль
│ ├── bootloader/ # [Проект 2] Загрузчик с поддержкой A/B обновления
│ └── app/ # [Проект 3] Основная боевая прошивка (FreeRTOS)
├── tests/ # Host-тесты (unit + integration)
│ ├── unit/
│ └── integration/
└── tools/ # Скрипты для прошивки, провизии, HIL-тестов
├── flash_usb.py # Прошивка через USB ROM (blhost / nxp-spsdk)
├── flash_remote.sh # Прошивка на удалённый сервер через SSH
└── provision.py # Производственная провизия (SPT / nxpimage)
```
---
## Три firmware-проекта
### 1. Тестовая прошивка (`firmware/test/`)
Bare-metal прошивка для входного контроля на производстве. Проверяет базовую работоспособность всех интерфейсов и периферии: CAN, UART, SDRAM, QSPI Flash, uSD, RGB-интерфейс, гальванически развязанные входы, светодиоды, кнопки. Загружается через USB ROM (Serial Download Mode). **Рекомендуется начать разработку с этого проекта** — он наиболее прост и позволяет полностью отладить окружение сборки.
### 2. Загрузчик (`firmware/bootloader/`)
Отвечает за обновление боевой прошивки в полевых условиях. Поддерживает схему A/B с шифрованием (обновление через uSD). Обновление самого загрузчика — только через внешний инструмент (USB ROM + blhost), не через себя. Загружается на производстве вместе с первой версией боевой прошивки единым blob-ом.
### 3. Боевая прошивка (`firmware/app/`)
Основная прошивка на базе FreeRTOS. Включает FatFS, бизнес-логику, модули. Обновляется через загрузчик по схеме A/B.
---
## Управление зависимостями
| Зависимость | Подход | Причина |
|---|---|---|
| NXP SDK | vendored | Стабильная версия, обновлений не планируется |
| FreeRTOS | vendored | Стабильная версия |
| FatFS, mbedTLS и др. | vendored | Стабильные версии |
| Unity + fff | vendored | Маленькие, стабильные |
| lib/hal (второй разработчик) | **submodule** | Активно развивается параллельно |
**Принцип:** всё что не меняется — vendored (скачано и закоммичено). Submodule только для активно развивающихся зависимостей. Это обеспечивает полностью автономную сборку после `git clone` без доступа к интернету.
## Devcontainer — состав окружения
| Инструмент | Назначение |
|---|---|
| `arm-none-eabi-gcc` | Сборка firmware для таргета |
| `arm-none-eabi-gdb` | Отладка через GDB server |
| `host-gcc` | Сборка и запуск host-тестов |
| `CMake + Ninja` | Система сборки |
| `CTest` | Запуск тестов (Unity + fff) |
| `clangd` | Language server для VSCode |
| `clang-format` | Форматирование кода |
| `clang-tidy` | Статический анализ |
| `cppcheck` | Дополнительный статический анализ |
| `lcov / gcovr` | Покрытие host-тестов |
| `Python 3 + nxp-spsdk` | Прошивка (blhost, nxpimage), производственная провизия |
| `srec_cat` | Манипуляции с бинарными образами |
---
## Конфигурация платы (NXP Config Tools)
Файл `bsp/board/board.mex` — источник истины для конфигурации пинов и тактирования. Открывается в NXP Config Tools (Pins Tool + Clocks Tool) для генерации `pin_mux.c/h` и `clock_config.c/h`. Используется **один раз** при старте проекта или при изменении аппаратной схемы. Коммитится в репозиторий вместе со сгенерированным кодом.
---
## Доставка прошивки
| Сценарий | Способ |
|---|---|
| Разработка (с ПК разработчика) | `rsync` / `scp` → SSH на ПК-сервер → OpenOCD/blhost |
| Производство | `nxp-spsdk` (blhost + nxpimage) через USB ROM |
| Готовое ПО на сервер | GitLab CI/CD → GitLab Package Registry → сервер подтягивает |
| Обновление загрузчика | USB ROM + blhost (только так, не через сам загрузчик) |
| Обновление боевой прошивки в поле | Через загрузчик, схема A/B, uSD |

BIN
docs/schematic.pdf Normal file
View file

Binary file not shown.

11
sdk/CMSIS/Core/Include/CMSIS_Include_core_cm.cmake Normal file
View file

@ -0,0 +1,11 @@
# Add set(CONFIG_USE_CMSIS_Include_core_cm true) in config.cmake to use this component
include_guard(GLOBAL)
message("${CMAKE_CURRENT_LIST_FILE} component is included.")
target_include_directories(${MCUX_SDK_PROJECT_NAME} PUBLIC
${CMAKE_CURRENT_LIST_DIR}/.
${CMAKE_CURRENT_LIST_DIR}/m-profile
)

708
sdk/CMSIS/Core/Include/cmsis_clang.h Normal file
View file

@ -0,0 +1,708 @@
/**************************************************************************//**
* @file cmsis_clang.h
* @brief CMSIS compiler LLVM/Clang header file
* @version V6.0.0
* @date 27. July 2024
******************************************************************************/
/*
* Copyright (c) 2009-2023 Arm Limited. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef __CMSIS_CLANG_H
#define __CMSIS_CLANG_H
#pragma clang system_header /* treat file as system include file */
#if (__ARM_ACLE >= 200)
#include <arm_acle.h>
#else
#error Compiler must support ACLE V2.0
#endif /* (__ARM_ACLE >= 200) */
/* Fallback for __has_builtin */
#ifndef __has_builtin
#define __has_builtin(x) (0)
#endif
/* CMSIS compiler specific defines */
#ifndef __ASM
#define __ASM __asm
#endif
#ifndef __INLINE
#define __INLINE inline
#endif
#ifndef __STATIC_INLINE
#define __STATIC_INLINE static inline
#endif
#ifndef __STATIC_FORCEINLINE
#define __STATIC_FORCEINLINE __attribute__((always_inline)) static inline
#endif
#ifndef __NO_RETURN
#define __NO_RETURN __attribute__((__noreturn__))
#endif
#ifndef CMSIS_DEPRECATED
#define CMSIS_DEPRECATED __attribute__((deprecated))
#endif
#ifndef __USED
#define __USED __attribute__((used))
#endif
#ifndef __WEAK
#define __WEAK __attribute__((weak))
#endif
#ifndef __PACKED
#define __PACKED __attribute__((packed, aligned(1)))
#endif
#ifndef __PACKED_STRUCT
#define __PACKED_STRUCT struct __attribute__((packed, aligned(1)))
#endif
#ifndef __PACKED_UNION
#define __PACKED_UNION union __attribute__((packed, aligned(1)))
#endif
#ifndef __UNALIGNED_UINT16_WRITE
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wpacked"
__PACKED_STRUCT T_UINT16_WRITE { uint16_t v; };
#pragma clang diagnostic pop
#define __UNALIGNED_UINT16_WRITE(addr, val) (void)((((struct T_UINT16_WRITE *)(void *)(addr))->v) = (val))
#endif
#ifndef __UNALIGNED_UINT16_READ
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wpacked"
__PACKED_STRUCT T_UINT16_READ { uint16_t v; };
#pragma clang diagnostic pop
#define __UNALIGNED_UINT16_READ(addr) (((const struct T_UINT16_READ *)(const void *)(addr))->v)
#endif
#ifndef __UNALIGNED_UINT32_WRITE
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wpacked"
__PACKED_STRUCT T_UINT32_WRITE { uint32_t v; };
#pragma clang diagnostic pop
#define __UNALIGNED_UINT32_WRITE(addr, val) (void)((((struct T_UINT32_WRITE *)(void *)(addr))->v) = (val))
#endif
#ifndef __UNALIGNED_UINT32_READ
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wpacked"
__PACKED_STRUCT T_UINT32_READ { uint32_t v; };
#pragma clang diagnostic pop
#define __UNALIGNED_UINT32_READ(addr) (((const struct T_UINT32_READ *)(const void *)(addr))->v)
#endif
#ifndef __ALIGNED
#define __ALIGNED(x) __attribute__((aligned(x)))
#endif
#ifndef __RESTRICT
#define __RESTRICT __restrict
#endif
#ifndef __COMPILER_BARRIER
#define __COMPILER_BARRIER() __ASM volatile("":::"memory")
#endif
#ifndef __NO_INIT
#define __NO_INIT __attribute__ ((section (".noinit")))
#endif
#ifndef __ALIAS
#define __ALIAS(x) __attribute__ ((alias(x)))
#endif
/* ########################## Core Instruction Access ######################### */
/** \defgroup CMSIS_Core_InstructionInterface CMSIS Core Instruction Interface
Access to dedicated instructions
@{
*/
/* Define macros for porting to both thumb1 and thumb2.
* For thumb1, use low register (r0-r7), specified by constraint "l"
* Otherwise, use general registers, specified by constraint "r" */
#if defined (__thumb__) && !defined (__thumb2__)
#define __CMSIS_GCC_OUT_REG(r) "=l" (r)
#define __CMSIS_GCC_RW_REG(r) "+l" (r)
#define __CMSIS_GCC_USE_REG(r) "l" (r)
#else
#define __CMSIS_GCC_OUT_REG(r) "=r" (r)
#define __CMSIS_GCC_RW_REG(r) "+r" (r)
#define __CMSIS_GCC_USE_REG(r) "r" (r)
#endif
/**
\brief No Operation
\details No Operation does nothing. This instruction can be used for code alignment purposes.
*/
#define __NOP() __nop()
/**
\brief Wait For Interrupt
\details Wait For Interrupt is a hint instruction that suspends execution until one of a number of events occurs.
*/
#define __WFI() __wfi()
/**
\brief Wait For Event
\details Wait For Event is a hint instruction that permits the processor to enter
a low-power state until one of a number of events occurs.
*/
#define __WFE() __wfe()
/**
\brief Send Event
\details Send Event is a hint instruction. It causes an event to be signaled to the CPU.
*/
#define __SEV() __sev()
/**
\brief Instruction Synchronization Barrier
\details Instruction Synchronization Barrier flushes the pipeline in the processor,
so that all instructions following the ISB are fetched from cache or memory,
after the instruction has been completed.
*/
#define __ISB() __isb(0xF)
/**
\brief Data Synchronization Barrier
\details Acts as a special kind of Data Memory Barrier.
It completes when all explicit memory accesses before this instruction complete.
*/
#define __DSB() __dsb(0xF)
/**
\brief Data Memory Barrier
\details Ensures the apparent order of the explicit memory operations before
and after the instruction, without ensuring their completion.
*/
#define __DMB() __dmb(0xF)
/**
\brief Reverse byte order (32 bit)
\details Reverses the byte order in unsigned integer value. For example, 0x12345678 becomes 0x78563412.
\param [in] value Value to reverse
\return Reversed value
*/
#define __REV(value) __rev(value)
/**
\brief Reverse byte order (16 bit)
\details Reverses the byte order within each halfword of a word. For example, 0x12345678 becomes 0x34127856.
\param [in] value Value to reverse
\return Reversed value
*/
#define __REV16(value) __rev16(value)
/**
\brief Reverse byte order (16 bit)
\details Reverses the byte order in a 16-bit value and returns the signed 16-bit result. For example, 0x0080 becomes 0x8000.
\param [in] value Value to reverse
\return Reversed value
*/
#define __REVSH(value) __revsh(value)
/**
\brief Rotate Right in unsigned value (32 bit)
\details Rotate Right (immediate) provides the value of the contents of a register rotated by a variable number of bits.
\param [in] op1 Value to rotate
\param [in] op2 Number of Bits to rotate
\return Rotated value
*/
#define __ROR(op1, op2) __ror(op1, op2)
/**
\brief Breakpoint
\details Causes the processor to enter Debug state.
Debug tools can use this to investigate system state when the instruction at a particular address is reached.
\param [in] value is ignored by the processor.
If required, a debugger can use it to store additional information about the breakpoint.
*/
#define __BKPT(value) __ASM volatile ("bkpt "#value)
/**
\brief Reverse bit order of value
\details Reverses the bit order of the given value.
\param [in] value Value to reverse
\return Reversed value
*/
#define __RBIT(value) __rbit(value)
/**
\brief Count leading zeros
\details Counts the number of leading zeros of a data value.
\param [in] value Value to count the leading zeros
\return number of leading zeros in value
*/
#define __CLZ(value) __clz(value)
#if ((__ARM_FEATURE_SAT >= 1) && \
(__ARM_ARCH_ISA_THUMB >= 2) )
/* __ARM_FEATURE_SAT is wrong for Armv8-M Baseline devices */
/**
\brief Signed Saturate
\details Saturates a signed value.
\param [in] value Value to be saturated
\param [in] sat Bit position to saturate to (1..32)
\return Saturated value
*/
#define __SSAT(value, sat) __ssat(value, sat)
/**
\brief Unsigned Saturate
\details Saturates an unsigned value.
\param [in] value Value to be saturated
\param [in] sat Bit position to saturate to (0..31)
\return Saturated value
*/
#define __USAT(value, sat) __usat(value, sat)
#else /* (__ARM_FEATURE_SAT >= 1) */
/**
\brief Signed Saturate
\details Saturates a signed value.
\param [in] value Value to be saturated
\param [in] sat Bit position to saturate to (1..32)
\return Saturated value
*/
__STATIC_FORCEINLINE int32_t __SSAT(int32_t val, uint32_t sat)
{
if ((sat >= 1U) && (sat <= 32U))
{
const int32_t max = (int32_t)((1U << (sat - 1U)) - 1U);
const int32_t min = -1 - max ;
if (val > max)
{
return (max);
}
else if (val < min)
{
return (min);
}
}
return (val);
}
/**
\brief Unsigned Saturate
\details Saturates an unsigned value.
\param [in] value Value to be saturated
\param [in] sat Bit position to saturate to (0..31)
\return Saturated value
*/
__STATIC_FORCEINLINE uint32_t __USAT(int32_t val, uint32_t sat)
{
if (sat <= 31U)
{
const uint32_t max = ((1U << sat) - 1U);
if (val > (int32_t)max)
{
return (max);
}
else if (val < 0)
{
return (0U);
}
}
return ((uint32_t)val);
}
#endif /* (__ARM_FEATURE_SAT >= 1) */
#if (__ARM_FEATURE_LDREX >= 1)
/**
\brief Remove the exclusive lock
\details Removes the exclusive lock which is created by LDREX.
*/
#define __CLREX __builtin_arm_clrex
/**
\brief LDR Exclusive (8 bit)
\details Executes a exclusive LDR instruction for 8 bit value.
\param [in] ptr Pointer to data
\return value of type uint8_t at (*ptr)
*/
#define __LDREXB (uint8_t)__builtin_arm_ldrex
/**
\brief STR Exclusive (8 bit)
\details Executes a exclusive STR instruction for 8 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
\return 0 Function succeeded
\return 1 Function failed
*/
#define __STREXB (uint32_t)__builtin_arm_strex
#endif /* (__ARM_FEATURE_LDREX >= 1) */
#if (__ARM_FEATURE_LDREX >= 2)
/**
\brief LDR Exclusive (16 bit)
\details Executes a exclusive LDR instruction for 16 bit values.
\param [in] ptr Pointer to data
\return value of type uint16_t at (*ptr)
*/
#define __LDREXH (uint16_t)__builtin_arm_ldrex
/**
\brief STR Exclusive (16 bit)
\details Executes a exclusive STR instruction for 16 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
\return 0 Function succeeded
\return 1 Function failed
*/
#define __STREXH (uint32_t)__builtin_arm_strex
#endif /* (__ARM_FEATURE_LDREX >= 2) */
#if (__ARM_FEATURE_LDREX >= 4)
/**
\brief LDR Exclusive (32 bit)
\details Executes a exclusive LDR instruction for 32 bit values.
\param [in] ptr Pointer to data
\return value of type uint32_t at (*ptr)
*/
#define __LDREXW (uint32_t)__builtin_arm_ldrex
/**
\brief STR Exclusive (32 bit)
\details Executes a exclusive STR instruction for 32 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
\return 0 Function succeeded
\return 1 Function failed
*/
#define __STREXW (uint32_t)__builtin_arm_strex
#endif /* (__ARM_FEATURE_LDREX >= 4) */
#if (__ARM_ARCH_ISA_THUMB >= 2)
/**
\brief Rotate Right with Extend (32 bit)
\details Moves each bit of a bitstring right by one bit.
The carry input is shifted in at the left end of the bitstring.
\param [in] value Value to rotate
\return Rotated value
*/
__STATIC_FORCEINLINE uint32_t __RRX(uint32_t value)
{
uint32_t result;
__ASM volatile ("rrx %0, %1" : "=r" (result) : "r" (value));
return (result);
}
/**
\brief LDRT Unprivileged (8 bit)
\details Executes a Unprivileged LDRT instruction for 8 bit value.
\param [in] ptr Pointer to data
\return value of type uint8_t at (*ptr)
*/
__STATIC_FORCEINLINE uint8_t __LDRBT(volatile uint8_t *ptr)
{
uint32_t result;
__ASM volatile ("ldrbt %0, %1" : "=r" (result) : "Q" (*ptr) );
return ((uint8_t)result); /* Add explicit type cast here */
}
/**
\brief LDRT Unprivileged (16 bit)
\details Executes a Unprivileged LDRT instruction for 16 bit values.
\param [in] ptr Pointer to data
\return value of type uint16_t at (*ptr)
*/
__STATIC_FORCEINLINE uint16_t __LDRHT(volatile uint16_t *ptr)
{
uint32_t result;
__ASM volatile ("ldrht %0, %1" : "=r" (result) : "Q" (*ptr) );
return ((uint16_t)result); /* Add explicit type cast here */
}
/**
\brief LDRT Unprivileged (32 bit)
\details Executes a Unprivileged LDRT instruction for 32 bit values.
\param [in] ptr Pointer to data
\return value of type uint32_t at (*ptr)
*/
__STATIC_FORCEINLINE uint32_t __LDRT(volatile uint32_t *ptr)
{
uint32_t result;
__ASM volatile ("ldrt %0, %1" : "=r" (result) : "Q" (*ptr) );
return (result);
}
#endif /* (__ARM_ARCH_ISA_THUMB >= 2) */
#if (__ARM_ARCH >= 8)
/**
\brief Load-Acquire (8 bit)
\details Executes a LDAB instruction for 8 bit value.
\param [in] ptr Pointer to data
\return value of type uint8_t at (*ptr)
*/
__STATIC_FORCEINLINE uint8_t __LDAB(volatile uint8_t *ptr)
{
uint32_t result;
__ASM volatile ("ldab %0, %1" : "=r" (result) : "Q" (*ptr) : "memory" );
return ((uint8_t)result); /* Add explicit type cast here */
}
/**
\brief Load-Acquire (16 bit)
\details Executes a LDAH instruction for 16 bit values.
\param [in] ptr Pointer to data
\return value of type uint16_t at (*ptr)
*/
__STATIC_FORCEINLINE uint16_t __LDAH(volatile uint16_t *ptr)
{
uint32_t result;
__ASM volatile ("ldah %0, %1" : "=r" (result) : "Q" (*ptr) : "memory" );
return ((uint16_t)result); /* Add explicit type cast here */
}
/**
\brief Load-Acquire (32 bit)
\details Executes a LDA instruction for 32 bit values.
\param [in] ptr Pointer to data
\return value of type uint32_t at (*ptr)
*/
__STATIC_FORCEINLINE uint32_t __LDA(volatile uint32_t *ptr)
{
uint32_t result;
__ASM volatile ("lda %0, %1" : "=r" (result) : "Q" (*ptr) : "memory" );
return (result);
}
/**
\brief Store-Release (8 bit)
\details Executes a STLB instruction for 8 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
*/
__STATIC_FORCEINLINE void __STLB(uint8_t value, volatile uint8_t *ptr)
{
__ASM volatile ("stlb %1, %0" : "=Q" (*ptr) : "r" ((uint32_t)value) : "memory" );
}
/**
\brief Store-Release (16 bit)
\details Executes a STLH instruction for 16 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
*/
__STATIC_FORCEINLINE void __STLH(uint16_t value, volatile uint16_t *ptr)
{
__ASM volatile ("stlh %1, %0" : "=Q" (*ptr) : "r" ((uint32_t)value) : "memory" );
}
/**
\brief Store-Release (32 bit)
\details Executes a STL instruction for 32 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
*/
__STATIC_FORCEINLINE void __STL(uint32_t value, volatile uint32_t *ptr)
{
__ASM volatile ("stl %1, %0" : "=Q" (*ptr) : "r" ((uint32_t)value) : "memory" );
}
/**
\brief Load-Acquire Exclusive (8 bit)
\details Executes a LDAB exclusive instruction for 8 bit value.
\param [in] ptr Pointer to data
\return value of type uint8_t at (*ptr)
*/
#define __LDAEXB (uint8_t)__builtin_arm_ldaex
/**
\brief Load-Acquire Exclusive (16 bit)
\details Executes a LDAH exclusive instruction for 16 bit values.
\param [in] ptr Pointer to data
\return value of type uint16_t at (*ptr)
*/
#define __LDAEXH (uint16_t)__builtin_arm_ldaex
/**
\brief Load-Acquire Exclusive (32 bit)
\details Executes a LDA exclusive instruction for 32 bit values.
\param [in] ptr Pointer to data
\return value of type uint32_t at (*ptr)
*/
#define __LDAEX (uint32_t)__builtin_arm_ldaex
/**
\brief Store-Release Exclusive (8 bit)
\details Executes a STLB exclusive instruction for 8 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
\return 0 Function succeeded
\return 1 Function failed
*/
#define __STLEXB (uint32_t)__builtin_arm_stlex
/**
\brief Store-Release Exclusive (16 bit)
\details Executes a STLH exclusive instruction for 16 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
\return 0 Function succeeded
\return 1 Function failed
*/
#define __STLEXH (uint32_t)__builtin_arm_stlex
/**
\brief Store-Release Exclusive (32 bit)
\details Executes a STL exclusive instruction for 32 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
\return 0 Function succeeded
\return 1 Function failed
*/
#define __STLEX (uint32_t)__builtin_arm_stlex
#endif /* (__ARM_ARCH >= 8) */
/** @}*/ /* end of group CMSIS_Core_InstructionInterface */
/* ########################### Core Function Access ########################### */
/** \ingroup CMSIS_Core_FunctionInterface
\defgroup CMSIS_Core_RegAccFunctions CMSIS Core Register Access Functions
@{
*/
/**
\brief Enable IRQ Interrupts
\details Enables IRQ interrupts by clearing special-purpose register PRIMASK.
Can only be executed in Privileged modes.
*/
__STATIC_FORCEINLINE void __enable_irq(void)
{
__ASM volatile ("cpsie i" : : : "memory");
}
/**
\brief Disable IRQ Interrupts
\details Disables IRQ interrupts by setting special-purpose register PRIMASK.
Can only be executed in Privileged modes.
*/
__STATIC_FORCEINLINE void __disable_irq(void)
{
__ASM volatile ("cpsid i" : : : "memory");
}
#if (__ARM_ARCH_ISA_THUMB >= 2)
/**
\brief Enable FIQ
\details Enables FIQ interrupts by clearing special-purpose register FAULTMASK.
Can only be executed in Privileged modes.
*/
__STATIC_FORCEINLINE void __enable_fault_irq(void)
{
__ASM volatile ("cpsie f" : : : "memory");
}
/**
\brief Disable FIQ
\details Disables FIQ interrupts by setting special-purpose register FAULTMASK.
Can only be executed in Privileged modes.
*/
__STATIC_FORCEINLINE void __disable_fault_irq(void)
{
__ASM volatile ("cpsid f" : : : "memory");
}
#endif
/**
\brief Get FPSCR
\details Returns the current value of the Floating Point Status/Control register.
\return Floating Point Status/Control register value
*/
__STATIC_FORCEINLINE uint32_t __get_FPSCR(void)
{
#if (defined(__ARM_FP) && (__ARM_FP >= 1))
return (__builtin_arm_get_fpscr());
#else
return (0U);
#endif
}
/**
\brief Set FPSCR
\details Assigns the given value to the Floating Point Status/Control register.
\param [in] fpscr Floating Point Status/Control value to set
*/
__STATIC_FORCEINLINE void __set_FPSCR(uint32_t fpscr)
{
#if (defined(__ARM_FP) && (__ARM_FP >= 1))
__builtin_arm_set_fpscr(fpscr);
#else
(void)fpscr;
#endif
}
/** @} end of CMSIS_Core_RegAccFunctions */
// Include the profile specific settings:
#if __ARM_ARCH_PROFILE == 'A'
#include "./a-profile/cmsis_clang_a.h"
#elif __ARM_ARCH_PROFILE == 'R'
#include "./r-profile/cmsis_clang_r.h"
#elif __ARM_ARCH_PROFILE == 'M'
#include "./m-profile/cmsis_clang_m.h"
#else
#error "Unknown Arm architecture profile"
#endif
#endif /* __CMSIS_CLANG_H */

292
sdk/CMSIS/Core/Include/cmsis_compiler.h Normal file
View file

@ -0,0 +1,292 @@
/*
* Copyright (c) 2009-2023 Arm Limited. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* CMSIS Compiler Generic Header File
*/
#ifndef __CMSIS_COMPILER_H
#define __CMSIS_COMPILER_H
#include <stdint.h>
/*
* Arm Compiler above 6.10.1 (armclang)
*/
#if defined (__ARMCC_VERSION) && (__ARMCC_VERSION >= 6100100)
#include "cmsis_armclang.h"
/*
* TI Arm Clang Compiler (tiarmclang)
*/
#elif defined (__ti__)
#include "cmsis_tiarmclang.h"
/*
* LLVM/Clang Compiler
*/
#elif defined ( __clang__ )
#include "cmsis_clang.h"
/*
* GNU Compiler
*/
#elif defined ( __GNUC__ )
#include "cmsis_gcc.h"
/*
* IAR Compiler
*/
#elif defined ( __ICCARM__ )
#if __ARM_ARCH_PROFILE == 'A'
#include "a-profile/cmsis_iccarm_a.h"
#elif __ARM_ARCH_PROFILE == 'R'
#include "r-profile/cmsis_iccarm_r.h"
#elif __ARM_ARCH_PROFILE == 'M'
#include "m-profile/cmsis_iccarm_m.h"
#else
#error "Unknown Arm architecture profile"
#endif
/*
* TI Arm Compiler (armcl)
*/
#elif defined ( __TI_ARM__ )
#include <cmsis_ccs.h>
#ifndef __ASM
#define __ASM __asm
#endif
#ifndef __INLINE
#define __INLINE inline
#endif
#ifndef __STATIC_INLINE
#define __STATIC_INLINE static inline
#endif
#ifndef __STATIC_FORCEINLINE
#define __STATIC_FORCEINLINE __STATIC_INLINE
#endif
#ifndef __NO_RETURN
#define __NO_RETURN __attribute__((noreturn))
#endif
#ifndef __USED
#define __USED __attribute__((used))
#endif
#ifndef __WEAK
#define __WEAK __attribute__((weak))
#endif
#ifndef __PACKED
#define __PACKED __attribute__((packed))
#endif
#ifndef __PACKED_STRUCT
#define __PACKED_STRUCT struct __attribute__((packed))
#endif
#ifndef __PACKED_UNION
#define __PACKED_UNION union __attribute__((packed))
#endif
#ifndef __UNALIGNED_UINT16_WRITE
__PACKED_STRUCT T_UINT16_WRITE { uint16_t v; };
#define __UNALIGNED_UINT16_WRITE(addr, val) (void)((((struct T_UINT16_WRITE *)(void*)(addr))->v) = (val))
#endif
#ifndef __UNALIGNED_UINT16_READ
__PACKED_STRUCT T_UINT16_READ { uint16_t v; };
#define __UNALIGNED_UINT16_READ(addr) (((const struct T_UINT16_READ *)(const void *)(addr))->v)
#endif
#ifndef __UNALIGNED_UINT32_WRITE
__PACKED_STRUCT T_UINT32_WRITE { uint32_t v; };
#define __UNALIGNED_UINT32_WRITE(addr, val) (void)((((struct T_UINT32_WRITE *)(void *)(addr))->v) = (val))
#endif
#ifndef __UNALIGNED_UINT32_READ
__PACKED_STRUCT T_UINT32_READ { uint32_t v; };
#define __UNALIGNED_UINT32_READ(addr) (((const struct T_UINT32_READ *)(const void *)(addr))->v)
#endif
#ifndef __ALIGNED
#define __ALIGNED(x) __attribute__((aligned(x)))
#endif
#ifndef __RESTRICT
#define __RESTRICT __restrict
#endif
#ifndef __COMPILER_BARRIER
#warning No compiler specific solution for __COMPILER_BARRIER. __COMPILER_BARRIER is ignored.
#define __COMPILER_BARRIER() (void)0
#endif
#ifndef __NO_INIT
#define __NO_INIT __attribute__ ((section (".noinit")))
#endif
#ifndef __ALIAS
#define __ALIAS(x) __attribute__ ((alias(x)))
#endif
/*
* TASKING Compiler
*/
#elif defined ( __TASKING__ )
/*
* The CMSIS functions have been implemented as intrinsics in the compiler.
* Please use "carm -?i" to get an up to date list of all intrinsics,
* Including the CMSIS ones.
*/
#ifndef __ASM
#define __ASM __asm
#endif
#ifndef __INLINE
#define __INLINE inline
#endif
#ifndef __STATIC_INLINE
#define __STATIC_INLINE static inline
#endif
#ifndef __STATIC_FORCEINLINE
#define __STATIC_FORCEINLINE __STATIC_INLINE
#endif
#ifndef __NO_RETURN
#define __NO_RETURN __attribute__((noreturn))
#endif
#ifndef __USED
#define __USED __attribute__((used))
#endif
#ifndef __WEAK
#define __WEAK __attribute__((weak))
#endif
#ifndef __PACKED
#define __PACKED __packed__
#endif
#ifndef __PACKED_STRUCT
#define __PACKED_STRUCT struct __packed__
#endif
#ifndef __PACKED_UNION
#define __PACKED_UNION union __packed__
#endif
#ifndef __UNALIGNED_UINT16_WRITE
__PACKED_STRUCT T_UINT16_WRITE { uint16_t v; };
#define __UNALIGNED_UINT16_WRITE(addr, val) (void)((((struct T_UINT16_WRITE *)(void *)(addr))->v) = (val))
#endif
#ifndef __UNALIGNED_UINT16_READ
__PACKED_STRUCT T_UINT16_READ { uint16_t v; };
#define __UNALIGNED_UINT16_READ(addr) (((const struct T_UINT16_READ *)(const void *)(addr))->v)
#endif
#ifndef __UNALIGNED_UINT32_WRITE
__PACKED_STRUCT T_UINT32_WRITE { uint32_t v; };
#define __UNALIGNED_UINT32_WRITE(addr, val) (void)((((struct T_UINT32_WRITE *)(void *)(addr))->v) = (val))
#endif
#ifndef __UNALIGNED_UINT32_READ
__PACKED_STRUCT T_UINT32_READ { uint32_t v; };
#define __UNALIGNED_UINT32_READ(addr) (((const struct T_UINT32_READ *)(const void *)(addr))->v)
#endif
#ifndef __ALIGNED
#define __ALIGNED(x) __align(x)
#endif
#ifndef __RESTRICT
#warning No compiler specific solution for __RESTRICT. __RESTRICT is ignored.
#define __RESTRICT
#endif
#ifndef __COMPILER_BARRIER
#warning No compiler specific solution for __COMPILER_BARRIER. __COMPILER_BARRIER is ignored.
#define __COMPILER_BARRIER() (void)0
#endif
#ifndef __NO_INIT
#define __NO_INIT __attribute__ ((section (".noinit")))
#endif
#ifndef __ALIAS
#define __ALIAS(x) __attribute__ ((alias(x)))
#endif
/*
* COSMIC Compiler
*/
#elif defined ( __CSMC__ )
#include <cmsis_csm.h>
#ifndef __ASM
#define __ASM _asm
#endif
#ifndef __INLINE
#define __INLINE inline
#endif
#ifndef __STATIC_INLINE
#define __STATIC_INLINE static inline
#endif
#ifndef __STATIC_FORCEINLINE
#define __STATIC_FORCEINLINE __STATIC_INLINE
#endif
#ifndef __NO_RETURN
// NO RETURN is automatically detected hence no warning here
#define __NO_RETURN
#endif
#ifndef __USED
#warning No compiler specific solution for __USED. __USED is ignored.
#define __USED
#endif
#ifndef __WEAK
#define __WEAK __weak
#endif
#ifndef __PACKED
#define __PACKED @packed
#endif
#ifndef __PACKED_STRUCT
#define __PACKED_STRUCT @packed struct
#endif
#ifndef __PACKED_UNION
#define __PACKED_UNION @packed union
#endif
#ifndef __UNALIGNED_UINT16_WRITE
__PACKED_STRUCT T_UINT16_WRITE { uint16_t v; };
#define __UNALIGNED_UINT16_WRITE(addr, val) (void)((((struct T_UINT16_WRITE *)(void *)(addr))->v) = (val))
#endif
#ifndef __UNALIGNED_UINT16_READ
__PACKED_STRUCT T_UINT16_READ { uint16_t v; };
#define __UNALIGNED_UINT16_READ(addr) (((const struct T_UINT16_READ *)(const void *)(addr))->v)
#endif
#ifndef __UNALIGNED_UINT32_WRITE
__PACKED_STRUCT T_UINT32_WRITE { uint32_t v; };
#define __UNALIGNED_UINT32_WRITE(addr, val) (void)((((struct T_UINT32_WRITE *)(void *)(addr))->v) = (val))
#endif
#ifndef __UNALIGNED_UINT32_READ
__PACKED_STRUCT T_UINT32_READ { uint32_t v; };
#define __UNALIGNED_UINT32_READ(addr) (((const struct T_UINT32_READ *)(const void *)(addr))->v)
#endif
#ifndef __ALIGNED
#warning No compiler specific solution for __ALIGNED. __ALIGNED is ignored.
#define __ALIGNED(x)
#endif
#ifndef __RESTRICT
#warning No compiler specific solution for __RESTRICT. __RESTRICT is ignored.
#define __RESTRICT
#endif
#ifndef __COMPILER_BARRIER
#warning No compiler specific solution for __COMPILER_BARRIER. __COMPILER_BARRIER is ignored.
#define __COMPILER_BARRIER() (void)0
#endif
#ifndef __NO_INIT
#define __NO_INIT __attribute__ ((section (".noinit")))
#endif
#ifndef __ALIAS
#define __ALIAS(x) __attribute__ ((alias(x)))
#endif
#else
#error Unknown compiler.
#endif
#endif /* __CMSIS_COMPILER_H */

1006
sdk/CMSIS/Core/Include/cmsis_gcc.h Normal file
View file

File diff suppressed because it is too large Load diff

44
sdk/CMSIS/Core/Include/cmsis_version.h Normal file
View file

@ -0,0 +1,44 @@
/*
* Copyright (c) 2009-2023 ARM Limited. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* CMSIS Core Version Definitions
*/
#if defined ( __ICCARM__ )
#pragma system_include /* treat file as system include file for MISRA check */
#elif defined (__clang__)
#pragma clang system_header /* treat file as system include file */
#endif
#ifndef __CMSIS_VERSION_H
#define __CMSIS_VERSION_H
/* CMSIS-Core(M) Version definitions */
#define __CM_CMSIS_VERSION_MAIN ( 6U) /*!< \brief [31:16] CMSIS-Core(M) main version */
#define __CM_CMSIS_VERSION_SUB ( 1U) /*!< \brief [15:0] CMSIS-Core(M) sub version */
#define __CM_CMSIS_VERSION ((__CM_CMSIS_VERSION_MAIN << 16U) | \
__CM_CMSIS_VERSION_SUB ) /*!< \brief CMSIS Core(M) version number */
/* CMSIS-Core(A) Version definitions */
#define __CA_CMSIS_VERSION_MAIN ( 6U) /*!< \brief [31:16] CMSIS-Core(A) main version */
#define __CA_CMSIS_VERSION_SUB ( 1U) /*!< \brief [15:0] CMSIS-Core(A) sub version */
#define __CA_CMSIS_VERSION ((__CA_CMSIS_VERSION_MAIN << 16U) | \
__CA_CMSIS_VERSION_SUB ) /*!< \brief CMSIS-Core(A) version number */
#endif

2468
sdk/CMSIS/Core/Include/core_cm7.h Normal file
View file

File diff suppressed because it is too large Load diff

439
sdk/CMSIS/Core/Include/m-profile/armv7m_cachel1.h Normal file
View file

@ -0,0 +1,439 @@
/*
* Copyright (c) 2020-2021 Arm Limited. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* CMSIS-Core(M) Level 1 Cache API for Armv7-M and later
*/
#ifndef ARM_ARMV7M_CACHEL1_H
#define ARM_ARMV7M_CACHEL1_H
#if defined ( __ICCARM__ )
#pragma system_include /* treat file as system include file for MISRA check */
#elif defined (__clang__)
#pragma clang system_header /* treat file as system include file */
#endif
/**
\ingroup CMSIS_Core_FunctionInterface
\defgroup CMSIS_Core_CacheFunctions Cache Functions
\brief Functions that configure Instruction and Data cache.
@{
*/
/* Cache Size ID Register Macros */
#define CCSIDR_WAYS(x) (((x) & SCB_CCSIDR_ASSOCIATIVITY_Msk) >> SCB_CCSIDR_ASSOCIATIVITY_Pos)
#define CCSIDR_SETS(x) (((x) & SCB_CCSIDR_NUMSETS_Msk ) >> SCB_CCSIDR_NUMSETS_Pos )
#ifndef __SCB_DCACHE_LINE_SIZE
#define __SCB_DCACHE_LINE_SIZE 32U /*!< Cortex-M7 cache line size is fixed to 32 bytes (8 words). See also register SCB_CCSIDR */
#endif
#ifndef __SCB_ICACHE_LINE_SIZE
#define __SCB_ICACHE_LINE_SIZE 32U /*!< Cortex-M7 cache line size is fixed to 32 bytes (8 words). See also register SCB_CCSIDR */
#endif
/**
\brief Enable I-Cache
\details Turns on I-Cache
*/
__STATIC_FORCEINLINE void SCB_EnableICache (void)
{
#if defined (__ICACHE_PRESENT) && (__ICACHE_PRESENT == 1U)
if (SCB->CCR & SCB_CCR_IC_Msk) return; /* return if ICache is already enabled */
__DSB();
__ISB();
SCB->ICIALLU = 0UL; /* invalidate I-Cache */
__DSB();
__ISB();
SCB->CCR |= (uint32_t)SCB_CCR_IC_Msk; /* enable I-Cache */
__DSB();
__ISB();
#endif
}
/**
\brief Disable I-Cache
\details Turns off I-Cache
*/
__STATIC_FORCEINLINE void SCB_DisableICache (void)
{
#if defined (__ICACHE_PRESENT) && (__ICACHE_PRESENT == 1U)
__DSB();
__ISB();
SCB->CCR &= ~(uint32_t)SCB_CCR_IC_Msk; /* disable I-Cache */
SCB->ICIALLU = 0UL; /* invalidate I-Cache */
__DSB();
__ISB();
#endif
}
/**
\brief Invalidate I-Cache
\details Invalidates I-Cache
*/
__STATIC_FORCEINLINE void SCB_InvalidateICache (void)
{
#if defined (__ICACHE_PRESENT) && (__ICACHE_PRESENT == 1U)
__DSB();
__ISB();
SCB->ICIALLU = 0UL;
__DSB();
__ISB();
#endif
}
/**
\brief I-Cache Invalidate by address
\details Invalidates I-Cache for the given address.
I-Cache is invalidated starting from a 32 byte aligned address in 32 byte granularity.
I-Cache memory blocks which are part of given address + given size are invalidated.
\param[in] addr address
\param[in] isize size of memory block (in number of bytes)
*/
__STATIC_FORCEINLINE void SCB_InvalidateICache_by_Addr (volatile void *addr, int32_t isize)
{
#if defined (__ICACHE_PRESENT) && (__ICACHE_PRESENT == 1U)
if ( isize > 0 ) {
int32_t op_size = isize + (((uint32_t)addr) & (__SCB_ICACHE_LINE_SIZE - 1U));
uint32_t op_addr = (uint32_t)addr /* & ~(__SCB_ICACHE_LINE_SIZE - 1U) */;
__DSB();
do {
SCB->ICIMVAU = op_addr; /* register accepts only 32byte aligned values, only bits 31..5 are valid */
op_addr += __SCB_ICACHE_LINE_SIZE;
op_size -= __SCB_ICACHE_LINE_SIZE;
} while ( op_size > 0 );
__DSB();
__ISB();
}
#endif
}
/**
\brief Enable D-Cache
\details Turns on D-Cache
*/
__STATIC_FORCEINLINE void SCB_EnableDCache (void)
{
#if defined (__DCACHE_PRESENT) && (__DCACHE_PRESENT == 1U)
uint32_t ccsidr;
uint32_t sets;
uint32_t ways;
if (SCB->CCR & SCB_CCR_DC_Msk) return; /* return if DCache is already enabled */
SCB->CSSELR = 0U; /* select Level 1 data cache */
__DSB();
ccsidr = SCB->CCSIDR;
/* invalidate D-Cache */
sets = (uint32_t)(CCSIDR_SETS(ccsidr));
do {
ways = (uint32_t)(CCSIDR_WAYS(ccsidr));
do {
SCB->DCISW = (((sets << SCB_DCISW_SET_Pos) & SCB_DCISW_SET_Msk) |
((ways << SCB_DCISW_WAY_Pos) & SCB_DCISW_WAY_Msk) );
#if defined ( __CC_ARM )
__schedule_barrier();
#endif
} while (ways-- != 0U);
} while(sets-- != 0U);
__DSB();
SCB->CCR |= (uint32_t)SCB_CCR_DC_Msk; /* enable D-Cache */
__DSB();
__ISB();
#endif
}
/**
\brief Disable D-Cache
\details Turns off D-Cache
*/
__STATIC_FORCEINLINE void SCB_DisableDCache (void)
{
#if defined (__DCACHE_PRESENT) && (__DCACHE_PRESENT == 1U)
struct {
uint32_t ccsidr;
uint32_t sets;
uint32_t ways;
} locals
#if ((defined(__GNUC__) || defined(__clang__)) && !defined(__OPTIMIZE__))
__ALIGNED(__SCB_DCACHE_LINE_SIZE)
#endif
;
SCB->CSSELR = 0U; /* select Level 1 data cache */
__DSB();
SCB->CCR &= ~(uint32_t)SCB_CCR_DC_Msk; /* disable D-Cache */
__DSB();
#if !defined(__OPTIMIZE__)
/*
* For the endless loop issue with no optimization builds.
* More details, see https://github.com/ARM-software/CMSIS_5/issues/620
*
* The issue only happens when local variables are in stack. If
* local variables are saved in general purpose register, then the function
* is OK.
*
* When local variables are in stack, after disabling the cache, flush the
* local variables cache line for data consistency.
*/
/* Clean and invalidate the local variable cache. */
#if defined(__ICCARM__)
/* As we can't align the stack to the cache line size, invalidate each of the variables */
SCB->DCCIMVAC = (uint32_t)&locals.sets;
SCB->DCCIMVAC = (uint32_t)&locals.ways;
SCB->DCCIMVAC = (uint32_t)&locals.ccsidr;
#else
SCB->DCCIMVAC = (uint32_t)&locals;
#endif
__DSB();
__ISB();
#endif
locals.ccsidr = SCB->CCSIDR;
/* clean & invalidate D-Cache */
locals.sets = (uint32_t)(CCSIDR_SETS(locals.ccsidr));
do {
locals.ways = (uint32_t)(CCSIDR_WAYS(locals.ccsidr));
do {
SCB->DCCISW = (((locals.sets << SCB_DCCISW_SET_Pos) & SCB_DCCISW_SET_Msk) |
((locals.ways << SCB_DCCISW_WAY_Pos) & SCB_DCCISW_WAY_Msk) );
#if defined ( __CC_ARM )
__schedule_barrier();
#endif
} while (locals.ways-- != 0U);
} while(locals.sets-- != 0U);
__DSB();
__ISB();
#endif
}
/**
\brief Invalidate D-Cache
\details Invalidates D-Cache
*/
__STATIC_FORCEINLINE void SCB_InvalidateDCache (void)
{
#if defined (__DCACHE_PRESENT) && (__DCACHE_PRESENT == 1U)
uint32_t ccsidr;
uint32_t sets;
uint32_t ways;
SCB->CSSELR = 0U; /* select Level 1 data cache */
__DSB();
ccsidr = SCB->CCSIDR;
/* invalidate D-Cache */
sets = (uint32_t)(CCSIDR_SETS(ccsidr));
do {
ways = (uint32_t)(CCSIDR_WAYS(ccsidr));
do {
SCB->DCISW = (((sets << SCB_DCISW_SET_Pos) & SCB_DCISW_SET_Msk) |
((ways << SCB_DCISW_WAY_Pos) & SCB_DCISW_WAY_Msk) );
#if defined ( __CC_ARM )
__schedule_barrier();
#endif
} while (ways-- != 0U);
} while(sets-- != 0U);
__DSB();
__ISB();
#endif
}
/**
\brief Clean D-Cache
\details Cleans D-Cache
*/
__STATIC_FORCEINLINE void SCB_CleanDCache (void)
{
#if defined (__DCACHE_PRESENT) && (__DCACHE_PRESENT == 1U)
uint32_t ccsidr;
uint32_t sets;
uint32_t ways;
SCB->CSSELR = 0U; /* select Level 1 data cache */
__DSB();
ccsidr = SCB->CCSIDR;
/* clean D-Cache */
sets = (uint32_t)(CCSIDR_SETS(ccsidr));
do {
ways = (uint32_t)(CCSIDR_WAYS(ccsidr));
do {
SCB->DCCSW = (((sets << SCB_DCCSW_SET_Pos) & SCB_DCCSW_SET_Msk) |
((ways << SCB_DCCSW_WAY_Pos) & SCB_DCCSW_WAY_Msk) );
#if defined ( __CC_ARM )
__schedule_barrier();
#endif
} while (ways-- != 0U);
} while(sets-- != 0U);
__DSB();
__ISB();
#endif
}
/**
\brief Clean & Invalidate D-Cache
\details Cleans and Invalidates D-Cache
*/
__STATIC_FORCEINLINE void SCB_CleanInvalidateDCache (void)
{
#if defined (__DCACHE_PRESENT) && (__DCACHE_PRESENT == 1U)
uint32_t ccsidr;
uint32_t sets;
uint32_t ways;
SCB->CSSELR = 0U; /* select Level 1 data cache */
__DSB();
ccsidr = SCB->CCSIDR;
/* clean & invalidate D-Cache */
sets = (uint32_t)(CCSIDR_SETS(ccsidr));
do {
ways = (uint32_t)(CCSIDR_WAYS(ccsidr));
do {
SCB->DCCISW = (((sets << SCB_DCCISW_SET_Pos) & SCB_DCCISW_SET_Msk) |
((ways << SCB_DCCISW_WAY_Pos) & SCB_DCCISW_WAY_Msk) );
#if defined ( __CC_ARM )
__schedule_barrier();
#endif
} while (ways-- != 0U);
} while(sets-- != 0U);
__DSB();
__ISB();
#endif
}
/**
\brief D-Cache Invalidate by address
\details Invalidates D-Cache for the given address.
D-Cache is invalidated starting from a 32 byte aligned address in 32 byte granularity.
D-Cache memory blocks which are part of given address + given size are invalidated.
\param[in] addr address
\param[in] dsize size of memory block (in number of bytes)
*/
__STATIC_FORCEINLINE void SCB_InvalidateDCache_by_Addr (volatile void *addr, int32_t dsize)
{
#if defined (__DCACHE_PRESENT) && (__DCACHE_PRESENT == 1U)
if ( dsize > 0 ) {
int32_t op_size = dsize + (((uint32_t)addr) & (__SCB_DCACHE_LINE_SIZE - 1U));
uint32_t op_addr = (uint32_t)addr /* & ~(__SCB_DCACHE_LINE_SIZE - 1U) */;
__DSB();
do {
SCB->DCIMVAC = op_addr; /* register accepts only 32byte aligned values, only bits 31..5 are valid */
op_addr += __SCB_DCACHE_LINE_SIZE;
op_size -= __SCB_DCACHE_LINE_SIZE;
} while ( op_size > 0 );
__DSB();
__ISB();
}
#endif
}
/**
\brief D-Cache Clean by address
\details Cleans D-Cache for the given address
D-Cache is cleaned starting from a 32 byte aligned address in 32 byte granularity.
D-Cache memory blocks which are part of given address + given size are cleaned.
\param[in] addr address
\param[in] dsize size of memory block (in number of bytes)
*/
__STATIC_FORCEINLINE void SCB_CleanDCache_by_Addr (volatile void *addr, int32_t dsize)
{
#if defined (__DCACHE_PRESENT) && (__DCACHE_PRESENT == 1U)
if ( dsize > 0 ) {
int32_t op_size = dsize + (((uint32_t)addr) & (__SCB_DCACHE_LINE_SIZE - 1U));
uint32_t op_addr = (uint32_t)addr /* & ~(__SCB_DCACHE_LINE_SIZE - 1U) */;
__DSB();
do {
SCB->DCCMVAC = op_addr; /* register accepts only 32byte aligned values, only bits 31..5 are valid */
op_addr += __SCB_DCACHE_LINE_SIZE;
op_size -= __SCB_DCACHE_LINE_SIZE;
} while ( op_size > 0 );
__DSB();
__ISB();
}
#endif
}
/**
\brief D-Cache Clean and Invalidate by address
\details Cleans and invalidates D_Cache for the given address
D-Cache is cleaned and invalidated starting from a 32 byte aligned address in 32 byte granularity.
D-Cache memory blocks which are part of given address + given size are cleaned and invalidated.
\param[in] addr address (aligned to 32-byte boundary)
\param[in] dsize size of memory block (in number of bytes)
*/
__STATIC_FORCEINLINE void SCB_CleanInvalidateDCache_by_Addr (volatile void *addr, int32_t dsize)
{
#if defined (__DCACHE_PRESENT) && (__DCACHE_PRESENT == 1U)
if ( dsize > 0 ) {
int32_t op_size = dsize + (((uint32_t)addr) & (__SCB_DCACHE_LINE_SIZE - 1U));
uint32_t op_addr = (uint32_t)addr /* & ~(__SCB_DCACHE_LINE_SIZE - 1U) */;
__DSB();
do {
SCB->DCCIMVAC = op_addr; /* register accepts only 32byte aligned values, only bits 31..5 are valid */
op_addr += __SCB_DCACHE_LINE_SIZE;
op_size -= __SCB_DCACHE_LINE_SIZE;
} while ( op_size > 0 );
__DSB();
__ISB();
}
#endif
}
/*@} end of CMSIS_Core_CacheFunctions */
#endif /* ARM_ARMV7M_CACHEL1_H */

273
sdk/CMSIS/Core/Include/m-profile/armv7m_mpu.h Normal file
View file

@ -0,0 +1,273 @@
/*
* Copyright (c) 2017-2020 Arm Limited. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* CMSIS-Core(M) MPU API for Armv7-M MPU
*/
#ifndef ARM_MPU_ARMV7_H
#define ARM_MPU_ARMV7_H
#if defined ( __ICCARM__ )
#pragma system_include /* treat file as system include file for MISRA check */
#elif defined (__clang__)
#pragma clang system_header /* treat file as system include file */
#endif
#define ARM_MPU_REGION_SIZE_32B ((uint8_t)0x04U) ///!< MPU Region Size 32 Bytes
#define ARM_MPU_REGION_SIZE_64B ((uint8_t)0x05U) ///!< MPU Region Size 64 Bytes
#define ARM_MPU_REGION_SIZE_128B ((uint8_t)0x06U) ///!< MPU Region Size 128 Bytes
#define ARM_MPU_REGION_SIZE_256B ((uint8_t)0x07U) ///!< MPU Region Size 256 Bytes
#define ARM_MPU_REGION_SIZE_512B ((uint8_t)0x08U) ///!< MPU Region Size 512 Bytes
#define ARM_MPU_REGION_SIZE_1KB ((uint8_t)0x09U) ///!< MPU Region Size 1 KByte
#define ARM_MPU_REGION_SIZE_2KB ((uint8_t)0x0AU) ///!< MPU Region Size 2 KBytes
#define ARM_MPU_REGION_SIZE_4KB ((uint8_t)0x0BU) ///!< MPU Region Size 4 KBytes
#define ARM_MPU_REGION_SIZE_8KB ((uint8_t)0x0CU) ///!< MPU Region Size 8 KBytes
#define ARM_MPU_REGION_SIZE_16KB ((uint8_t)0x0DU) ///!< MPU Region Size 16 KBytes
#define ARM_MPU_REGION_SIZE_32KB ((uint8_t)0x0EU) ///!< MPU Region Size 32 KBytes
#define ARM_MPU_REGION_SIZE_64KB ((uint8_t)0x0FU) ///!< MPU Region Size 64 KBytes
#define ARM_MPU_REGION_SIZE_128KB ((uint8_t)0x10U) ///!< MPU Region Size 128 KBytes
#define ARM_MPU_REGION_SIZE_256KB ((uint8_t)0x11U) ///!< MPU Region Size 256 KBytes
#define ARM_MPU_REGION_SIZE_512KB ((uint8_t)0x12U) ///!< MPU Region Size 512 KBytes
#define ARM_MPU_REGION_SIZE_1MB ((uint8_t)0x13U) ///!< MPU Region Size 1 MByte
#define ARM_MPU_REGION_SIZE_2MB ((uint8_t)0x14U) ///!< MPU Region Size 2 MBytes
#define ARM_MPU_REGION_SIZE_4MB ((uint8_t)0x15U) ///!< MPU Region Size 4 MBytes
#define ARM_MPU_REGION_SIZE_8MB ((uint8_t)0x16U) ///!< MPU Region Size 8 MBytes
#define ARM_MPU_REGION_SIZE_16MB ((uint8_t)0x17U) ///!< MPU Region Size 16 MBytes
#define ARM_MPU_REGION_SIZE_32MB ((uint8_t)0x18U) ///!< MPU Region Size 32 MBytes
#define ARM_MPU_REGION_SIZE_64MB ((uint8_t)0x19U) ///!< MPU Region Size 64 MBytes
#define ARM_MPU_REGION_SIZE_128MB ((uint8_t)0x1AU) ///!< MPU Region Size 128 MBytes
#define ARM_MPU_REGION_SIZE_256MB ((uint8_t)0x1BU) ///!< MPU Region Size 256 MBytes
#define ARM_MPU_REGION_SIZE_512MB ((uint8_t)0x1CU) ///!< MPU Region Size 512 MBytes
#define ARM_MPU_REGION_SIZE_1GB ((uint8_t)0x1DU) ///!< MPU Region Size 1 GByte
#define ARM_MPU_REGION_SIZE_2GB ((uint8_t)0x1EU) ///!< MPU Region Size 2 GBytes
#define ARM_MPU_REGION_SIZE_4GB ((uint8_t)0x1FU) ///!< MPU Region Size 4 GBytes
#define ARM_MPU_AP_NONE 0U ///!< MPU Access Permission no access
#define ARM_MPU_AP_PRIV 1U ///!< MPU Access Permission privileged access only
#define ARM_MPU_AP_URO 2U ///!< MPU Access Permission unprivileged access read-only
#define ARM_MPU_AP_FULL 3U ///!< MPU Access Permission full access
#define ARM_MPU_AP_PRO 5U ///!< MPU Access Permission privileged access read-only
#define ARM_MPU_AP_RO 6U ///!< MPU Access Permission read-only access
/** MPU Region Base Address Register Value
*
* \param Region The region to be configured, number 0 to 15.
* \param BaseAddress The base address for the region.
*/
#define ARM_MPU_RBAR(Region, BaseAddress) \
(((BaseAddress) & MPU_RBAR_ADDR_Msk) | \
((Region) & MPU_RBAR_REGION_Msk) | \
(MPU_RBAR_VALID_Msk))
/**
* MPU Memory Access Attributes
*
* \param TypeExtField Type extension field, allows you to configure memory access type, for example strongly ordered, peripheral.
* \param IsShareable Region is shareable between multiple bus masters.
* \param IsCacheable Region is cacheable, i.e. its value may be kept in cache.
* \param IsBufferable Region is bufferable, i.e. using write-back caching. Cacheable but non-bufferable regions use write-through policy.
*/
#define ARM_MPU_ACCESS_(TypeExtField, IsShareable, IsCacheable, IsBufferable) \
((((TypeExtField) << MPU_RASR_TEX_Pos) & MPU_RASR_TEX_Msk) | \
(((IsShareable) << MPU_RASR_S_Pos) & MPU_RASR_S_Msk) | \
(((IsCacheable) << MPU_RASR_C_Pos) & MPU_RASR_C_Msk) | \
(((IsBufferable) << MPU_RASR_B_Pos) & MPU_RASR_B_Msk))
/**
* MPU Region Attribute and Size Register Value
*
* \param DisableExec Instruction access disable bit, 1= disable instruction fetches.
* \param AccessPermission Data access permissions, allows you to configure read/write access for User and Privileged mode.
* \param AccessAttributes Memory access attribution, see \ref ARM_MPU_ACCESS_.
* \param SubRegionDisable Sub-region disable field.
* \param Size Region size of the region to be configured, for example 4K, 8K.
*/
#define ARM_MPU_RASR_EX(DisableExec, AccessPermission, AccessAttributes, SubRegionDisable, Size) \
((((DisableExec) << MPU_RASR_XN_Pos) & MPU_RASR_XN_Msk) | \
(((AccessPermission) << MPU_RASR_AP_Pos) & MPU_RASR_AP_Msk) | \
(((AccessAttributes) & (MPU_RASR_TEX_Msk | MPU_RASR_S_Msk | MPU_RASR_C_Msk | MPU_RASR_B_Msk))) | \
(((SubRegionDisable) << MPU_RASR_SRD_Pos) & MPU_RASR_SRD_Msk) | \
(((Size) << MPU_RASR_SIZE_Pos) & MPU_RASR_SIZE_Msk) | \
(((MPU_RASR_ENABLE_Msk))))
/**
* MPU Region Attribute and Size Register Value
*
* \param DisableExec Instruction access disable bit, 1= disable instruction fetches.
* \param AccessPermission Data access permissions, allows you to configure read/write access for User and Privileged mode.
* \param TypeExtField Type extension field, allows you to configure memory access type, for example strongly ordered, peripheral.
* \param IsShareable Region is shareable between multiple bus masters.
* \param IsCacheable Region is cacheable, i.e. its value may be kept in cache.
* \param IsBufferable Region is bufferable, i.e. using write-back caching. Cacheable but non-bufferable regions use write-through policy.
* \param SubRegionDisable Sub-region disable field.
* \param Size Region size of the region to be configured, for example 4K, 8K.
*/
#define ARM_MPU_RASR(DisableExec, AccessPermission, TypeExtField, IsShareable, IsCacheable, IsBufferable, SubRegionDisable, Size) \
ARM_MPU_RASR_EX(DisableExec, AccessPermission, ARM_MPU_ACCESS_(TypeExtField, IsShareable, IsCacheable, IsBufferable), SubRegionDisable, Size)
/**
* MPU Memory Access Attribute for strongly ordered memory.
* - TEX: 000b
* - Shareable
* - Non-cacheable
* - Non-bufferable
*/
#define ARM_MPU_ACCESS_ORDERED ARM_MPU_ACCESS_(0U, 1U, 0U, 0U)
/**
* MPU Memory Access Attribute for device memory.
* - TEX: 000b (if shareable) or 010b (if non-shareable)
* - Shareable or non-shareable
* - Non-cacheable
* - Bufferable (if shareable) or non-bufferable (if non-shareable)
*
* \param IsShareable Configures the device memory as shareable or non-shareable.
*/
#define ARM_MPU_ACCESS_DEVICE(IsShareable) ((IsShareable) ? ARM_MPU_ACCESS_(0U, 1U, 0U, 1U) : ARM_MPU_ACCESS_(2U, 0U, 0U, 0U))
/**
* MPU Memory Access Attribute for normal memory.
* - TEX: 1BBb (reflecting outer cacheability rules)
* - Shareable or non-shareable
* - Cacheable or non-cacheable (reflecting inner cacheability rules)
* - Bufferable or non-bufferable (reflecting inner cacheability rules)
*
* \param OuterCp Configures the outer cache policy.
* \param InnerCp Configures the inner cache policy.
* \param IsShareable Configures the memory as shareable or non-shareable.
*/
#define ARM_MPU_ACCESS_NORMAL(OuterCp, InnerCp, IsShareable) ARM_MPU_ACCESS_((4U | (OuterCp)), IsShareable, ((InnerCp) >> 1U), ((InnerCp) & 1U))
/**
* MPU Memory Access Attribute non-cacheable policy.
*/
#define ARM_MPU_CACHEP_NOCACHE 0U
/**
* MPU Memory Access Attribute write-back, write and read allocate policy.
*/
#define ARM_MPU_CACHEP_WB_WRA 1U
/**
* MPU Memory Access Attribute write-through, no write allocate policy.
*/
#define ARM_MPU_CACHEP_WT_NWA 2U
/**
* MPU Memory Access Attribute write-back, no write allocate policy.
*/
#define ARM_MPU_CACHEP_WB_NWA 3U
/**
* Struct for a single MPU Region
*/
typedef struct {
uint32_t RBAR; //!< The region base address register value (RBAR)
uint32_t RASR; //!< The region attribute and size register value (RASR) \ref MPU_RASR
} ARM_MPU_Region_t;
/** Enable the MPU.
* \param MPU_Control Default access permissions for unconfigured regions.
*/
__STATIC_INLINE void ARM_MPU_Enable(uint32_t MPU_Control)
{
__DMB();
MPU->CTRL = MPU_Control | MPU_CTRL_ENABLE_Msk;
#ifdef SCB_SHCSR_MEMFAULTENA_Msk
SCB->SHCSR |= SCB_SHCSR_MEMFAULTENA_Msk;
#endif
__DSB();
__ISB();
}
/** Disable the MPU.
*/
__STATIC_INLINE void ARM_MPU_Disable(void)
{
__DMB();
#ifdef SCB_SHCSR_MEMFAULTENA_Msk
SCB->SHCSR &= ~SCB_SHCSR_MEMFAULTENA_Msk;
#endif
MPU->CTRL &= ~MPU_CTRL_ENABLE_Msk;
__DSB();
__ISB();
}
/** Clear and disable the given MPU region.
* \param rnr Region number to be cleared.
*/
__STATIC_INLINE void ARM_MPU_ClrRegion(uint32_t rnr)
{
MPU->RNR = rnr;
MPU->RASR = 0U;
}
/** Configure an MPU region.
* \param rbar Value for RBAR register.
* \param rasr Value for RASR register.
*/
__STATIC_INLINE void ARM_MPU_SetRegion(uint32_t rbar, uint32_t rasr)
{
MPU->RBAR = rbar;
MPU->RASR = rasr;
}
/** Configure the given MPU region.
* \param rnr Region number to be configured.
* \param rbar Value for RBAR register.
* \param rasr Value for RASR register.
*/
__STATIC_INLINE void ARM_MPU_SetRegionEx(uint32_t rnr, uint32_t rbar, uint32_t rasr)
{
MPU->RNR = rnr;
MPU->RBAR = rbar;
MPU->RASR = rasr;
}
/** Memcpy with strictly ordered memory access, e.g. used by code in ARM_MPU_Load().
* \param dst Destination data is copied to.
* \param src Source data is copied from.
* \param len Amount of data words to be copied.
*/
__STATIC_INLINE void ARM_MPU_OrderedMemcpy(volatile uint32_t* dst, const uint32_t* __RESTRICT src, uint32_t len)
{
uint32_t i;
for (i = 0U; i < len; ++i)
{
dst[i] = src[i];
}
}
/** Load the given number of MPU regions from a table.
* \param table Pointer to the MPU configuration table.
* \param cnt Amount of regions to be configured.
*/
__STATIC_INLINE void ARM_MPU_Load(ARM_MPU_Region_t const* table, uint32_t cnt)
{
const uint32_t rowWordSize = sizeof(ARM_MPU_Region_t)/4U;
while (cnt > MPU_TYPE_RALIASES) {
ARM_MPU_OrderedMemcpy(&(MPU->RBAR), &(table->RBAR), MPU_TYPE_RALIASES*rowWordSize);
table += MPU_TYPE_RALIASES;
cnt -= MPU_TYPE_RALIASES;
}
ARM_MPU_OrderedMemcpy(&(MPU->RBAR), &(table->RBAR), cnt*rowWordSize);
}
#endif

824
sdk/CMSIS/Core/Include/m-profile/cmsis_clang_m.h Normal file
View file

@ -0,0 +1,824 @@
/*
* Copyright (c) 2009-2024 Arm Limited. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* CMSIS-Core(M) Compiler LLVM/Clang Header File
*/
#ifndef __CMSIS_CLANG_M_H
#define __CMSIS_CLANG_M_H
#pragma clang system_header /* treat file as system include file */
#ifndef __CMSIS_CLANG_H
#error "This file must not be included directly"
#endif
#if (__ARM_ACLE >= 200)
#include <arm_acle.h>
#else
#error Compiler must support ACLE V2.0
#endif /* (__ARM_ACLE >= 200) */
/* Fallback for __has_builtin */
#ifndef __has_builtin
#define __has_builtin(x) (0)
#endif
/* ######################### Startup and Lowlevel Init ######################## */
#ifndef __PROGRAM_START
#define __PROGRAM_START _start
#endif
#ifndef __INITIAL_SP
#define __INITIAL_SP __stack
#endif
#ifndef __STACK_LIMIT
#define __STACK_LIMIT __stack_limit
#endif
#ifndef __VECTOR_TABLE
#define __VECTOR_TABLE __Vectors
#endif
#ifndef __VECTOR_TABLE_ATTRIBUTE
#define __VECTOR_TABLE_ATTRIBUTE __attribute__((used, section(".vectors")))
#endif
#if (__ARM_FEATURE_CMSE == 3)
#ifndef __STACK_SEAL
#define __STACK_SEAL __stack_seal
#endif
#ifndef __TZ_STACK_SEAL_SIZE
#define __TZ_STACK_SEAL_SIZE 8U
#endif
#ifndef __TZ_STACK_SEAL_VALUE
#define __TZ_STACK_SEAL_VALUE 0xFEF5EDA5FEF5EDA5ULL
#endif
__STATIC_FORCEINLINE void __TZ_set_STACKSEAL_S (uint32_t* stackTop) {
*((uint64_t *)stackTop) = __TZ_STACK_SEAL_VALUE;
}
#endif
#if (__ARM_ARCH_ISA_THUMB >= 2)
/**
\brief STRT Unprivileged (8 bit)
\details Executes a Unprivileged STRT instruction for 8 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
*/
__STATIC_FORCEINLINE void __STRBT(uint8_t value, volatile uint8_t *ptr)
{
__ASM volatile ("strbt %1, %0" : "=Q" (*ptr) : "r" ((uint32_t)value) );
}
/**
\brief STRT Unprivileged (16 bit)
\details Executes a Unprivileged STRT instruction for 16 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
*/
__STATIC_FORCEINLINE void __STRHT(uint16_t value, volatile uint16_t *ptr)
{
__ASM volatile ("strht %1, %0" : "=Q" (*ptr) : "r" ((uint32_t)value) );
}
/**
\brief STRT Unprivileged (32 bit)
\details Executes a Unprivileged STRT instruction for 32 bit values.
\param [in] value Value to store
\param [in] ptr Pointer to location
*/
__STATIC_FORCEINLINE void __STRT(uint32_t value, volatile uint32_t *ptr)
{
__ASM volatile ("strt %1, %0" : "=Q" (*ptr) : "r" (value) );
}
#endif /* (__ARM_ARCH_ISA_THUMB >= 2) */
/* ########################### Core Function Access ########################### */
/** \ingroup CMSIS_Core_FunctionInterface
\defgroup CMSIS_Core_RegAccFunctions CMSIS Core Register Access Functions
@{
*/
/**
\brief Get Control Register
\details Returns the content of the Control Register.
\return Control Register value
*/
__STATIC_FORCEINLINE uint32_t __get_CONTROL(void)
{
uint32_t result;
__ASM volatile ("MRS %0, control" : "=r" (result) );
return (result);
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Control Register (non-secure)
\details Returns the content of the non-secure Control Register when in secure mode.
\return non-secure Control Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_CONTROL_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, control_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Control Register
\details Writes the given value to the Control Register.
\param [in] control Control Register value to set
*/
__STATIC_FORCEINLINE void __set_CONTROL(uint32_t control)
{
__ASM volatile ("MSR control, %0" : : "r" (control) : "memory");
__ISB();
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Control Register (non-secure)
\details Writes the given value to the non-secure Control Register when in secure state.
\param [in] control Control Register value to set
*/
__STATIC_FORCEINLINE void __TZ_set_CONTROL_NS(uint32_t control)
{
__ASM volatile ("MSR control_ns, %0" : : "r" (control) : "memory");
__ISB();
}
#endif
/**
\brief Get IPSR Register
\details Returns the content of the IPSR Register.
\return IPSR Register value
*/
__STATIC_FORCEINLINE uint32_t __get_IPSR(void)
{
uint32_t result;
__ASM volatile ("MRS %0, ipsr" : "=r" (result) );
return (result);
}
/**
\brief Get APSR Register
\details Returns the content of the APSR Register.
\return APSR Register value
*/
__STATIC_FORCEINLINE uint32_t __get_APSR(void)
{
uint32_t result;
__ASM volatile ("MRS %0, apsr" : "=r" (result) );
return (result);
}
/**
\brief Get xPSR Register
\details Returns the content of the xPSR Register.
\return xPSR Register value
*/
__STATIC_FORCEINLINE uint32_t __get_xPSR(void)
{
uint32_t result;
__ASM volatile ("MRS %0, xpsr" : "=r" (result) );
return (result);
}
/**
\brief Get Process Stack Pointer
\details Returns the current value of the Process Stack Pointer (PSP).
\return PSP Register value
*/
__STATIC_FORCEINLINE uint32_t __get_PSP(void)
{
uint32_t result;
__ASM volatile ("MRS %0, psp" : "=r" (result) );
return (result);
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Process Stack Pointer (non-secure)
\details Returns the current value of the non-secure Process Stack Pointer (PSP) when in secure state.
\return PSP Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_PSP_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, psp_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Process Stack Pointer
\details Assigns the given value to the Process Stack Pointer (PSP).
\param [in] topOfProcStack Process Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __set_PSP(uint32_t topOfProcStack)
{
__ASM volatile ("MSR psp, %0" : : "r" (topOfProcStack) : );
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Process Stack Pointer (non-secure)
\details Assigns the given value to the non-secure Process Stack Pointer (PSP) when in secure state.
\param [in] topOfProcStack Process Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __TZ_set_PSP_NS(uint32_t topOfProcStack)
{
__ASM volatile ("MSR psp_ns, %0" : : "r" (topOfProcStack) : );
}
#endif
/**
\brief Get Main Stack Pointer
\details Returns the current value of the Main Stack Pointer (MSP).
\return MSP Register value
*/
__STATIC_FORCEINLINE uint32_t __get_MSP(void)
{
uint32_t result;
__ASM volatile ("MRS %0, msp" : "=r" (result) );
return (result);
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Main Stack Pointer (non-secure)
\details Returns the current value of the non-secure Main Stack Pointer (MSP) when in secure state.
\return MSP Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_MSP_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, msp_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Main Stack Pointer
\details Assigns the given value to the Main Stack Pointer (MSP).
\param [in] topOfMainStack Main Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __set_MSP(uint32_t topOfMainStack)
{
__ASM volatile ("MSR msp, %0" : : "r" (topOfMainStack) : );
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Main Stack Pointer (non-secure)
\details Assigns the given value to the non-secure Main Stack Pointer (MSP) when in secure state.
\param [in] topOfMainStack Main Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __TZ_set_MSP_NS(uint32_t topOfMainStack)
{
__ASM volatile ("MSR msp_ns, %0" : : "r" (topOfMainStack) : );
}
#endif
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Stack Pointer (non-secure)
\details Returns the current value of the non-secure Stack Pointer (SP) when in secure state.
\return SP Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_SP_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, sp_ns" : "=r" (result) );
return (result);
}
/**
\brief Set Stack Pointer (non-secure)
\details Assigns the given value to the non-secure Stack Pointer (SP) when in secure state.
\param [in] topOfStack Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __TZ_set_SP_NS(uint32_t topOfStack)
{
__ASM volatile ("MSR sp_ns, %0" : : "r" (topOfStack) : );
}
#endif
/**
\brief Get Priority Mask
\details Returns the current state of the priority mask bit from the Priority Mask Register.
\return Priority Mask value
*/
__STATIC_FORCEINLINE uint32_t __get_PRIMASK(void)
{
uint32_t result;
__ASM volatile ("MRS %0, primask" : "=r" (result) );
return (result);
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Priority Mask (non-secure)
\details Returns the current state of the non-secure priority mask bit from the Priority Mask Register when in secure state.
\return Priority Mask value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_PRIMASK_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, primask_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Priority Mask
\details Assigns the given value to the Priority Mask Register.
\param [in] priMask Priority Mask
*/
__STATIC_FORCEINLINE void __set_PRIMASK(uint32_t priMask)
{
__ASM volatile ("MSR primask, %0" : : "r" (priMask) : "memory");
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Priority Mask (non-secure)
\details Assigns the given value to the non-secure Priority Mask Register when in secure state.
\param [in] priMask Priority Mask
*/
__STATIC_FORCEINLINE void __TZ_set_PRIMASK_NS(uint32_t priMask)
{
__ASM volatile ("MSR primask_ns, %0" : : "r" (priMask) : "memory");
}
#endif
#if (__ARM_ARCH_ISA_THUMB >= 2)
/**
\brief Get Base Priority
\details Returns the current value of the Base Priority register.
\return Base Priority register value
*/
__STATIC_FORCEINLINE uint32_t __get_BASEPRI(void)
{
uint32_t result;
__ASM volatile ("MRS %0, basepri" : "=r" (result) );
return (result);
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Base Priority (non-secure)
\details Returns the current value of the non-secure Base Priority register when in secure state.
\return Base Priority register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_BASEPRI_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, basepri_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Base Priority
\details Assigns the given value to the Base Priority register.
\param [in] basePri Base Priority value to set
*/
__STATIC_FORCEINLINE void __set_BASEPRI(uint32_t basePri)
{
__ASM volatile ("MSR basepri, %0" : : "r" (basePri) : "memory");
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Base Priority (non-secure)
\details Assigns the given value to the non-secure Base Priority register when in secure state.
\param [in] basePri Base Priority value to set
*/
__STATIC_FORCEINLINE void __TZ_set_BASEPRI_NS(uint32_t basePri)
{
__ASM volatile ("MSR basepri_ns, %0" : : "r" (basePri) : "memory");
}
#endif
/**
\brief Set Base Priority with condition
\details Assigns the given value to the Base Priority register only if BASEPRI masking is disabled,
or the new value increases the BASEPRI priority level.
\param [in] basePri Base Priority value to set
*/
__STATIC_FORCEINLINE void __set_BASEPRI_MAX(uint32_t basePri)
{
__ASM volatile ("MSR basepri_max, %0" : : "r" (basePri) : "memory");
}
/**
\brief Get Fault Mask
\details Returns the current value of the Fault Mask register.
\return Fault Mask register value
*/
__STATIC_FORCEINLINE uint32_t __get_FAULTMASK(void)
{
uint32_t result;
__ASM volatile ("MRS %0, faultmask" : "=r" (result) );
return (result);
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Fault Mask (non-secure)
\details Returns the current value of the non-secure Fault Mask register when in secure state.
\return Fault Mask register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_FAULTMASK_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, faultmask_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Fault Mask
\details Assigns the given value to the Fault Mask register.
\param [in] faultMask Fault Mask value to set
*/
__STATIC_FORCEINLINE void __set_FAULTMASK(uint32_t faultMask)
{
__ASM volatile ("MSR faultmask, %0" : : "r" (faultMask) : "memory");
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Fault Mask (non-secure)
\details Assigns the given value to the non-secure Fault Mask register when in secure state.
\param [in] faultMask Fault Mask value to set
*/
__STATIC_FORCEINLINE void __TZ_set_FAULTMASK_NS(uint32_t faultMask)
{
__ASM volatile ("MSR faultmask_ns, %0" : : "r" (faultMask) : "memory");
}
#endif
#endif /* (__ARM_ARCH_ISA_THUMB >= 2) */
#if (__ARM_ARCH >= 8)
/**
\brief Get Process Stack Pointer Limit
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence zero is returned always in non-secure
mode.
\details Returns the current value of the Process Stack Pointer Limit (PSPLIM).
\return PSPLIM Register value
*/
__STATIC_FORCEINLINE uint32_t __get_PSPLIM(void)
{
#if (((__ARM_ARCH_8M_MAIN__ < 1) && \
(__ARM_ARCH_8_1M_MAIN__ < 1) ) && \
(__ARM_FEATURE_CMSE < 3) )
/* without main extensions, the non-secure PSPLIM is RAZ/WI */
return (0U);
#else
uint32_t result;
__ASM volatile ("MRS %0, psplim" : "=r" (result) );
return (result);
#endif
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Process Stack Pointer Limit (non-secure)
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence zero is returned always.
\details Returns the current value of the non-secure Process Stack Pointer Limit (PSPLIM) when in secure state.
\return PSPLIM Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_PSPLIM_NS(void)
{
#if ((__ARM_ARCH_8M_MAIN__ < 1) && \
(__ARM_ARCH_8_1M_MAIN__ < 1) )
/* without main extensions, the non-secure PSPLIM is RAZ/WI */
return (0U);
#else
uint32_t result;
__ASM volatile ("MRS %0, psplim_ns" : "=r" (result) );
return (result);
#endif
}
#endif
/**
\brief Set Process Stack Pointer Limit
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence the write is silently ignored in non-secure
mode.
\details Assigns the given value to the Process Stack Pointer Limit (PSPLIM).
\param [in] ProcStackPtrLimit Process Stack Pointer Limit value to set
*/
__STATIC_FORCEINLINE void __set_PSPLIM(uint32_t ProcStackPtrLimit)
{
#if (((__ARM_ARCH_8M_MAIN__ < 1) && \
(__ARM_ARCH_8_1M_MAIN__ < 1) ) && \
(__ARM_FEATURE_CMSE < 3) )
/* without main extensions, the non-secure PSPLIM is RAZ/WI */
(void)ProcStackPtrLimit;
#else
__ASM volatile ("MSR psplim, %0" : : "r" (ProcStackPtrLimit));
#endif
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Process Stack Pointer (non-secure)
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence the write is silently ignored.
\details Assigns the given value to the non-secure Process Stack Pointer Limit (PSPLIM) when in secure state.
\param [in] ProcStackPtrLimit Process Stack Pointer Limit value to set
*/
__STATIC_FORCEINLINE void __TZ_set_PSPLIM_NS(uint32_t ProcStackPtrLimit)
{
#if ((__ARM_ARCH_8M_MAIN__ < 1) && \
(__ARM_ARCH_8_1M_MAIN__ < 1) )
/* without main extensions, the non-secure PSPLIM is RAZ/WI */
(void)ProcStackPtrLimit;
#else
__ASM volatile ("MSR psplim_ns, %0\n" : : "r" (ProcStackPtrLimit));
#endif
}
#endif
/**
\brief Get Main Stack Pointer Limit
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence zero is returned always.
\details Returns the current value of the Main Stack Pointer Limit (MSPLIM).
\return MSPLIM Register value
*/
__STATIC_FORCEINLINE uint32_t __get_MSPLIM(void)
{
#if (((__ARM_ARCH_8M_MAIN__ < 1) && \
(__ARM_ARCH_8_1M_MAIN__ < 1) ) && \
(__ARM_FEATURE_CMSE < 3) )
/* without main extensions, the non-secure MSPLIM is RAZ/WI */
return (0U);
#else
uint32_t result;
__ASM volatile ("MRS %0, msplim" : "=r" (result) );
return (result);
#endif
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Main Stack Pointer Limit (non-secure)
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence zero is returned always.
\details Returns the current value of the non-secure Main Stack Pointer Limit(MSPLIM) when in secure state.
\return MSPLIM Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_MSPLIM_NS(void)
{
#if ((__ARM_ARCH_8M_MAIN__ < 1) && \
(__ARM_ARCH_8_1M_MAIN__ < 1) )
/* without main extensions, the non-secure MSPLIM is RAZ/WI */
return (0U);
#else
uint32_t result;
__ASM volatile ("MRS %0, msplim_ns" : "=r" (result) );
return (result);
#endif
}
#endif
/**
\brief Set Main Stack Pointer Limit
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence the write is silently ignored.
\details Assigns the given value to the Main Stack Pointer Limit (MSPLIM).
\param [in] MainStackPtrLimit Main Stack Pointer Limit value to set
*/
__STATIC_FORCEINLINE void __set_MSPLIM(uint32_t MainStackPtrLimit)
{
#if (((__ARM_ARCH_8M_MAIN__ < 1) && \
(__ARM_ARCH_8_1M_MAIN__ < 1) ) && \
(__ARM_FEATURE_CMSE < 3) )
/* without main extensions, the non-secure MSPLIM is RAZ/WI */
(void)MainStackPtrLimit;
#else
__ASM volatile ("MSR msplim, %0" : : "r" (MainStackPtrLimit));
#endif
}
#if (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Main Stack Pointer Limit (non-secure)
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence the write is silently ignored.
\details Assigns the given value to the non-secure Main Stack Pointer Limit (MSPLIM) when in secure state.
\param [in] MainStackPtrLimit Main Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __TZ_set_MSPLIM_NS(uint32_t MainStackPtrLimit)
{
#if ((__ARM_ARCH_8M_MAIN__ < 1) && \
(__ARM_ARCH_8_1M_MAIN__ < 1) )
/* without main extensions, the non-secure MSPLIM is RAZ/WI */
(void)MainStackPtrLimit;
#else
__ASM volatile ("MSR msplim_ns, %0" : : "r" (MainStackPtrLimit));
#endif
}
#endif
#endif /* (__ARM_ARCH >= 8) */
/* ################### Compiler specific Intrinsics ########################### */
/** \defgroup CMSIS_SIMD_intrinsics CMSIS SIMD Intrinsics
Access to dedicated SIMD instructions
@{
*/
#if (defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1))
#define __SADD8 __sadd8
#define __QADD8 __qadd8
#define __SHADD8 __shadd8
#define __UADD8 __uadd8
#define __UQADD8 __uqadd8
#define __UHADD8 __uhadd8
#define __SSUB8 __ssub8
#define __QSUB8 __qsub8
#define __SHSUB8 __shsub8
#define __USUB8 __usub8
#define __UQSUB8 __uqsub8
#define __UHSUB8 __uhsub8
#define __SADD16 __sadd16
#define __QADD16 __qadd16
#define __SHADD16 __shadd16
#define __UADD16 __uadd16
#define __UQADD16 __uqadd16
#define __UHADD16 __uhadd16
#define __SSUB16 __ssub16
#define __QSUB16 __qsub16
#define __SHSUB16 __shsub16
#define __USUB16 __usub16
#define __UQSUB16 __uqsub16
#define __UHSUB16 __uhsub16
#define __SASX __sasx
#define __QASX __qasx
#define __SHASX __shasx
#define __UASX __uasx
#define __UQASX __uqasx
#define __UHASX __uhasx
#define __SSAX __ssax
#define __QSAX __qsax
#define __SHSAX __shsax
#define __USAX __usax
#define __UQSAX __uqsax
#define __UHSAX __uhsax
#define __USAD8 __usad8
#define __USADA8 __usada8
#define __SSAT16 __ssat16
#define __USAT16 __usat16
#define __UXTB16 __uxtb16
#define __UXTAB16 __uxtab16
#define __SXTB16 __sxtb16
#define __SXTAB16 __sxtab16
#define __SMUAD __smuad
#define __SMUADX __smuadx
#define __SMLAD __smlad
#define __SMLADX __smladx
#define __SMLALD __smlald
#define __SMLALDX __smlaldx
#define __SMUSD __smusd
#define __SMUSDX __smusdx
#define __SMLSD __smlsd
#define __SMLSDX __smlsdx
#define __SMLSLD __smlsld
#define __SMLSLDX __smlsldx
#define __SEL __sel
#define __QADD __qadd
#define __QSUB __qsub
#define __PKHBT(ARG1,ARG2,ARG3) \
__extension__ \
({ \
uint32_t __RES, __ARG1 = (ARG1), __ARG2 = (ARG2); \
__ASM ("pkhbt %0, %1, %2, lsl %3" : "=r" (__RES) : "r" (__ARG1), "r" (__ARG2), "I" (ARG3) ); \
__RES; \
})
#define __PKHTB(ARG1,ARG2,ARG3) \
__extension__ \
({ \
uint32_t __RES, __ARG1 = (ARG1), __ARG2 = (ARG2); \
if (ARG3 == 0) \
__ASM ("pkhtb %0, %1, %2" : "=r" (__RES) : "r" (__ARG1), "r" (__ARG2) ); \
else \
__ASM ("pkhtb %0, %1, %2, asr %3" : "=r" (__RES) : "r" (__ARG1), "r" (__ARG2), "I" (ARG3) ); \
__RES; \
})
#define __SXTB16_RORn(ARG1, ARG2) __SXTB16(__ROR(ARG1, ARG2))
#define __SXTAB16_RORn(ARG1, ARG2, ARG3) __SXTAB16(ARG1, __ROR(ARG2, ARG3))
__STATIC_FORCEINLINE int32_t __SMMLA (int32_t op1, int32_t op2, int32_t op3)
{
int32_t result;
__ASM volatile ("smmla %0, %1, %2, %3" : "=r" (result): "r" (op1), "r" (op2), "r" (op3) );
return (result);
}
#endif /* (defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1)) */
/** @} end of group CMSIS_SIMD_intrinsics */
/** @} end of CMSIS_Core_RegAccFunctions */
#endif /* __CMSIS_CLANG_M_H */

717
sdk/CMSIS/Core/Include/m-profile/cmsis_gcc_m.h Normal file
View file

@ -0,0 +1,717 @@
/*
* Copyright (c) 2009-2023 Arm Limited. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* CMSIS-Core(M) Compiler GCC Header File
*/
#ifndef __CMSIS_GCC_M_H
#define __CMSIS_GCC_M_H
#ifndef __CMSIS_GCC_H
#error "This file must not be included directly"
#endif
#include <arm_acle.h>
/* ######################### Startup and Lowlevel Init ######################## */
#ifndef __PROGRAM_START
/**
\brief Initializes data and bss sections
\details This default implementations initialized all data and additional bss
sections relying on .copy.table and .zero.table specified properly
in the used linker script.
*/
__STATIC_FORCEINLINE __NO_RETURN void __cmsis_start(void)
{
extern void _start(void) __NO_RETURN;
typedef struct __copy_table {
uint32_t const* src;
uint32_t* dest;
uint32_t wlen;
} __copy_table_t;
typedef struct __zero_table {
uint32_t* dest;
uint32_t wlen;
} __zero_table_t;
extern const __copy_table_t __copy_table_start__;
extern const __copy_table_t __copy_table_end__;
extern const __zero_table_t __zero_table_start__;
extern const __zero_table_t __zero_table_end__;
for (__copy_table_t const* pTable = &__copy_table_start__; pTable < &__copy_table_end__; ++pTable) {
for(uint32_t i=0u; i<pTable->wlen; ++i) {
pTable->dest[i] = pTable->src[i];
}
}
for (__zero_table_t const* pTable = &__zero_table_start__; pTable < &__zero_table_end__; ++pTable) {
for(uint32_t i=0u; i<pTable->wlen; ++i) {
pTable->dest[i] = 0u;
}
}
_start();
}
#define __PROGRAM_START __cmsis_start
#endif
#ifndef __INITIAL_SP
#define __INITIAL_SP __StackTop
#endif
#ifndef __STACK_LIMIT
#define __STACK_LIMIT __StackLimit
#endif
#ifndef __VECTOR_TABLE
#define __VECTOR_TABLE __Vectors
#endif
#ifndef __VECTOR_TABLE_ATTRIBUTE
#define __VECTOR_TABLE_ATTRIBUTE __attribute__((used, section(".vectors")))
#endif
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
#ifndef __STACK_SEAL
#define __STACK_SEAL __StackSeal
#endif
#ifndef __TZ_STACK_SEAL_SIZE
#define __TZ_STACK_SEAL_SIZE 8U
#endif
#ifndef __TZ_STACK_SEAL_VALUE
#define __TZ_STACK_SEAL_VALUE 0xFEF5EDA5FEF5EDA5ULL
#endif
__STATIC_FORCEINLINE void __TZ_set_STACKSEAL_S (uint32_t* stackTop) {
*((uint64_t *)stackTop) = __TZ_STACK_SEAL_VALUE;
}
#endif
/* ########################### Core Function Access ########################### */
/** \ingroup CMSIS_Core_FunctionInterface
\defgroup CMSIS_Core_RegAccFunctions CMSIS Core Register Access Functions
@{
*/
/**
\brief Get Control Register
\details Returns the content of the Control Register.
\return Control Register value
*/
__STATIC_FORCEINLINE uint32_t __get_CONTROL(void)
{
uint32_t result;
__ASM volatile ("MRS %0, control" : "=r" (result) );
return (result);
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Control Register (non-secure)
\details Returns the content of the non-secure Control Register when in secure mode.
\return non-secure Control Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_CONTROL_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, control_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Control Register
\details Writes the given value to the Control Register.
\param [in] control Control Register value to set
*/
__STATIC_FORCEINLINE void __set_CONTROL(uint32_t control)
{
__ASM volatile ("MSR control, %0" : : "r" (control) : "memory");
__ISB();
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Control Register (non-secure)
\details Writes the given value to the non-secure Control Register when in secure state.
\param [in] control Control Register value to set
*/
__STATIC_FORCEINLINE void __TZ_set_CONTROL_NS(uint32_t control)
{
__ASM volatile ("MSR control_ns, %0" : : "r" (control) : "memory");
__ISB();
}
#endif
/**
\brief Get IPSR Register
\details Returns the content of the IPSR Register.
\return IPSR Register value
*/
__STATIC_FORCEINLINE uint32_t __get_IPSR(void)
{
uint32_t result;
__ASM volatile ("MRS %0, ipsr" : "=r" (result) );
return (result);
}
/**
\brief Get APSR Register
\details Returns the content of the APSR Register.
\return APSR Register value
*/
__STATIC_FORCEINLINE uint32_t __get_APSR(void)
{
uint32_t result;
__ASM volatile ("MRS %0, apsr" : "=r" (result) );
return (result);
}
/**
\brief Get xPSR Register
\details Returns the content of the xPSR Register.
\return xPSR Register value
*/
__STATIC_FORCEINLINE uint32_t __get_xPSR(void)
{
uint32_t result;
__ASM volatile ("MRS %0, xpsr" : "=r" (result) );
return (result);
}
/**
\brief Get Process Stack Pointer
\details Returns the current value of the Process Stack Pointer (PSP).
\return PSP Register value
*/
__STATIC_FORCEINLINE uint32_t __get_PSP(void)
{
uint32_t result;
__ASM volatile ("MRS %0, psp" : "=r" (result) );
return (result);
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Process Stack Pointer (non-secure)
\details Returns the current value of the non-secure Process Stack Pointer (PSP) when in secure state.
\return PSP Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_PSP_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, psp_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Process Stack Pointer
\details Assigns the given value to the Process Stack Pointer (PSP).
\param [in] topOfProcStack Process Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __set_PSP(uint32_t topOfProcStack)
{
__ASM volatile ("MSR psp, %0" : : "r" (topOfProcStack) : );
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Process Stack Pointer (non-secure)
\details Assigns the given value to the non-secure Process Stack Pointer (PSP) when in secure state.
\param [in] topOfProcStack Process Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __TZ_set_PSP_NS(uint32_t topOfProcStack)
{
__ASM volatile ("MSR psp_ns, %0" : : "r" (topOfProcStack) : );
}
#endif
/**
\brief Get Main Stack Pointer
\details Returns the current value of the Main Stack Pointer (MSP).
\return MSP Register value
*/
__STATIC_FORCEINLINE uint32_t __get_MSP(void)
{
uint32_t result;
__ASM volatile ("MRS %0, msp" : "=r" (result) );
return (result);
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Main Stack Pointer (non-secure)
\details Returns the current value of the non-secure Main Stack Pointer (MSP) when in secure state.
\return MSP Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_MSP_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, msp_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Main Stack Pointer
\details Assigns the given value to the Main Stack Pointer (MSP).
\param [in] topOfMainStack Main Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __set_MSP(uint32_t topOfMainStack)
{
__ASM volatile ("MSR msp, %0" : : "r" (topOfMainStack) : );
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Main Stack Pointer (non-secure)
\details Assigns the given value to the non-secure Main Stack Pointer (MSP) when in secure state.
\param [in] topOfMainStack Main Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __TZ_set_MSP_NS(uint32_t topOfMainStack)
{
__ASM volatile ("MSR msp_ns, %0" : : "r" (topOfMainStack) : );
}
#endif
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Stack Pointer (non-secure)
\details Returns the current value of the non-secure Stack Pointer (SP) when in secure state.
\return SP Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_SP_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, sp_ns" : "=r" (result) );
return (result);
}
/**
\brief Set Stack Pointer (non-secure)
\details Assigns the given value to the non-secure Stack Pointer (SP) when in secure state.
\param [in] topOfStack Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __TZ_set_SP_NS(uint32_t topOfStack)
{
__ASM volatile ("MSR sp_ns, %0" : : "r" (topOfStack) : );
}
#endif
/**
\brief Get Priority Mask
\details Returns the current state of the priority mask bit from the Priority Mask Register.
\return Priority Mask value
*/
__STATIC_FORCEINLINE uint32_t __get_PRIMASK(void)
{
uint32_t result;
__ASM volatile ("MRS %0, primask" : "=r" (result) );
return (result);
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Priority Mask (non-secure)
\details Returns the current state of the non-secure priority mask bit from the Priority Mask Register when in secure state.
\return Priority Mask value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_PRIMASK_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, primask_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Priority Mask
\details Assigns the given value to the Priority Mask Register.
\param [in] priMask Priority Mask
*/
__STATIC_FORCEINLINE void __set_PRIMASK(uint32_t priMask)
{
__ASM volatile ("MSR primask, %0" : : "r" (priMask) : "memory");
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Priority Mask (non-secure)
\details Assigns the given value to the non-secure Priority Mask Register when in secure state.
\param [in] priMask Priority Mask
*/
__STATIC_FORCEINLINE void __TZ_set_PRIMASK_NS(uint32_t priMask)
{
__ASM volatile ("MSR primask_ns, %0" : : "r" (priMask) : "memory");
}
#endif
#if (__ARM_ARCH_ISA_THUMB >= 2)
/**
\brief Get Base Priority
\details Returns the current value of the Base Priority register.
\return Base Priority register value
*/
__STATIC_FORCEINLINE uint32_t __get_BASEPRI(void)
{
uint32_t result;
__ASM volatile ("MRS %0, basepri" : "=r" (result) );
return (result);
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Base Priority (non-secure)
\details Returns the current value of the non-secure Base Priority register when in secure state.
\return Base Priority register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_BASEPRI_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, basepri_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Base Priority
\details Assigns the given value to the Base Priority register.
\param [in] basePri Base Priority value to set
*/
__STATIC_FORCEINLINE void __set_BASEPRI(uint32_t basePri)
{
__ASM volatile ("MSR basepri, %0" : : "r" (basePri) : "memory");
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Base Priority (non-secure)
\details Assigns the given value to the non-secure Base Priority register when in secure state.
\param [in] basePri Base Priority value to set
*/
__STATIC_FORCEINLINE void __TZ_set_BASEPRI_NS(uint32_t basePri)
{
__ASM volatile ("MSR basepri_ns, %0" : : "r" (basePri) : "memory");
}
#endif
/**
\brief Set Base Priority with condition
\details Assigns the given value to the Base Priority register only if BASEPRI masking is disabled,
or the new value increases the BASEPRI priority level.
\param [in] basePri Base Priority value to set
*/
__STATIC_FORCEINLINE void __set_BASEPRI_MAX(uint32_t basePri)
{
__ASM volatile ("MSR basepri_max, %0" : : "r" (basePri) : "memory");
}
/**
\brief Get Fault Mask
\details Returns the current value of the Fault Mask register.
\return Fault Mask register value
*/
__STATIC_FORCEINLINE uint32_t __get_FAULTMASK(void)
{
uint32_t result;
__ASM volatile ("MRS %0, faultmask" : "=r" (result) );
return (result);
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Fault Mask (non-secure)
\details Returns the current value of the non-secure Fault Mask register when in secure state.
\return Fault Mask register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_FAULTMASK_NS(void)
{
uint32_t result;
__ASM volatile ("MRS %0, faultmask_ns" : "=r" (result) );
return (result);
}
#endif
/**
\brief Set Fault Mask
\details Assigns the given value to the Fault Mask register.
\param [in] faultMask Fault Mask value to set
*/
__STATIC_FORCEINLINE void __set_FAULTMASK(uint32_t faultMask)
{
__ASM volatile ("MSR faultmask, %0" : : "r" (faultMask) : "memory");
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Fault Mask (non-secure)
\details Assigns the given value to the non-secure Fault Mask register when in secure state.
\param [in] faultMask Fault Mask value to set
*/
__STATIC_FORCEINLINE void __TZ_set_FAULTMASK_NS(uint32_t faultMask)
{
__ASM volatile ("MSR faultmask_ns, %0" : : "r" (faultMask) : "memory");
}
#endif
#endif /* (__ARM_ARCH_ISA_THUMB >= 2) */
#if (__ARM_ARCH >= 8)
/**
\brief Get Process Stack Pointer Limit
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence zero is returned always in non-secure
mode.
\details Returns the current value of the Process Stack Pointer Limit (PSPLIM).
\return PSPLIM Register value
*/
__STATIC_FORCEINLINE uint32_t __get_PSPLIM(void)
{
#if (!(defined (__ARM_ARCH_8M_MAIN__ ) && (__ARM_ARCH_8M_MAIN__ == 1)) && \
!(defined (__ARM_ARCH_8_1M_MAIN__ ) && (__ARM_ARCH_8_1M_MAIN__ == 1)) && \
(!defined (__ARM_FEATURE_CMSE ) || (__ARM_FEATURE_CMSE < 3)))
/* without main extensions, the non-secure PSPLIM is RAZ/WI */
return (0U);
#else
uint32_t result;
__ASM volatile ("MRS %0, psplim" : "=r" (result) );
return (result);
#endif
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Process Stack Pointer Limit (non-secure)
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence zero is returned always.
\details Returns the current value of the non-secure Process Stack Pointer Limit (PSPLIM) when in secure state.
\return PSPLIM Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_PSPLIM_NS(void)
{
#if (!(defined (__ARM_ARCH_8M_MAIN__ ) && (__ARM_ARCH_8M_MAIN__ == 1)) && \
!(defined (__ARM_ARCH_8_1M_MAIN__ ) && (__ARM_ARCH_8_1M_MAIN__ == 1)))
/* without main extensions, the non-secure PSPLIM is RAZ/WI */
return (0U);
#else
uint32_t result;
__ASM volatile ("MRS %0, psplim_ns" : "=r" (result) );
return (result);
#endif
}
#endif
/**
\brief Set Process Stack Pointer Limit
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence the write is silently ignored in non-secure
mode.
\details Assigns the given value to the Process Stack Pointer Limit (PSPLIM).
\param [in] ProcStackPtrLimit Process Stack Pointer Limit value to set
*/
__STATIC_FORCEINLINE void __set_PSPLIM(uint32_t ProcStackPtrLimit)
{
#if (!(defined (__ARM_ARCH_8M_MAIN__ ) && (__ARM_ARCH_8M_MAIN__ == 1)) && \
!(defined (__ARM_ARCH_8_1M_MAIN__ ) && (__ARM_ARCH_8_1M_MAIN__ == 1)) && \
(!defined (__ARM_FEATURE_CMSE ) || (__ARM_FEATURE_CMSE < 3)))
/* without main extensions, the non-secure PSPLIM is RAZ/WI */
(void)ProcStackPtrLimit;
#else
__ASM volatile ("MSR psplim, %0" : : "r" (ProcStackPtrLimit));
#endif
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Process Stack Pointer (non-secure)
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence the write is silently ignored.
\details Assigns the given value to the non-secure Process Stack Pointer Limit (PSPLIM) when in secure state.
\param [in] ProcStackPtrLimit Process Stack Pointer Limit value to set
*/
__STATIC_FORCEINLINE void __TZ_set_PSPLIM_NS(uint32_t ProcStackPtrLimit)
{
#if (!(defined (__ARM_ARCH_8M_MAIN__ ) && (__ARM_ARCH_8M_MAIN__ == 1)) && \
!(defined (__ARM_ARCH_8_1M_MAIN__ ) && (__ARM_ARCH_8_1M_MAIN__ == 1)))
/* without main extensions, the non-secure PSPLIM is RAZ/WI */
(void)ProcStackPtrLimit;
#else
__ASM volatile ("MSR psplim_ns, %0\n" : : "r" (ProcStackPtrLimit));
#endif
}
#endif
/**
\brief Get Main Stack Pointer Limit
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence zero is returned always.
\details Returns the current value of the Main Stack Pointer Limit (MSPLIM).
\return MSPLIM Register value
*/
__STATIC_FORCEINLINE uint32_t __get_MSPLIM(void)
{
#if (!(defined (__ARM_ARCH_8M_MAIN__ ) && (__ARM_ARCH_8M_MAIN__ == 1)) && \
!(defined (__ARM_ARCH_8_1M_MAIN__ ) && (__ARM_ARCH_8_1M_MAIN__ == 1)) && \
(!defined (__ARM_FEATURE_CMSE ) || (__ARM_FEATURE_CMSE < 3)))
/* without main extensions, the non-secure MSPLIM is RAZ/WI */
return (0U);
#else
uint32_t result;
__ASM volatile ("MRS %0, msplim" : "=r" (result) );
return (result);
#endif
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Get Main Stack Pointer Limit (non-secure)
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence zero is returned always.
\details Returns the current value of the non-secure Main Stack Pointer Limit(MSPLIM) when in secure state.
\return MSPLIM Register value
*/
__STATIC_FORCEINLINE uint32_t __TZ_get_MSPLIM_NS(void)
{
#if (!(defined (__ARM_ARCH_8M_MAIN__ ) && (__ARM_ARCH_8M_MAIN__ == 1)) && \
!(defined (__ARM_ARCH_8_1M_MAIN__ ) && (__ARM_ARCH_8_1M_MAIN__ == 1)))
/* without main extensions, the non-secure MSPLIM is RAZ/WI */
return (0U);
#else
uint32_t result;
__ASM volatile ("MRS %0, msplim_ns" : "=r" (result) );
return (result);
#endif
}
#endif
/**
\brief Set Main Stack Pointer Limit
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence the write is silently ignored.
\details Assigns the given value to the Main Stack Pointer Limit (MSPLIM).
\param [in] MainStackPtrLimit Main Stack Pointer Limit value to set
*/
__STATIC_FORCEINLINE void __set_MSPLIM(uint32_t MainStackPtrLimit)
{
#if (!(defined (__ARM_ARCH_8M_MAIN__ ) && (__ARM_ARCH_8M_MAIN__ == 1)) && \
!(defined (__ARM_ARCH_8_1M_MAIN__ ) && (__ARM_ARCH_8_1M_MAIN__ == 1)) && \
(!defined (__ARM_FEATURE_CMSE ) || (__ARM_FEATURE_CMSE < 3)))
/* without main extensions, the non-secure MSPLIM is RAZ/WI */
(void)MainStackPtrLimit;
#else
__ASM volatile ("MSR msplim, %0" : : "r" (MainStackPtrLimit));
#endif
}
#if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3)
/**
\brief Set Main Stack Pointer Limit (non-secure)
Devices without ARMv8-M Main Extensions (i.e. Cortex-M23) lack the non-secure
Stack Pointer Limit register hence the write is silently ignored.
\details Assigns the given value to the non-secure Main Stack Pointer Limit (MSPLIM) when in secure state.
\param [in] MainStackPtrLimit Main Stack Pointer value to set
*/
__STATIC_FORCEINLINE void __TZ_set_MSPLIM_NS(uint32_t MainStackPtrLimit)
{
#if (!(defined (__ARM_ARCH_8M_MAIN__ ) && (__ARM_ARCH_8M_MAIN__ == 1)) && \
!(defined (__ARM_ARCH_8_1M_MAIN__ ) && (__ARM_ARCH_8_1M_MAIN__ == 1)))
/* without main extensions, the non-secure MSPLIM is RAZ/WI */
(void)MainStackPtrLimit;
#else
__ASM volatile ("MSR msplim_ns, %0" : : "r" (MainStackPtrLimit));
#endif
}
#endif
#endif /* (__ARM_ARCH >= 8) */
/*@} end of CMSIS_Core_RegAccFunctions */
#endif /* __CMSIS_GCC_M_H */

11
sdk/CMSIS/DSP/CMSIS_DSP_Include.cmake Normal file
View file

@ -0,0 +1,11 @@
# Add set(CONFIG_USE_CMSIS_DSP_Include true) in config.cmake to use this component
include_guard(GLOBAL)
message("${CMAKE_CURRENT_LIST_FILE} component is included.")
target_include_directories(${MCUX_SDK_PROJECT_NAME} PUBLIC
${CMAKE_CURRENT_LIST_DIR}/Include
${CMAKE_CURRENT_LIST_DIR}/PrivateInclude
)

54
sdk/CMSIS/DSP/CMSIS_DSP_Source.cmake Normal file
View file

@ -0,0 +1,54 @@
# Add set(CONFIG_USE_CMSIS_DSP_Source true) in config.cmake to use this component
include_guard(GLOBAL)
message("${CMAKE_CURRENT_LIST_FILE} component is included.")
target_sources(${MCUX_SDK_PROJECT_NAME} PRIVATE
${CMAKE_CURRENT_LIST_DIR}/Source/BasicMathFunctions/BasicMathFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/BasicMathFunctions/BasicMathFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/BayesFunctions/BayesFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/BayesFunctions/BayesFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/CommonTables/CommonTables.c
${CMAKE_CURRENT_LIST_DIR}/Source/CommonTables/CommonTablesF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/ComplexMathFunctions/ComplexMathFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/ComplexMathFunctions/ComplexMathFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/ControllerFunctions/ControllerFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/DistanceFunctions/DistanceFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/DistanceFunctions/DistanceFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/FastMathFunctions/FastMathFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/FastMathFunctions/FastMathFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/FilteringFunctions/FilteringFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/FilteringFunctions/FilteringFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/InterpolationFunctions/InterpolationFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/InterpolationFunctions/InterpolationFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/MatrixFunctions/MatrixFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/MatrixFunctions/MatrixFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/QuaternionMathFunctions/QuaternionMathFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/SVMFunctions/SVMFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/SVMFunctions/SVMFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/StatisticsFunctions/StatisticsFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/StatisticsFunctions/StatisticsFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/SupportFunctions/SupportFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/SupportFunctions/SupportFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/TransformFunctions/TransformFunctions.c
${CMAKE_CURRENT_LIST_DIR}/Source/TransformFunctions/TransformFunctionsF16.c
${CMAKE_CURRENT_LIST_DIR}/Source/WindowFunctions/WindowFunctions.c
)
target_include_directories(${MCUX_SDK_PROJECT_NAME} PUBLIC
${CMAKE_CURRENT_LIST_DIR}/Include
${CMAKE_CURRENT_LIST_DIR}/PrivateInclude
${CMAKE_CURRENT_LIST_DIR}/Source/DistanceFunctions
)
if(CONFIG_USE_COMPONENT_CONFIGURATION)
message("===>Import configuration from ${CMAKE_CURRENT_LIST_FILE}")
target_compile_definitions(${MCUX_SDK_PROJECT_NAME} PUBLIC
-DDISABLEFLOAT16
)
endif()

318
sdk/CMSIS/DSP/Include/arm_common_tables.h Normal file
View file

@ -0,0 +1,318 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_common_tables.h
* Description: Extern declaration for common tables
*
* @version V1.10.0
* @date 08 July 2021
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_COMMON_TABLES_H
#define ARM_COMMON_TABLES_H
#include "arm_math_types.h"
#include "dsp/fast_math_functions.h"
#ifdef __cplusplus
extern "C"
{
#endif
/* Double Precision Float CFFT twiddles */
extern const uint16_t armBitRevTable[1024];
extern const uint64_t twiddleCoefF64_16[32];
extern const uint64_t twiddleCoefF64_32[64];
extern const uint64_t twiddleCoefF64_64[128];
extern const uint64_t twiddleCoefF64_128[256];
extern const uint64_t twiddleCoefF64_256[512];
extern const uint64_t twiddleCoefF64_512[1024];
extern const uint64_t twiddleCoefF64_1024[2048];
extern const uint64_t twiddleCoefF64_2048[4096];
extern const uint64_t twiddleCoefF64_4096[8192];
extern const float32_t twiddleCoef_16[32];
extern const float32_t twiddleCoef_32[64];
extern const float32_t twiddleCoef_64[128];
extern const float32_t twiddleCoef_128[256];
extern const float32_t twiddleCoef_256[512];
extern const float32_t twiddleCoef_512[1024];
extern const float32_t twiddleCoef_1024[2048];
extern const float32_t twiddleCoef_2048[4096];
extern const float32_t twiddleCoef_4096[8192];
#define twiddleCoef twiddleCoef_4096
/* Q31 */
extern const q31_t twiddleCoef_16_q31[24];
extern const q31_t twiddleCoef_32_q31[48];
extern const q31_t twiddleCoef_64_q31[96];
extern const q31_t twiddleCoef_128_q31[192];
extern const q31_t twiddleCoef_256_q31[384];
extern const q31_t twiddleCoef_512_q31[768];
extern const q31_t twiddleCoef_1024_q31[1536];
extern const q31_t twiddleCoef_2048_q31[3072];
extern const q31_t twiddleCoef_4096_q31[6144];
extern const q15_t twiddleCoef_16_q15[24];
extern const q15_t twiddleCoef_32_q15[48];
extern const q15_t twiddleCoef_64_q15[96];
extern const q15_t twiddleCoef_128_q15[192];
extern const q15_t twiddleCoef_256_q15[384];
extern const q15_t twiddleCoef_512_q15[768];
extern const q15_t twiddleCoef_1024_q15[1536];
extern const q15_t twiddleCoef_2048_q15[3072];
extern const q15_t twiddleCoef_4096_q15[6144];
/* Double Precision Float RFFT twiddles */
extern const uint64_t twiddleCoefF64_rfft_32[32];
extern const uint64_t twiddleCoefF64_rfft_64[64];
extern const uint64_t twiddleCoefF64_rfft_128[128];
extern const uint64_t twiddleCoefF64_rfft_256[256];
extern const uint64_t twiddleCoefF64_rfft_512[512];
extern const uint64_t twiddleCoefF64_rfft_1024[1024];
extern const uint64_t twiddleCoefF64_rfft_2048[2048];
extern const uint64_t twiddleCoefF64_rfft_4096[4096];
extern const float32_t twiddleCoef_rfft_32[32];
extern const float32_t twiddleCoef_rfft_64[64];
extern const float32_t twiddleCoef_rfft_128[128];
extern const float32_t twiddleCoef_rfft_256[256];
extern const float32_t twiddleCoef_rfft_512[512];
extern const float32_t twiddleCoef_rfft_1024[1024];
extern const float32_t twiddleCoef_rfft_2048[2048];
extern const float32_t twiddleCoef_rfft_4096[4096];
/* Double precision floating-point bit reversal tables */
#define ARMBITREVINDEXTABLEF64_16_TABLE_LENGTH ((uint16_t)12)
extern const uint16_t armBitRevIndexTableF64_16[ARMBITREVINDEXTABLEF64_16_TABLE_LENGTH];
#define ARMBITREVINDEXTABLEF64_32_TABLE_LENGTH ((uint16_t)24)
extern const uint16_t armBitRevIndexTableF64_32[ARMBITREVINDEXTABLEF64_32_TABLE_LENGTH];
#define ARMBITREVINDEXTABLEF64_64_TABLE_LENGTH ((uint16_t)56)
extern const uint16_t armBitRevIndexTableF64_64[ARMBITREVINDEXTABLEF64_64_TABLE_LENGTH];
#define ARMBITREVINDEXTABLEF64_128_TABLE_LENGTH ((uint16_t)112)
extern const uint16_t armBitRevIndexTableF64_128[ARMBITREVINDEXTABLEF64_128_TABLE_LENGTH];
#define ARMBITREVINDEXTABLEF64_256_TABLE_LENGTH ((uint16_t)240)
extern const uint16_t armBitRevIndexTableF64_256[ARMBITREVINDEXTABLEF64_256_TABLE_LENGTH];
#define ARMBITREVINDEXTABLEF64_512_TABLE_LENGTH ((uint16_t)480)
extern const uint16_t armBitRevIndexTableF64_512[ARMBITREVINDEXTABLEF64_512_TABLE_LENGTH];
#define ARMBITREVINDEXTABLEF64_1024_TABLE_LENGTH ((uint16_t)992)
extern const uint16_t armBitRevIndexTableF64_1024[ARMBITREVINDEXTABLEF64_1024_TABLE_LENGTH];
#define ARMBITREVINDEXTABLEF64_2048_TABLE_LENGTH ((uint16_t)1984)
extern const uint16_t armBitRevIndexTableF64_2048[ARMBITREVINDEXTABLEF64_2048_TABLE_LENGTH];
#define ARMBITREVINDEXTABLEF64_4096_TABLE_LENGTH ((uint16_t)4032)
extern const uint16_t armBitRevIndexTableF64_4096[ARMBITREVINDEXTABLEF64_4096_TABLE_LENGTH];
/* floating-point bit reversal tables */
#define ARMBITREVINDEXTABLE_16_TABLE_LENGTH ((uint16_t)20)
extern const uint16_t armBitRevIndexTable16[ARMBITREVINDEXTABLE_16_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_32_TABLE_LENGTH ((uint16_t)48)
extern const uint16_t armBitRevIndexTable32[ARMBITREVINDEXTABLE_32_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_64_TABLE_LENGTH ((uint16_t)56)
extern const uint16_t armBitRevIndexTable64[ARMBITREVINDEXTABLE_64_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_128_TABLE_LENGTH ((uint16_t)208)
extern const uint16_t armBitRevIndexTable128[ARMBITREVINDEXTABLE_128_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_256_TABLE_LENGTH ((uint16_t)440)
extern const uint16_t armBitRevIndexTable256[ARMBITREVINDEXTABLE_256_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_512_TABLE_LENGTH ((uint16_t)448)
extern const uint16_t armBitRevIndexTable512[ARMBITREVINDEXTABLE_512_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_1024_TABLE_LENGTH ((uint16_t)1800)
extern const uint16_t armBitRevIndexTable1024[ARMBITREVINDEXTABLE_1024_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_2048_TABLE_LENGTH ((uint16_t)3808)
extern const uint16_t armBitRevIndexTable2048[ARMBITREVINDEXTABLE_2048_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_4096_TABLE_LENGTH ((uint16_t)4032)
extern const uint16_t armBitRevIndexTable4096[ARMBITREVINDEXTABLE_4096_TABLE_LENGTH];
/* fixed-point bit reversal tables */
#define ARMBITREVINDEXTABLE_FIXED_16_TABLE_LENGTH ((uint16_t)12)
extern const uint16_t armBitRevIndexTable_fixed_16[ARMBITREVINDEXTABLE_FIXED_16_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_FIXED_32_TABLE_LENGTH ((uint16_t)24)
extern const uint16_t armBitRevIndexTable_fixed_32[ARMBITREVINDEXTABLE_FIXED_32_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_FIXED_64_TABLE_LENGTH ((uint16_t)56)
extern const uint16_t armBitRevIndexTable_fixed_64[ARMBITREVINDEXTABLE_FIXED_64_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_FIXED_128_TABLE_LENGTH ((uint16_t)112)
extern const uint16_t armBitRevIndexTable_fixed_128[ARMBITREVINDEXTABLE_FIXED_128_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_FIXED_256_TABLE_LENGTH ((uint16_t)240)
extern const uint16_t armBitRevIndexTable_fixed_256[ARMBITREVINDEXTABLE_FIXED_256_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_FIXED_512_TABLE_LENGTH ((uint16_t)480)
extern const uint16_t armBitRevIndexTable_fixed_512[ARMBITREVINDEXTABLE_FIXED_512_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_FIXED_1024_TABLE_LENGTH ((uint16_t)992)
extern const uint16_t armBitRevIndexTable_fixed_1024[ARMBITREVINDEXTABLE_FIXED_1024_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_FIXED_2048_TABLE_LENGTH ((uint16_t)1984)
extern const uint16_t armBitRevIndexTable_fixed_2048[ARMBITREVINDEXTABLE_FIXED_2048_TABLE_LENGTH];
#define ARMBITREVINDEXTABLE_FIXED_4096_TABLE_LENGTH ((uint16_t)4032)
extern const uint16_t armBitRevIndexTable_fixed_4096[ARMBITREVINDEXTABLE_FIXED_4096_TABLE_LENGTH];
extern const float32_t realCoefA[8192];
extern const float32_t realCoefB[8192];
extern const q31_t realCoefAQ31[8192];
extern const q31_t realCoefBQ31[8192];
extern const q15_t realCoefAQ15[8192];
extern const q15_t realCoefBQ15[8192];
extern const float32_t Weights_128[256];
extern const float32_t cos_factors_128[128];
extern const float32_t Weights_512[1024];
extern const float32_t cos_factors_512[512];
extern const float32_t Weights_2048[4096];
extern const float32_t cos_factors_2048[2048];
extern const float32_t Weights_8192[16384];
extern const float32_t cos_factors_8192[8192];
extern const q15_t WeightsQ15_128[256];
extern const q15_t cos_factorsQ15_128[128];
extern const q15_t WeightsQ15_512[1024];
extern const q15_t cos_factorsQ15_512[512];
extern const q15_t WeightsQ15_2048[4096];
extern const q15_t cos_factorsQ15_2048[2048];
extern const q15_t WeightsQ15_8192[16384];
extern const q15_t cos_factorsQ15_8192[8192];
extern const q31_t WeightsQ31_128[256];
extern const q31_t cos_factorsQ31_128[128];
extern const q31_t WeightsQ31_512[1024];
extern const q31_t cos_factorsQ31_512[512];
extern const q31_t WeightsQ31_2048[4096];
extern const q31_t cos_factorsQ31_2048[2048];
extern const q31_t WeightsQ31_8192[16384];
extern const q31_t cos_factorsQ31_8192[8192];
extern const q15_t armRecipTableQ15[64];
extern const q31_t armRecipTableQ31[64];
/* Tables for Fast Math Sine and Cosine */
extern const float32_t sinTable_f32[FAST_MATH_TABLE_SIZE + 1];
extern const q31_t sinTable_q31[FAST_MATH_TABLE_SIZE + 1];
extern const q15_t sinTable_q15[FAST_MATH_TABLE_SIZE + 1];
/* Accurate scalar sqrt */
extern const q31_t sqrt_initial_lut_q31[32];
extern const q15_t sqrt_initial_lut_q15[16];
#if (defined(ARM_MATH_MVEI) || defined(ARM_MATH_HELIUM)) && !defined(ARM_MATH_AUTOVECTORIZE)
extern const q15_t sqrtTable_Q15[256];
extern const q31_t sqrtTable_Q31[256];
extern const unsigned char hwLUT[256];
#endif
#if (defined(ARM_MATH_MVEF) || defined(ARM_MATH_HELIUM)) && !defined(ARM_MATH_AUTOVECTORIZE)
extern const float32_t exp_tab[8];
extern const float32_t __logf_lut_f32[8];
#endif
#ifdef __cplusplus
}
#endif
#endif /* ARM_COMMON_TABLES_H */

95
sdk/CMSIS/DSP/Include/arm_common_tables_f16.h Normal file
View file

@ -0,0 +1,95 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_common_tables_f16.h
* Description: Extern declaration for common tables
*
* @version V1.10.0
* @date 08 July 2021
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_COMMON_TABLES_F16_H
#define ARM_COMMON_TABLES_F16_H
#include "arm_math_types_f16.h"
#ifdef __cplusplus
extern "C"
{
#endif
/* F16 */
#if !defined(__CC_ARM) && defined(ARM_FLOAT16_SUPPORTED)
extern const float16_t twiddleCoefF16_16[32];
extern const float16_t twiddleCoefF16_32[64];
extern const float16_t twiddleCoefF16_64[128];
extern const float16_t twiddleCoefF16_128[256];
extern const float16_t twiddleCoefF16_256[512];
extern const float16_t twiddleCoefF16_512[1024];
extern const float16_t twiddleCoefF16_1024[2048];
extern const float16_t twiddleCoefF16_2048[4096];
extern const float16_t twiddleCoefF16_4096[8192];
#define twiddleCoefF16 twiddleCoefF16_4096
extern const float16_t twiddleCoefF16_rfft_32[32];
extern const float16_t twiddleCoefF16_rfft_64[64];
extern const float16_t twiddleCoefF16_rfft_128[128];
extern const float16_t twiddleCoefF16_rfft_256[256];
extern const float16_t twiddleCoefF16_rfft_512[512];
extern const float16_t twiddleCoefF16_rfft_1024[1024];
extern const float16_t twiddleCoefF16_rfft_2048[2048];
extern const float16_t twiddleCoefF16_rfft_4096[4096];
#endif /* ARMAC5 */
#if !defined(__CC_ARM) && defined(ARM_FLOAT16_SUPPORTED)
#if (defined(ARM_MATH_MVEF) || defined(ARM_MATH_HELIUM)) && !defined(ARM_MATH_AUTOVECTORIZE)
extern const float16_t exp_tab_f16[8];
extern const float16_t __logf_lut_f16[8];
#endif
#endif
#ifdef __cplusplus
}
#endif
#endif /* _ARM_COMMON_TABLES_F16_H */

86
sdk/CMSIS/DSP/Include/arm_const_structs.h Normal file
View file

@ -0,0 +1,86 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_const_structs.h
* Description: Constant structs that are initialized for user convenience.
* For example, some can be given as arguments to the arm_cfft_f32() function.
*
* @version V1.10.0
* @date 08 July 2021
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_CONST_STRUCTS_H
#define ARM_CONST_STRUCTS_H
#include "arm_math_types.h"
#include "arm_common_tables.h"
#include "dsp/transform_functions.h"
#ifdef __cplusplus
extern "C"
{
#endif
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len16;
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len32;
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len64;
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len128;
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len256;
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len512;
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len1024;
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len2048;
extern const arm_cfft_instance_f64 arm_cfft_sR_f64_len4096;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len16;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len32;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len64;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len128;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len256;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len512;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len1024;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len2048;
extern const arm_cfft_instance_f32 arm_cfft_sR_f32_len4096;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len16;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len32;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len64;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len128;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len256;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len512;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len1024;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len2048;
extern const arm_cfft_instance_q31 arm_cfft_sR_q31_len4096;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len16;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len32;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len64;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len128;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len256;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len512;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len1024;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len2048;
extern const arm_cfft_instance_q15 arm_cfft_sR_q15_len4096;
#ifdef __cplusplus
}
#endif
#endif

59
sdk/CMSIS/DSP/Include/arm_const_structs_f16.h Normal file
View file

@ -0,0 +1,59 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_const_structs_f16.h
* Description: Constant structs that are initialized for user convenience.
* For example, some can be given as arguments to the arm_cfft_f16() function.
*
* @version V1.10.0
* @date 08 July 2021
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_CONST_STRUCTS_F16_H
#define ARM_CONST_STRUCTS_F16_H
#include "arm_math_types_f16.h"
#include "arm_common_tables.h"
#include "arm_common_tables_f16.h"
#include "dsp/transform_functions_f16.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if !defined(__CC_ARM) && defined(ARM_FLOAT16_SUPPORTED)
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len16;
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len32;
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len64;
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len128;
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len256;
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len512;
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len1024;
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len2048;
extern const arm_cfft_instance_f16 arm_cfft_sR_f16_len4096;
#endif
#ifdef __cplusplus
}
#endif
#endif

749
sdk/CMSIS/DSP/Include/arm_helium_utils.h Normal file
View file

@ -0,0 +1,749 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_helium_utils.h
* Description: Utility functions for Helium development
*
* @version V1.10.0
* @date 08 July 2021
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_UTILS_HELIUM_H_
#define ARM_UTILS_HELIUM_H_
#ifdef __cplusplus
extern "C"
{
#endif
/***************************************
Definitions available for MVEF and MVEI
***************************************/
#if (defined (ARM_MATH_HELIUM) || defined(ARM_MATH_MVEF) || defined(ARM_MATH_MVEI)) && !defined(ARM_MATH_AUTOVECTORIZE)
#define INACTIVELANE 0 /* inactive lane content */
#endif /* defined (ARM_MATH_HELIUM) || defined(ARM_MATH_MVEF) || defined(ARM_MATH_MVEI) */
/***************************************
Definitions available for MVEF only
***************************************/
#if (defined (ARM_MATH_HELIUM) || defined(ARM_MATH_MVEF)) && !defined(ARM_MATH_AUTOVECTORIZE)
__STATIC_FORCEINLINE float32_t vecAddAcrossF32Mve(float32x4_t in)
{
float32_t acc;
acc = vgetq_lane(in, 0) + vgetq_lane(in, 1) +
vgetq_lane(in, 2) + vgetq_lane(in, 3);
return acc;
}
/* newton initial guess */
#define INVSQRT_MAGIC_F32 0x5f3759df
#define INV_NEWTON_INIT_F32 0x7EF127EA
#define INVSQRT_NEWTON_MVE_F32(invSqrt, xHalf, xStart)\
{ \
float32x4_t tmp; \
\
/* tmp = xhalf * x * x */ \
tmp = vmulq(xStart, xStart); \
tmp = vmulq(tmp, xHalf); \
/* (1.5f - xhalf * x * x) */ \
tmp = vsubq(vdupq_n_f32(1.5f), tmp); \
/* x = x*(1.5f-xhalf*x*x); */ \
invSqrt = vmulq(tmp, xStart); \
}
#endif /* defined (ARM_MATH_HELIUM) || defined(ARM_MATH_MVEF) */
/***************************************
Definitions available for f16 datatype with HW acceleration only
***************************************/
#if defined(ARM_FLOAT16_SUPPORTED)
#if defined (ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
__STATIC_FORCEINLINE float16_t vecAddAcrossF16Mve(float16x8_t in)
{
float16x8_t tmpVec;
_Float16 acc;
tmpVec = (float16x8_t) vrev32q_s16((int16x8_t) in);
in = vaddq_f16(tmpVec, in);
tmpVec = (float16x8_t) vrev64q_s32((int32x4_t) in);
in = vaddq_f16(tmpVec, in);
acc = (_Float16)vgetq_lane_f16(in, 0) + (_Float16)vgetq_lane_f16(in, 4);
return acc;
}
__STATIC_FORCEINLINE float16x8_t __mve_cmplx_sum_intra_vec_f16(
float16x8_t vecIn)
{
float16x8_t vecTmp, vecOut;
uint32_t tmp = 0;
vecTmp = (float16x8_t) vrev64q_s32((int32x4_t) vecIn);
// TO TRACK : using canonical addition leads to unefficient code generation for f16
// vecTmp = vecTmp + vecAccCpx0;
/*
* Compute
* re0+re1 | im0+im1 | re0+re1 | im0+im1
* re2+re3 | im2+im3 | re2+re3 | im2+im3
*/
vecTmp = vaddq_f16(vecTmp, vecIn);
vecOut = vecTmp;
/*
* shift left, random tmp insertion in bottom
*/
vecOut = vreinterpretq_f16_s32(vshlcq_s32(vreinterpretq_s32_f16(vecOut) , &tmp, 32));
/*
* Compute:
* DONTCARE | DONTCARE | re0+re1+re0+re1 |im0+im1+im0+im1
* re0+re1+re2+re3 | im0+im1+im2+im3 | re2+re3+re2+re3 |im2+im3+im2+im3
*/
vecOut = vaddq_f16(vecOut, vecTmp);
/*
* Cmplx sum is in 4rd & 5th f16 elt
* return full vector
*/
return vecOut;
}
#define mve_cmplx_sum_intra_r_i_f16(vec, Re, Im) \
{ \
float16x8_t vecOut = __mve_cmplx_sum_intra_vec_f16(vec); \
Re = vgetq_lane(vecOut, 4); \
Im = vgetq_lane(vecOut, 5); \
}
__STATIC_FORCEINLINE void mve_cmplx_sum_intra_vec_f16(
float16x8_t vecIn,
float16_t *pOut)
{
float16x8_t vecOut = __mve_cmplx_sum_intra_vec_f16(vecIn);
/*
* Cmplx sum is in 4rd & 5th f16 elt
* use 32-bit extraction
*/
*(float32_t *) pOut = ((float32x4_t) vecOut)[2];
}
#define INVSQRT_MAGIC_F16 0x59ba /* ( 0x1ba = 0x3759df >> 13) */
/* canonical version of INVSQRT_NEWTON_MVE_F16 leads to bad performance */
#define INVSQRT_NEWTON_MVE_F16(invSqrt, xHalf, xStart) \
{ \
float16x8_t tmp; \
\
/* tmp = xhalf * x * x */ \
tmp = vmulq(xStart, xStart); \
tmp = vmulq(tmp, xHalf); \
/* (1.5f - xhalf * x * x) */ \
tmp = vsubq(vdupq_n_f16((float16_t)1.5), tmp); \
/* x = x*(1.5f-xhalf*x*x); */ \
invSqrt = vmulq(tmp, xStart); \
}
#endif
#endif
/***************************************
Definitions available for MVEI and MVEF only
***************************************/
#if (defined (ARM_MATH_HELIUM) || defined(ARM_MATH_MVEF) || defined(ARM_MATH_MVEI)) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Following functions are used to transpose matrix in f32 and q31 cases */
__STATIC_INLINE arm_status arm_mat_trans_32bit_2x2_mve(
uint32_t * pDataSrc,
uint32_t * pDataDest)
{
static const uint32x4_t vecOffs = { 0, 2, 1, 3 };
/*
*
* | 0 1 | => | 0 2 |
* | 2 3 | | 1 3 |
*
*/
uint32x4_t vecIn = vldrwq_u32((uint32_t const *)pDataSrc);
vstrwq_scatter_shifted_offset_u32(pDataDest, vecOffs, vecIn);
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_trans_32bit_3x3_mve(
uint32_t * pDataSrc,
uint32_t * pDataDest)
{
const uint32x4_t vecOffs1 = { 0, 3, 6, 1};
const uint32x4_t vecOffs2 = { 4, 7, 2, 5};
/*
*
* | 0 1 2 | | 0 3 6 | 4 x 32 flattened version | 0 3 6 1 |
* | 3 4 5 | => | 1 4 7 | => | 4 7 2 5 |
* | 6 7 8 | | 2 5 8 | (row major) | 8 . . . |
*
*/
uint32x4_t vecIn1 = vldrwq_u32((uint32_t const *) pDataSrc);
uint32x4_t vecIn2 = vldrwq_u32((uint32_t const *) &pDataSrc[4]);
vstrwq_scatter_shifted_offset_u32(pDataDest, vecOffs1, vecIn1);
vstrwq_scatter_shifted_offset_u32(pDataDest, vecOffs2, vecIn2);
pDataDest[8] = pDataSrc[8];
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_trans_32bit_4x4_mve(uint32_t * pDataSrc, uint32_t * pDataDest)
{
/*
* 4x4 Matrix transposition
* is 4 x de-interleave operation
*
* 0 1 2 3 0 4 8 12
* 4 5 6 7 1 5 9 13
* 8 9 10 11 2 6 10 14
* 12 13 14 15 3 7 11 15
*/
uint32x4x4_t vecIn;
vecIn = vld4q((uint32_t const *) pDataSrc);
vstrwq(pDataDest, vecIn.val[0]);
pDataDest += 4;
vstrwq(pDataDest, vecIn.val[1]);
pDataDest += 4;
vstrwq(pDataDest, vecIn.val[2]);
pDataDest += 4;
vstrwq(pDataDest, vecIn.val[3]);
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_trans_32bit_generic_mve(
uint16_t srcRows,
uint16_t srcCols,
uint32_t * pDataSrc,
uint32_t * pDataDest)
{
uint32x4_t vecOffs;
uint32_t i;
uint32_t blkCnt;
uint32_t const *pDataC;
uint32_t *pDataDestR;
uint32x4_t vecIn;
vecOffs = vidupq_u32((uint32_t)0, 1);
vecOffs = vecOffs * srcCols;
i = srcCols;
do
{
pDataC = (uint32_t const *) pDataSrc;
pDataDestR = pDataDest;
blkCnt = srcRows >> 2;
while (blkCnt > 0U)
{
vecIn = vldrwq_gather_shifted_offset_u32(pDataC, vecOffs);
vstrwq(pDataDestR, vecIn);
pDataDestR += 4;
pDataC = pDataC + srcCols * 4;
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
}
/*
* tail
*/
blkCnt = srcRows & 3;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp32q(blkCnt);
vecIn = vldrwq_gather_shifted_offset_u32(pDataC, vecOffs);
vstrwq_p(pDataDestR, vecIn, p0);
}
pDataSrc += 1;
pDataDest += srcRows;
}
while (--i);
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_cmplx_trans_32bit(
uint16_t srcRows,
uint16_t srcCols,
uint32_t *pDataSrc,
uint16_t dstRows,
uint16_t dstCols,
uint32_t *pDataDest)
{
uint32_t i;
uint32_t const *pDataC;
uint32_t *pDataRow;
uint32_t *pDataDestR, *pDataDestRow;
uint32x4_t vecOffsRef, vecOffsCur;
uint32_t blkCnt;
uint32x4_t vecIn;
#ifdef ARM_MATH_MATRIX_CHECK
/*
* Check for matrix mismatch condition
*/
if ((srcRows != dstCols) || (srcCols != dstRows))
{
/*
* Set status as ARM_MATH_SIZE_MISMATCH
*/
return ARM_MATH_SIZE_MISMATCH;
}
#else
(void)dstRows;
(void)dstCols;
#endif
/* 2x2, 3x3 and 4x4 specialization to be added */
vecOffsRef[0] = 0;
vecOffsRef[1] = 1;
vecOffsRef[2] = srcCols << 1;
vecOffsRef[3] = (srcCols << 1) + 1;
pDataRow = pDataSrc;
pDataDestRow = pDataDest;
i = srcCols;
do
{
pDataC = (uint32_t const *) pDataRow;
pDataDestR = pDataDestRow;
vecOffsCur = vecOffsRef;
blkCnt = (srcRows * CMPLX_DIM) >> 2;
while (blkCnt > 0U)
{
vecIn = vldrwq_gather_shifted_offset(pDataC, vecOffsCur);
vstrwq(pDataDestR, vecIn);
pDataDestR += 4;
vecOffsCur = vaddq(vecOffsCur, (srcCols << 2));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
}
/*
* tail
* (will be merged thru tail predication)
*/
blkCnt = (srcRows * CMPLX_DIM) & 3;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp32q(blkCnt);
vecIn = vldrwq_gather_shifted_offset(pDataC, vecOffsCur);
vstrwq_p(pDataDestR, vecIn, p0);
}
pDataRow += CMPLX_DIM;
pDataDestRow += (srcRows * CMPLX_DIM);
}
while (--i);
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_trans_16bit_2x2(uint16_t * pDataSrc, uint16_t * pDataDest)
{
pDataDest[0] = pDataSrc[0];
pDataDest[3] = pDataSrc[3];
pDataDest[2] = pDataSrc[1];
pDataDest[1] = pDataSrc[2];
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_trans_16bit_3x3_mve(uint16_t * pDataSrc, uint16_t * pDataDest)
{
static const uint16_t stridesTr33[8] = { 0, 3, 6, 1, 4, 7, 2, 5 };
uint16x8_t vecOffs1;
uint16x8_t vecIn1;
/*
*
* | 0 1 2 | | 0 3 6 | 8 x 16 flattened version | 0 3 6 1 4 7 2 5 |
* | 3 4 5 | => | 1 4 7 | => | 8 . . . . . . . |
* | 6 7 8 | | 2 5 8 | (row major)
*
*/
vecOffs1 = vldrhq_u16((uint16_t const *) stridesTr33);
vecIn1 = vldrhq_u16((uint16_t const *) pDataSrc);
vstrhq_scatter_shifted_offset_u16(pDataDest, vecOffs1, vecIn1);
pDataDest[8] = pDataSrc[8];
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_trans_16bit_4x4_mve(uint16_t * pDataSrc, uint16_t * pDataDest)
{
static const uint16_t stridesTr44_1[8] = { 0, 4, 8, 12, 1, 5, 9, 13 };
static const uint16_t stridesTr44_2[8] = { 2, 6, 10, 14, 3, 7, 11, 15 };
uint16x8_t vecOffs1, vecOffs2;
uint16x8_t vecIn1, vecIn2;
uint16_t const * pDataSrcVec = (uint16_t const *) pDataSrc;
/*
* 4x4 Matrix transposition
*
* | 0 1 2 3 | | 0 4 8 12 | 8 x 16 flattened version
* | 4 5 6 7 | => | 1 5 9 13 | => [0 4 8 12 1 5 9 13]
* | 8 9 10 11 | | 2 6 10 14 | [2 6 10 14 3 7 11 15]
* | 12 13 14 15 | | 3 7 11 15 |
*/
vecOffs1 = vldrhq_u16((uint16_t const *) stridesTr44_1);
vecOffs2 = vldrhq_u16((uint16_t const *) stridesTr44_2);
vecIn1 = vldrhq_u16(pDataSrcVec);
pDataSrcVec += 8;
vecIn2 = vldrhq_u16(pDataSrcVec);
vstrhq_scatter_shifted_offset_u16(pDataDest, vecOffs1, vecIn1);
vstrhq_scatter_shifted_offset_u16(pDataDest, vecOffs2, vecIn2);
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_trans_16bit_generic(
uint16_t srcRows,
uint16_t srcCols,
uint16_t * pDataSrc,
uint16_t * pDataDest)
{
uint16x8_t vecOffs;
uint32_t i;
uint32_t blkCnt;
uint16_t const *pDataC;
uint16_t *pDataDestR;
uint16x8_t vecIn;
vecOffs = vidupq_u16((uint32_t)0, 1);
vecOffs = vecOffs * srcCols;
i = srcCols;
while(i > 0U)
{
pDataC = (uint16_t const *) pDataSrc;
pDataDestR = pDataDest;
blkCnt = srcRows >> 3;
while (blkCnt > 0U)
{
vecIn = vldrhq_gather_shifted_offset_u16(pDataC, vecOffs);
vstrhq_u16(pDataDestR, vecIn);
pDataDestR += 8;
pDataC = pDataC + srcCols * 8;
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
}
/*
* tail
*/
blkCnt = srcRows & 7;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp16q(blkCnt);
vecIn = vldrhq_gather_shifted_offset_u16(pDataC, vecOffs);
vstrhq_p_u16(pDataDestR, vecIn, p0);
}
pDataSrc += 1;
pDataDest += srcRows;
i--;
}
return (ARM_MATH_SUCCESS);
}
__STATIC_INLINE arm_status arm_mat_cmplx_trans_16bit(
uint16_t srcRows,
uint16_t srcCols,
uint16_t *pDataSrc,
uint16_t dstRows,
uint16_t dstCols,
uint16_t *pDataDest)
{
static const uint16_t loadCmplxCol[8] = { 0, 0, 1, 1, 2, 2, 3, 3 };
int i;
uint16x8_t vecOffsRef, vecOffsCur;
uint16_t const *pDataC;
uint16_t *pDataRow;
uint16_t *pDataDestR, *pDataDestRow;
uint32_t blkCnt;
uint16x8_t vecIn;
#ifdef ARM_MATH_MATRIX_CHECK
/*
* Check for matrix mismatch condition
*/
if ((srcRows != dstCols) || (srcCols != dstRows))
{
/*
* Set status as ARM_MATH_SIZE_MISMATCH
*/
return ARM_MATH_SIZE_MISMATCH;
}
#else
(void)dstRows;
(void)dstCols;
#endif
/*
* 2x2, 3x3 and 4x4 specialization to be added
*/
/*
* build [0, 1, 2xcol, 2xcol+1, 4xcol, 4xcol+1, 6xcol, 6xcol+1]
*/
vecOffsRef = vldrhq_u16((uint16_t const *) loadCmplxCol);
vecOffsRef = vmulq(vecOffsRef, (uint16_t) (srcCols * CMPLX_DIM))
+ viwdupq_u16((uint32_t)0, (uint16_t) 2, 1);
pDataRow = pDataSrc;
pDataDestRow = pDataDest;
i = srcCols;
do
{
pDataC = (uint16_t const *) pDataRow;
pDataDestR = pDataDestRow;
vecOffsCur = vecOffsRef;
blkCnt = (srcRows * CMPLX_DIM) >> 3;
while (blkCnt > 0U)
{
vecIn = vldrhq_gather_shifted_offset(pDataC, vecOffsCur);
vstrhq(pDataDestR, vecIn);
pDataDestR+= 8; // VEC_LANES_U16
vecOffsCur = vaddq(vecOffsCur, (srcCols << 3));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
}
/*
* tail
* (will be merged thru tail predication)
*/
blkCnt = (srcRows * CMPLX_DIM) & 0x7;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp16q(blkCnt);
vecIn = vldrhq_gather_shifted_offset(pDataC, vecOffsCur);
vstrhq_p(pDataDestR, vecIn, p0);
}
pDataRow += CMPLX_DIM;
pDataDestRow += (srcRows * CMPLX_DIM);
}
while (--i);
return (ARM_MATH_SUCCESS);
}
#endif /* MVEF and MVEI */
/***************************************
Definitions available for MVEI only
***************************************/
#if (defined (ARM_MATH_HELIUM) || defined(ARM_MATH_MVEI)) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_common_tables.h"
#define MVE_ASRL_SAT16(acc, shift) ((sqrshrl_sat48(acc, -(32-shift)) >> 32) & 0xffffffff)
#define MVE_ASRL_SAT32(acc, shift) ((sqrshrl(acc, -(32-shift)) >> 32) & 0xffffffff)
__STATIC_INLINE q31x4_t FAST_VSQRT_Q31(q31x4_t vecIn)
{
q63x2_t vecTmpLL;
q31x4_t vecTmp0, vecTmp1;
q31_t scale;
q63_t tmp64;
q31x4_t vecNrm, vecDst, vecIdx, vecSignBits;
vecSignBits = vclsq(vecIn);
vecSignBits = vbicq_n_s32(vecSignBits, 1);
/*
* in = in << no_of_sign_bits;
*/
vecNrm = vshlq(vecIn, vecSignBits);
/*
* index = in >> 24;
*/
vecIdx = vecNrm >> 24;
vecIdx = vecIdx << 1;
vecTmp0 = vldrwq_gather_shifted_offset_s32(sqrtTable_Q31, (uint32x4_t)vecIdx);
vecIdx = vecIdx + 1;
vecTmp1 = vldrwq_gather_shifted_offset_s32(sqrtTable_Q31, (uint32x4_t)vecIdx);
vecTmp1 = vqrdmulhq(vecTmp1, vecNrm);
vecTmp0 = vecTmp0 - vecTmp1;
vecTmp1 = vqrdmulhq(vecTmp0, vecTmp0);
vecTmp1 = vqrdmulhq(vecNrm, vecTmp1);
vecTmp1 = vdupq_n_s32(0x18000000) - vecTmp1;
vecTmp0 = vqrdmulhq(vecTmp0, vecTmp1);
vecTmpLL = vmullbq_int(vecNrm, vecTmp0);
/*
* scale elements 0, 2
*/
scale = 26 + (vecSignBits[0] >> 1);
tmp64 = asrl(vecTmpLL[0], scale);
vecDst[0] = (q31_t) tmp64;
scale = 26 + (vecSignBits[2] >> 1);
tmp64 = asrl(vecTmpLL[1], scale);
vecDst[2] = (q31_t) tmp64;
vecTmpLL = vmulltq_int(vecNrm, vecTmp0);
/*
* scale elements 1, 3
*/
scale = 26 + (vecSignBits[1] >> 1);
tmp64 = asrl(vecTmpLL[0], scale);
vecDst[1] = (q31_t) tmp64;
scale = 26 + (vecSignBits[3] >> 1);
tmp64 = asrl(vecTmpLL[1], scale);
vecDst[3] = (q31_t) tmp64;
/*
* set negative values to 0
*/
vecDst = vdupq_m(vecDst, 0, vcmpltq_n_s32(vecIn, 0));
return vecDst;
}
__STATIC_INLINE q15x8_t FAST_VSQRT_Q15(q15x8_t vecIn)
{
q31x4_t vecTmpLev, vecTmpLodd, vecSignL;
q15x8_t vecTmp0, vecTmp1;
q15x8_t vecNrm, vecDst, vecIdx, vecSignBits;
vecDst = vuninitializedq_s16();
vecSignBits = vclsq(vecIn);
vecSignBits = vbicq_n_s16(vecSignBits, 1);
/*
* in = in << no_of_sign_bits;
*/
vecNrm = vshlq(vecIn, vecSignBits);
vecIdx = vecNrm >> 8;
vecIdx = vecIdx << 1;
vecTmp0 = vldrhq_gather_shifted_offset_s16(sqrtTable_Q15, (uint16x8_t)vecIdx);
vecIdx = vecIdx + 1;
vecTmp1 = vldrhq_gather_shifted_offset_s16(sqrtTable_Q15, (uint16x8_t)vecIdx);
vecTmp1 = vqrdmulhq(vecTmp1, vecNrm);
vecTmp0 = vecTmp0 - vecTmp1;
vecTmp1 = vqrdmulhq(vecTmp0, vecTmp0);
vecTmp1 = vqrdmulhq(vecNrm, vecTmp1);
vecTmp1 = vdupq_n_s16(0x1800) - vecTmp1;
vecTmp0 = vqrdmulhq(vecTmp0, vecTmp1);
vecSignBits = vecSignBits >> 1;
vecTmpLev = vmullbq_int(vecNrm, vecTmp0);
vecTmpLodd = vmulltq_int(vecNrm, vecTmp0);
vecTmp0 = vecSignBits + 10;
/*
* negate sign to apply register based vshl
*/
vecTmp0 = -vecTmp0;
/*
* shift even elements
*/
vecSignL = vmovlbq(vecTmp0);
vecTmpLev = vshlq(vecTmpLev, vecSignL);
/*
* shift odd elements
*/
vecSignL = vmovltq(vecTmp0);
vecTmpLodd = vshlq(vecTmpLodd, vecSignL);
/*
* merge and narrow odd and even parts
*/
vecDst = vmovnbq_s32(vecDst, vecTmpLev);
vecDst = vmovntq_s32(vecDst, vecTmpLodd);
/*
* set negative values to 0
*/
vecDst = vdupq_m(vecDst, 0, vcmpltq_n_s16(vecIn, 0));
return vecDst;
}
#endif /* defined (ARM_MATH_HELIUM) || defined(ARM_MATH_MVEI) */
#ifdef __cplusplus
}
#endif
#endif

80
sdk/CMSIS/DSP/Include/arm_math.h Normal file
View file

@ -0,0 +1,80 @@
/******************************************************************************
* @file arm_math.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2021 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_MATH_H
#define ARM_MATH_H
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/basic_math_functions.h"
#include "dsp/interpolation_functions.h"
#include "dsp/bayes_functions.h"
#include "dsp/matrix_functions.h"
#include "dsp/complex_math_functions.h"
#include "dsp/statistics_functions.h"
#include "dsp/controller_functions.h"
#include "dsp/support_functions.h"
#include "dsp/distance_functions.h"
#include "dsp/svm_functions.h"
#include "dsp/fast_math_functions.h"
#include "dsp/transform_functions.h"
#include "dsp/filtering_functions.h"
#include "dsp/quaternion_math_functions.h"
#include "dsp/window_functions.h"
#ifdef __cplusplus
extern "C"
{
#endif
//#define TABLE_SPACING_Q31 0x400000
//#define TABLE_SPACING_Q15 0x80
#ifdef __cplusplus
}
#endif
#endif /* _ARM_MATH_H */
/**
*
* End of file.
*/

59
sdk/CMSIS/DSP/Include/arm_math_f16.h Normal file
View file

@ -0,0 +1,59 @@
/******************************************************************************
* @file arm_math_f16.h
* @brief Public header file for f16 function of the CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2021 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_MATH_F16_H
#define ARM_MATH_F16_H
#include "arm_math.h"
#ifdef __cplusplus
extern "C"
{
#endif
#include "arm_math_types_f16.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/basic_math_functions_f16.h"
#include "dsp/interpolation_functions_f16.h"
#include "dsp/bayes_functions_f16.h"
#include "dsp/matrix_functions_f16.h"
#include "dsp/complex_math_functions_f16.h"
#include "dsp/statistics_functions_f16.h"
#include "dsp/controller_functions_f16.h"
#include "dsp/support_functions_f16.h"
#include "dsp/distance_functions_f16.h"
#include "dsp/svm_functions_f16.h"
#include "dsp/fast_math_functions_f16.h"
#include "dsp/transform_functions_f16.h"
#include "dsp/filtering_functions_f16.h"
#ifdef __cplusplus
}
#endif
#endif /* _ARM_MATH_F16_H */

203
sdk/CMSIS/DSP/Include/arm_math_memory.h Normal file
View file

@ -0,0 +1,203 @@
/******************************************************************************
* @file arm_math_memory.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2021 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_MATH_MEMORY_H_
#define ARM_MATH_MEMORY_H_
#include "arm_math_types.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
@brief definition to read/write two 16 bit values.
@deprecated
*/
#if defined ( __CC_ARM )
#define __SIMD32_TYPE int32_t __packed
#elif defined ( __ARMCC_VERSION ) && ( __ARMCC_VERSION >= 6010050 )
#define __SIMD32_TYPE int32_t
#elif defined ( __GNUC__ )
#define __SIMD32_TYPE int32_t
#elif defined ( __ICCARM__ )
#define __SIMD32_TYPE int32_t __packed
#elif defined ( __TI_ARM__ )
#define __SIMD32_TYPE int32_t
#elif defined ( __CSMC__ )
#define __SIMD32_TYPE int32_t
#elif defined ( __TASKING__ )
#define __SIMD32_TYPE __un(aligned) int32_t
#elif defined(_MSC_VER )
#define __SIMD32_TYPE int32_t
#else
#error Unknown compiler
#endif
#define __SIMD32(addr) (*(__SIMD32_TYPE **) & (addr))
#define __SIMD32_CONST(addr) ( (__SIMD32_TYPE * ) (addr))
#define _SIMD32_OFFSET(addr) (*(__SIMD32_TYPE * ) (addr))
#define __SIMD64(addr) (*( int64_t **) & (addr))
/* SIMD replacement */
/**
@brief Read 2 Q15 from Q15 pointer.
@param[in] pQ15 points to input value
@return Q31 value
*/
__STATIC_FORCEINLINE q31_t read_q15x2 (
q15_t const * pQ15)
{
q31_t val;
#ifdef __ARM_FEATURE_UNALIGNED
memcpy (&val, pQ15, 4);
#else
val = (pQ15[1] << 16) | (pQ15[0] & 0x0FFFF) ;
#endif
return (val);
}
/**
@brief Read 2 Q15 from Q15 pointer and increment pointer afterwards.
@param[in] pQ15 points to input value
@return Q31 value
*/
#define read_q15x2_ia(pQ15) read_q15x2((*(pQ15) += 2) - 2)
/**
@brief Read 2 Q15 from Q15 pointer and decrement pointer afterwards.
@param[in] pQ15 points to input value
@return Q31 value
*/
#define read_q15x2_da(pQ15) read_q15x2((*(pQ15) -= 2) + 2)
/**
@brief Write 2 Q15 to Q15 pointer and increment pointer afterwards.
@param[in] pQ15 points to input value
@param[in] value Q31 value
*/
__STATIC_FORCEINLINE void write_q15x2_ia (
q15_t ** pQ15,
q31_t value)
{
q31_t val = value;
#ifdef __ARM_FEATURE_UNALIGNED
memcpy (*pQ15, &val, 4);
#else
(*pQ15)[0] = (q15_t)(val & 0x0FFFF);
(*pQ15)[1] = (q15_t)((val >> 16) & 0x0FFFF);
#endif
*pQ15 += 2;
}
/**
@brief Write 2 Q15 to Q15 pointer.
@param[in] pQ15 points to input value
@param[in] value Q31 value
*/
__STATIC_FORCEINLINE void write_q15x2 (
q15_t * pQ15,
q31_t value)
{
q31_t val = value;
#ifdef __ARM_FEATURE_UNALIGNED
memcpy (pQ15, &val, 4);
#else
pQ15[0] = (q15_t)(val & 0x0FFFF);
pQ15[1] = (q15_t)(val >> 16);
#endif
}
/**
@brief Read 4 Q7 from Q7 pointer
@param[in] pQ7 points to input value
@return Q31 value
*/
__STATIC_FORCEINLINE q31_t read_q7x4 (
q7_t const * pQ7)
{
q31_t val;
#ifdef __ARM_FEATURE_UNALIGNED
memcpy (&val, pQ7, 4);
#else
val =((pQ7[3] & 0x0FF) << 24) | ((pQ7[2] & 0x0FF) << 16) | ((pQ7[1] & 0x0FF) << 8) | (pQ7[0] & 0x0FF);
#endif
return (val);
}
/**
@brief Read 4 Q7 from Q7 pointer and increment pointer afterwards.
@param[in] pQ7 points to input value
@return Q31 value
*/
#define read_q7x4_ia(pQ7) read_q7x4((*(pQ7) += 4) - 4)
/**
@brief Read 4 Q7 from Q7 pointer and decrement pointer afterwards.
@param[in] pQ7 points to input value
@return Q31 value
*/
#define read_q7x4_da(pQ7) read_q7x4((*(pQ7) -= 4) + 4)
/**
@brief Write 4 Q7 to Q7 pointer and increment pointer afterwards.
@param[in] pQ7 points to input value
@param[in] value Q31 value
*/
__STATIC_FORCEINLINE void write_q7x4_ia (
q7_t ** pQ7,
q31_t value)
{
q31_t val = value;
#ifdef __ARM_FEATURE_UNALIGNED
memcpy (*pQ7, &val, 4);
#else
(*pQ7)[0] = (q7_t)(val & 0x0FF);
(*pQ7)[1] = (q7_t)((val >> 8) & 0x0FF);
(*pQ7)[2] = (q7_t)((val >> 16) & 0x0FF);
(*pQ7)[3] = (q7_t)((val >> 24) & 0x0FF);
#endif
*pQ7 += 4;
}
#ifdef __cplusplus
}
#endif
#endif /*ifndef _ARM_MATH_MEMORY_H_ */

645
sdk/CMSIS/DSP/Include/arm_math_types.h Normal file
View file

@ -0,0 +1,645 @@
/******************************************************************************
* @file arm_math_types.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2021 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_MATH_TYPES_H_
#define ARM_MATH_TYPES_H_
#if defined(ARM_DSP_CUSTOM_CONFIG)
#include "arm_dsp_config.h"
#endif
#ifndef ARM_DSP_ATTRIBUTE
#define ARM_DSP_ATTRIBUTE
#endif
#ifndef ARM_DSP_TABLE_ATTRIBUTE
#define ARM_DSP_TABLE_ATTRIBUTE
#endif
#ifdef __cplusplus
extern "C"
{
#endif
/* Compiler specific diagnostic adjustment */
#if defined ( __CC_ARM )
#elif defined ( __ARMCC_VERSION ) && ( __ARMCC_VERSION >= 6010050 )
#elif defined ( __APPLE_CC__ )
#pragma GCC diagnostic ignored "-Wold-style-cast"
#elif defined(__clang__)
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wsign-conversion"
#pragma GCC diagnostic ignored "-Wconversion"
#pragma GCC diagnostic ignored "-Wunused-parameter"
#elif defined ( __GNUC__ )
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wsign-conversion"
#pragma GCC diagnostic ignored "-Wconversion"
#pragma GCC diagnostic ignored "-Wunused-parameter"
// Disable some code having issue with GCC
#define ARM_DSP_BUILT_WITH_GCC
#elif defined ( __ICCARM__ )
#elif defined ( __TI_ARM__ )
#elif defined ( __CSMC__ )
#elif defined ( __TASKING__ )
#elif defined ( _MSC_VER )
#else
#error Unknown compiler
#endif
/* Included for instrinsics definitions */
#if defined (_MSC_VER )
#include <stdint.h>
#define __STATIC_FORCEINLINE static __forceinline
#define __STATIC_INLINE static __inline
#define __ALIGNED(x) __declspec(align(x))
#define __WEAK
#elif defined ( __APPLE_CC__ )
#include <stdint.h>
#define __ALIGNED(x) __attribute__((aligned(x)))
#define __STATIC_FORCEINLINE static inline __attribute__((always_inline))
#define __STATIC_INLINE static inline
#define __WEAK
#elif defined (__GNUC_PYTHON__)
#include <stdint.h>
#define __ALIGNED(x) __attribute__((aligned(x)))
#define __STATIC_FORCEINLINE static inline __attribute__((always_inline))
#define __STATIC_INLINE static inline
#define __WEAK
#else
#include "cmsis_compiler.h"
#endif
#include <string.h>
#include <math.h>
#include <float.h>
#include <limits.h>
/* evaluate ARM DSP feature */
#if (defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1))
#define ARM_MATH_DSP 1
#endif
#if defined(ARM_MATH_NEON)
#if defined(_MSC_VER) && defined(_M_ARM64EC)
#include <arm64_neon.h>
#else
#include <arm_neon.h>
#endif
#if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
#if !defined(ARM_MATH_NEON_FLOAT16)
#define ARM_MATH_NEON_FLOAT16
#endif
#endif
#endif
#if !defined(ARM_MATH_AUTOVECTORIZE)
#if defined(__ARM_FEATURE_MVE)
#if __ARM_FEATURE_MVE
#if !defined(ARM_MATH_MVEI)
#define ARM_MATH_MVEI
#endif
#endif
#if defined(__ARM_FEATURE_MVE) && (__ARM_FEATURE_MVE & 2)
#if !defined(ARM_MATH_MVEF)
#define ARM_MATH_MVEF
#endif
#if !defined(ARM_MATH_MVE_FLOAT16)
#define ARM_MATH_MVE_FLOAT16
#endif
#endif
#endif /* defined (__ARM_FEATURE_MVE) */
#endif /* !defined (ARM_MATH_AUTOVECTORIZE) */
#if defined (ARM_MATH_HELIUM)
#if !defined(ARM_MATH_MVEF)
#define ARM_MATH_MVEF
#endif
#if !defined(ARM_MATH_MVEI)
#define ARM_MATH_MVEI
#endif
#if !defined(ARM_MATH_MVE_FLOAT16)
#define ARM_MATH_MVE_FLOAT16
#endif
#endif
#if defined ( __CC_ARM )
/* Enter low optimization region - place directly above function definition */
#if defined( __ARM_ARCH_7EM__ )
#define LOW_OPTIMIZATION_ENTER \
_Pragma ("push") \
_Pragma ("O1")
#else
#define LOW_OPTIMIZATION_ENTER
#endif
/* Exit low optimization region - place directly after end of function definition */
#if defined ( __ARM_ARCH_7EM__ )
#define LOW_OPTIMIZATION_EXIT \
_Pragma ("pop")
#else
#define LOW_OPTIMIZATION_EXIT
#endif
/* Enter low optimization region - place directly above function definition */
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
/* Exit low optimization region - place directly after end of function definition */
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#elif defined (__ARMCC_VERSION ) && ( __ARMCC_VERSION >= 6010050 )
#define LOW_OPTIMIZATION_ENTER
#define LOW_OPTIMIZATION_EXIT
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#elif defined ( __APPLE_CC__ )
#define LOW_OPTIMIZATION_ENTER
#define LOW_OPTIMIZATION_EXIT
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#elif defined ( __GNUC__ )
#define LOW_OPTIMIZATION_ENTER \
__attribute__(( optimize("-O1") ))
#define LOW_OPTIMIZATION_EXIT
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#elif defined ( __ICCARM__ )
/* Enter low optimization region - place directly above function definition */
#if defined ( __ARM_ARCH_7EM__ )
#define LOW_OPTIMIZATION_ENTER \
_Pragma ("optimize=low")
#else
#define LOW_OPTIMIZATION_ENTER
#endif
/* Exit low optimization region - place directly after end of function definition */
#define LOW_OPTIMIZATION_EXIT
/* Enter low optimization region - place directly above function definition */
#if defined ( __ARM_ARCH_7EM__ )
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER \
_Pragma ("optimize=low")
#else
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
#endif
/* Exit low optimization region - place directly after end of function definition */
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#elif defined ( __TI_ARM__ )
#define LOW_OPTIMIZATION_ENTER
#define LOW_OPTIMIZATION_EXIT
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#elif defined ( __CSMC__ )
#define LOW_OPTIMIZATION_ENTER
#define LOW_OPTIMIZATION_EXIT
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#elif defined ( __TASKING__ )
#define LOW_OPTIMIZATION_ENTER
#define LOW_OPTIMIZATION_EXIT
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#elif defined ( _MSC_VER ) || defined(__GNUC_PYTHON__)
#define LOW_OPTIMIZATION_ENTER
#define LOW_OPTIMIZATION_EXIT
#define IAR_ONLY_LOW_OPTIMIZATION_ENTER
#define IAR_ONLY_LOW_OPTIMIZATION_EXIT
#endif
/* Compiler specific diagnostic adjustment */
#if defined ( __CC_ARM )
#elif defined ( __ARMCC_VERSION ) && ( __ARMCC_VERSION >= 6010050 )
#elif defined ( __APPLE_CC__ )
#elif defined ( __GNUC__ )
#pragma GCC diagnostic pop
#elif defined ( __ICCARM__ )
#elif defined ( __TI_ARM__ )
#elif defined ( __CSMC__ )
#elif defined ( __TASKING__ )
#elif defined ( _MSC_VER )
#else
#error Unknown compiler
#endif
#ifdef __cplusplus
}
#endif
#if defined(__ARM_FEATURE_MVE) && __ARM_FEATURE_MVE
#include <arm_mve.h>
#endif
#if defined(ARM_DSP_CONFIG_TABLES)
#error("-DARM_DSP_CONFIG_TABLES no more supported. Use the new initialization functions to let the linker optimize the code size.")
#endif
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup genericTypes Generic Types
* @{
*/
/**
* @brief 8-bit fractional data type in 1.7 format.
*/
typedef int8_t q7_t;
/**
* @brief 16-bit fractional data type in 1.15 format.
*/
typedef int16_t q15_t;
/**
* @brief 32-bit fractional data type in 1.31 format.
*/
typedef int32_t q31_t;
/**
* @brief 64-bit fractional data type in 1.63 format.
*/
typedef int64_t q63_t;
/**
* @brief 32-bit floating-point type definition.
*/
#if !defined(__ICCARM__) || !(__ARM_FEATURE_MVE & 2)
typedef float float32_t;
#endif
/**
* @brief 64-bit floating-point type definition.
*/
typedef double float64_t;
/**
* @brief vector types
*/
#if defined(ARM_MATH_NEON) || (defined (ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE))
/**
* @brief 64-bit fractional 128-bit vector data type in 1.63 format
*/
typedef int64x2_t q63x2_t;
/**
* @brief 32-bit fractional 128-bit vector data type in 1.31 format.
*/
typedef int32x4_t q31x4_t;
/**
* @brief 16-bit fractional 128-bit vector data type with 16-bit alignment in 1.15 format.
*/
typedef __ALIGNED(2) int16x8_t q15x8_t;
/**
* @brief 8-bit fractional 128-bit vector data type with 8-bit alignment in 1.7 format.
*/
typedef __ALIGNED(1) int8x16_t q7x16_t;
/**
* @brief 32-bit fractional 128-bit vector pair data type in 1.31 format.
*/
typedef int32x4x2_t q31x4x2_t;
/**
* @brief 32-bit fractional 128-bit vector quadruplet data type in 1.31 format.
*/
typedef int32x4x4_t q31x4x4_t;
/**
* @brief 16-bit fractional 128-bit vector pair data type in 1.15 format.
*/
typedef int16x8x2_t q15x8x2_t;
/**
* @brief 16-bit fractional 128-bit vector quadruplet data type in 1.15 format.
*/
typedef int16x8x4_t q15x8x4_t;
/**
* @brief 8-bit fractional 128-bit vector pair data type in 1.7 format.
*/
typedef int8x16x2_t q7x16x2_t;
/**
* @brief 8-bit fractional 128-bit vector quadruplet data type in 1.7 format.
*/
typedef int8x16x4_t q7x16x4_t;
/**
* @brief 32-bit fractional data type in 9.23 format.
*/
typedef int32_t q23_t;
/**
* @brief 32-bit fractional 128-bit vector data type in 9.23 format.
*/
typedef int32x4_t q23x4_t;
/**
* @brief 64-bit status 128-bit vector data type.
*/
typedef int64x2_t status64x2_t;
/**
* @brief 32-bit status 128-bit vector data type.
*/
typedef int32x4_t status32x4_t;
/**
* @brief 16-bit status 128-bit vector data type.
*/
typedef int16x8_t status16x8_t;
/**
* @brief 8-bit status 128-bit vector data type.
*/
typedef int8x16_t status8x16_t;
#endif
#if defined(ARM_MATH_NEON) || (defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)) /* floating point vector*/
/**
* @brief 32-bit floating-point 128-bit vector type
*/
typedef float32x4_t f32x4_t;
/**
* @brief 32-bit floating-point 128-bit vector pair data type
*/
typedef float32x4x2_t f32x4x2_t;
/**
* @brief 32-bit floating-point 128-bit vector quadruplet data type
*/
typedef float32x4x4_t f32x4x4_t;
/**
* @brief 32-bit ubiquitous 128-bit vector data type
*/
typedef union _any32x4_t
{
float32x4_t f;
int32x4_t i;
} any32x4_t;
#endif
#if defined(ARM_MATH_NEON)
/**
* @brief 32-bit fractional 64-bit vector data type in 1.31 format.
*/
typedef int32x2_t q31x2_t;
/**
* @brief 16-bit fractional 64-bit vector data type in 1.15 format.
*/
typedef __ALIGNED(2) int16x4_t q15x4_t;
/**
* @brief 8-bit fractional 64-bit vector data type in 1.7 format.
*/
typedef __ALIGNED(1) int8x8_t q7x8_t;
/**
* @brief 32-bit float 64-bit vector data type.
*/
typedef float32x2_t f32x2_t;
/**
* @brief 32-bit floating-point 128-bit vector triplet data type
*/
typedef float32x4x3_t f32x4x3_t;
/**
* @brief 32-bit fractional 128-bit vector triplet data type in 1.31 format
*/
typedef int32x4x3_t q31x4x3_t;
/**
* @brief 16-bit fractional 128-bit vector triplet data type in 1.15 format
*/
typedef int16x8x3_t q15x8x3_t;
/**
* @brief 8-bit fractional 128-bit vector triplet data type in 1.7 format
*/
typedef int8x16x3_t q7x16x3_t;
/**
* @brief 32-bit floating-point 64-bit vector pair data type
*/
typedef float32x2x2_t f32x2x2_t;
/**
* @brief 32-bit floating-point 64-bit vector triplet data type
*/
typedef float32x2x3_t f32x2x3_t;
/**
* @brief 32-bit floating-point 64-bit vector quadruplet data type
*/
typedef float32x2x4_t f32x2x4_t;
/**
* @brief 32-bit fractional 64-bit vector pair data type in 1.31 format
*/
typedef int32x2x2_t q31x2x2_t;
/**
* @brief 32-bit fractional 64-bit vector triplet data type in 1.31 format
*/
typedef int32x2x3_t q31x2x3_t;
/**
* @brief 32-bit fractional 64-bit vector quadruplet data type in 1.31 format
*/
typedef int32x4x3_t q31x2x4_t;
/**
* @brief 16-bit fractional 64-bit vector pair data type in 1.15 format
*/
typedef int16x4x2_t q15x4x2_t;
/**
* @brief 16-bit fractional 64-bit vector triplet data type in 1.15 format
*/
typedef int16x4x2_t q15x4x3_t;
/**
* @brief 16-bit fractional 64-bit vector quadruplet data type in 1.15 format
*/
typedef int16x4x3_t q15x4x4_t;
/**
* @brief 8-bit fractional 64-bit vector pair data type in 1.7 format
*/
typedef int8x8x2_t q7x8x2_t;
/**
* @brief 8-bit fractional 64-bit vector triplet data type in 1.7 format
*/
typedef int8x8x3_t q7x8x3_t;
/**
* @brief 8-bit fractional 64-bit vector quadruplet data type in 1.7 format
*/
typedef int8x8x4_t q7x8x4_t;
/**
* @brief 32-bit ubiquitous 64-bit vector data type
*/
typedef union _any32x2_t
{
float32x2_t f;
int32x2_t i;
} any32x2_t;
/**
* @brief 32-bit status 64-bit vector data type.
*/
typedef int32x4_t status32x2_t;
/**
* @brief 16-bit status 64-bit vector data type.
*/
typedef int16x8_t status16x4_t;
/**
* @brief 8-bit status 64-bit vector data type.
*/
typedef int8x16_t status8x8_t;
#endif
/**
* @brief Error status returned by some functions in the library.
*/
typedef enum
{
ARM_MATH_SUCCESS = 0, /**< No error */
ARM_MATH_ARGUMENT_ERROR = -1, /**< One or more arguments are incorrect */
ARM_MATH_LENGTH_ERROR = -2, /**< Length of data buffer is incorrect */
ARM_MATH_SIZE_MISMATCH = -3, /**< Size of matrices is not compatible with the operation */
ARM_MATH_NANINF = -4, /**< Not-a-number (NaN) or infinity is generated */
ARM_MATH_SINGULAR = -5, /**< Input matrix is singular and cannot be inverted */
ARM_MATH_TEST_FAILURE = -6, /**< Test Failed */
ARM_MATH_DECOMPOSITION_FAILURE = -7 /**< Decomposition Failed */
} arm_status;
/**
* @} // endgroup generic
*/
#define F64_MAX ((float64_t)DBL_MAX)
#define F32_MAX ((float32_t)FLT_MAX)
#define F64_MIN (-DBL_MAX)
#define F32_MIN (-FLT_MAX)
#define F64_ABSMAX ((float64_t)DBL_MAX)
#define F32_ABSMAX ((float32_t)FLT_MAX)
#define F64_ABSMIN ((float64_t)0.0)
#define F32_ABSMIN ((float32_t)0.0)
#define Q31_MAX ((q31_t)(0x7FFFFFFFL))
#define Q15_MAX ((q15_t)(0x7FFF))
#define Q7_MAX ((q7_t)(0x7F))
#define Q31_MIN ((q31_t)(0x80000000L))
#define Q15_MIN ((q15_t)(0x8000))
#define Q7_MIN ((q7_t)(0x80))
#define Q31_ABSMAX ((q31_t)(0x7FFFFFFFL))
#define Q15_ABSMAX ((q15_t)(0x7FFF))
#define Q7_ABSMAX ((q7_t)(0x7F))
#define Q31_ABSMIN ((q31_t)0)
#define Q15_ABSMIN ((q15_t)0)
#define Q7_ABSMIN ((q7_t)0)
/* Dimension C vector space */
#define CMPLX_DIM 2
#ifdef __cplusplus
}
#endif
#endif /*ifndef _ARM_MATH_TYPES_H_ */

163
sdk/CMSIS/DSP/Include/arm_math_types_f16.h Normal file
View file

@ -0,0 +1,163 @@
/******************************************************************************
* @file arm_math_types_f16.h
* @brief Public header file for f16 function of the CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2021 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_MATH_TYPES_F16_H
#define ARM_MATH_TYPES_F16_H
#include "arm_math_types.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if !defined( __CC_ARM )
/**
* @brief 16-bit floating-point type definition.
* This is already defined in arm_mve.h
*
* This is not fully supported on ARM AC5.
*/
/*
Check if the type __fp16 is available.
If it is not available, f16 version of the kernels
won't be built.
*/
#if defined(__ARM_FEATURE_MVE) && (__ARM_FEATURE_MVE & 2)
/* When Vector float16, this flag is always defined and can't be disabled */
#define ARM_FLOAT16_SUPPORTED
#else
#if !defined(DISABLEFLOAT16)
#if defined(__ARM_FP16_FORMAT_IEEE) || defined(__ARM_FP16_FORMAT_ALTERNATIVE)
typedef __fp16 float16_t;
#define ARM_FLOAT16_SUPPORTED
#endif
#endif
#endif
#if defined(ARM_MATH_NEON) || (defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)) /* floating point vector*/
#if defined(ARM_MATH_MVE_FLOAT16) || defined(ARM_MATH_NEON_FLOAT16)
/**
* @brief 16-bit floating-point 128-bit vector data type
*/
typedef __ALIGNED(2) float16x8_t f16x8_t;
/**
* @brief 16-bit floating-point 128-bit vector pair data type
*/
typedef float16x8x2_t f16x8x2_t;
/**
* @brief 16-bit floating-point 128-bit vector quadruplet data type
*/
typedef float16x8x4_t f16x8x4_t;
/**
* @brief 16-bit ubiquitous 128-bit vector data type
*/
typedef union _any16x8_t
{
float16x8_t f;
int16x8_t i;
} any16x8_t;
#endif
#endif
#if defined(ARM_MATH_NEON)
#if defined(ARM_MATH_NEON_FLOAT16)
/**
* @brief 16-bit float 64-bit vector data type.
*/
typedef __ALIGNED(2) float16x4_t f16x4_t;
/**
* @brief 16-bit floating-point 128-bit vector triplet data type
*/
typedef float16x8x3_t f16x8x3_t;
/**
* @brief 16-bit floating-point 64-bit vector pair data type
*/
typedef float16x4x2_t f16x4x2_t;
/**
* @brief 16-bit floating-point 64-bit vector triplet data type
*/
typedef float16x4x3_t f16x4x3_t;
/**
* @brief 16-bit floating-point 64-bit vector quadruplet data type
*/
typedef float16x4x4_t f16x4x4_t;
/**
* @brief 16-bit ubiquitous 64-bit vector data type
*/
typedef union _any16x4_t
{
float16x4_t f;
int16x4_t i;
} any16x4_t;
#endif
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
#if defined(__ICCARM__)
#define F16INFINITY ((float16_t) INFINITY)
#else
#define F16INFINITY ((float16_t)__builtin_inf())
#endif
#define F16_MAX ((float16_t)__FLT16_MAX__)
#define F16_MIN (-(_Float16)__FLT16_MAX__)
#define F16_ABSMAX ((float16_t)__FLT16_MAX__)
#define F16_ABSMIN ((float16_t)0.0f16)
#endif /* ARM_FLOAT16_SUPPORTED*/
#endif /* !defined( __CC_ARM ) */
#ifdef __cplusplus
}
#endif
#endif /* _ARM_MATH_F16_H */

193
sdk/CMSIS/DSP/Include/arm_mve_tables.h Normal file
View file

@ -0,0 +1,193 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_mve_tables.h
* Description: common tables like fft twiddle factors, Bitreverse, reciprocal etc
* used for MVE implementation only
*
* @version V1.10.0
* @date 04 October 2021
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_MVE_TABLES_H
#define ARM_MVE_TABLES_H
#include "arm_math_types.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
extern const uint32_t rearranged_twiddle_tab_stride1_arr_16_f32[2];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_16_f32[2];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_16_f32[2];
extern const float32_t rearranged_twiddle_stride1_16_f32[8];
extern const float32_t rearranged_twiddle_stride2_16_f32[8];
extern const float32_t rearranged_twiddle_stride3_16_f32[8];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_64_f32[3];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_64_f32[3];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_64_f32[3];
extern const float32_t rearranged_twiddle_stride1_64_f32[40];
extern const float32_t rearranged_twiddle_stride2_64_f32[40];
extern const float32_t rearranged_twiddle_stride3_64_f32[40];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_256_f32[4];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_256_f32[4];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_256_f32[4];
extern const float32_t rearranged_twiddle_stride1_256_f32[168];
extern const float32_t rearranged_twiddle_stride2_256_f32[168];
extern const float32_t rearranged_twiddle_stride3_256_f32[168];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_1024_f32[5];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_1024_f32[5];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_1024_f32[5];
extern const float32_t rearranged_twiddle_stride1_1024_f32[680];
extern const float32_t rearranged_twiddle_stride2_1024_f32[680];
extern const float32_t rearranged_twiddle_stride3_1024_f32[680];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_4096_f32[6];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_4096_f32[6];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_4096_f32[6];
extern const float32_t rearranged_twiddle_stride1_4096_f32[2728];
extern const float32_t rearranged_twiddle_stride2_4096_f32[2728];
extern const float32_t rearranged_twiddle_stride3_4096_f32[2728];
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
extern const uint32_t rearranged_twiddle_tab_stride1_arr_16_q31[2];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_16_q31[2];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_16_q31[2];
extern const q31_t rearranged_twiddle_stride1_16_q31[8];
extern const q31_t rearranged_twiddle_stride2_16_q31[8];
extern const q31_t rearranged_twiddle_stride3_16_q31[8];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_64_q31[3];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_64_q31[3];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_64_q31[3];
extern const q31_t rearranged_twiddle_stride1_64_q31[40];
extern const q31_t rearranged_twiddle_stride2_64_q31[40];
extern const q31_t rearranged_twiddle_stride3_64_q31[40];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_256_q31[4];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_256_q31[4];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_256_q31[4];
extern const q31_t rearranged_twiddle_stride1_256_q31[168];
extern const q31_t rearranged_twiddle_stride2_256_q31[168];
extern const q31_t rearranged_twiddle_stride3_256_q31[168];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_1024_q31[5];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_1024_q31[5];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_1024_q31[5];
extern const q31_t rearranged_twiddle_stride1_1024_q31[680];
extern const q31_t rearranged_twiddle_stride2_1024_q31[680];
extern const q31_t rearranged_twiddle_stride3_1024_q31[680];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_4096_q31[6];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_4096_q31[6];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_4096_q31[6];
extern const q31_t rearranged_twiddle_stride1_4096_q31[2728];
extern const q31_t rearranged_twiddle_stride2_4096_q31[2728];
extern const q31_t rearranged_twiddle_stride3_4096_q31[2728];
#endif /* defined(ARM_MATH_MVEI) */
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
extern const uint32_t rearranged_twiddle_tab_stride1_arr_16_q15[2];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_16_q15[2];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_16_q15[2];
extern const q15_t rearranged_twiddle_stride1_16_q15[8];
extern const q15_t rearranged_twiddle_stride2_16_q15[8];
extern const q15_t rearranged_twiddle_stride3_16_q15[8];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_64_q15[3];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_64_q15[3];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_64_q15[3];
extern const q15_t rearranged_twiddle_stride1_64_q15[40];
extern const q15_t rearranged_twiddle_stride2_64_q15[40];
extern const q15_t rearranged_twiddle_stride3_64_q15[40];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_256_q15[4];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_256_q15[4];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_256_q15[4];
extern const q15_t rearranged_twiddle_stride1_256_q15[168];
extern const q15_t rearranged_twiddle_stride2_256_q15[168];
extern const q15_t rearranged_twiddle_stride3_256_q15[168];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_1024_q15[5];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_1024_q15[5];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_1024_q15[5];
extern const q15_t rearranged_twiddle_stride1_1024_q15[680];
extern const q15_t rearranged_twiddle_stride2_1024_q15[680];
extern const q15_t rearranged_twiddle_stride3_1024_q15[680];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_4096_q15[6];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_4096_q15[6];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_4096_q15[6];
extern const q15_t rearranged_twiddle_stride1_4096_q15[2728];
extern const q15_t rearranged_twiddle_stride2_4096_q15[2728];
extern const q15_t rearranged_twiddle_stride3_4096_q15[2728];
#endif /* defined(ARM_MATH_MVEI) */
#ifdef __cplusplus
}
#endif
#endif /*_ARM_MVE_TABLES_H*/

97
sdk/CMSIS/DSP/Include/arm_mve_tables_f16.h Normal file
View file

@ -0,0 +1,97 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_mve_tables_f16.h
* Description: common tables like fft twiddle factors, Bitreverse, reciprocal etc
* used for MVE implementation only
*
* @version V1.10.0
* @date 04 October 2021
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_MVE_TABLES_F16_H
#define ARM_MVE_TABLES_F16_H
#include "arm_math_types_f16.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
extern const uint32_t rearranged_twiddle_tab_stride1_arr_16_f16[2];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_16_f16[2];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_16_f16[2];
extern const float16_t rearranged_twiddle_stride1_16_f16[8];
extern const float16_t rearranged_twiddle_stride2_16_f16[8];
extern const float16_t rearranged_twiddle_stride3_16_f16[8];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_64_f16[3];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_64_f16[3];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_64_f16[3];
extern const float16_t rearranged_twiddle_stride1_64_f16[40];
extern const float16_t rearranged_twiddle_stride2_64_f16[40];
extern const float16_t rearranged_twiddle_stride3_64_f16[40];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_256_f16[4];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_256_f16[4];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_256_f16[4];
extern const float16_t rearranged_twiddle_stride1_256_f16[168];
extern const float16_t rearranged_twiddle_stride2_256_f16[168];
extern const float16_t rearranged_twiddle_stride3_256_f16[168];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_1024_f16[5];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_1024_f16[5];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_1024_f16[5];
extern const float16_t rearranged_twiddle_stride1_1024_f16[680];
extern const float16_t rearranged_twiddle_stride2_1024_f16[680];
extern const float16_t rearranged_twiddle_stride3_1024_f16[680];
extern const uint32_t rearranged_twiddle_tab_stride1_arr_4096_f16[6];
extern const uint32_t rearranged_twiddle_tab_stride2_arr_4096_f16[6];
extern const uint32_t rearranged_twiddle_tab_stride3_arr_4096_f16[6];
extern const float16_t rearranged_twiddle_stride1_4096_f16[2728];
extern const float16_t rearranged_twiddle_stride2_4096_f16[2728];
extern const float16_t rearranged_twiddle_stride3_4096_f16[2728];
#endif /* defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE) */
#ifdef __cplusplus
}
#endif
#endif /*_ARM_MVE_TABLES_F16_H*/

373
sdk/CMSIS/DSP/Include/arm_vec_math.h Normal file
View file

@ -0,0 +1,373 @@
/******************************************************************************
* @file arm_vec_math.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2021 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_VEC_MATH_H
#define ARM_VEC_MATH_H
#include "arm_math_types.h"
#include "arm_common_tables.h"
#include "arm_helium_utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if (defined(ARM_MATH_MVEF) || defined(ARM_MATH_HELIUM)) && !defined(ARM_MATH_AUTOVECTORIZE)
#define INV_NEWTON_INIT_F32 0x7EF127EA
static const float32_t __logf_rng_f32=0.693147180f;
/* fast inverse approximation (3x newton) */
__STATIC_INLINE f32x4_t vrecip_medprec_f32(
f32x4_t x)
{
q31x4_t m;
f32x4_t b;
any32x4_t xinv;
f32x4_t ax = vabsq(x);
xinv.f = ax;
m = 0x3F800000 - (xinv.i & 0x7F800000);
xinv.i = xinv.i + m;
xinv.f = 1.41176471f - 0.47058824f * xinv.f;
xinv.i = xinv.i + m;
b = 2.0f - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f - xinv.f * ax;
xinv.f = xinv.f * b;
xinv.f = vdupq_m(xinv.f, F32_MAX, vcmpeqq(x, 0.0f));
/*
* restore sign
*/
xinv.f = vnegq_m(xinv.f, xinv.f, vcmpltq(x, 0.0f));
return xinv.f;
}
/* fast inverse approximation (4x newton) */
__STATIC_INLINE f32x4_t vrecip_hiprec_f32(
f32x4_t x)
{
q31x4_t m;
f32x4_t b;
any32x4_t xinv;
f32x4_t ax = vabsq(x);
xinv.f = ax;
m = 0x3F800000 - (xinv.i & 0x7F800000);
xinv.i = xinv.i + m;
xinv.f = 1.41176471f - 0.47058824f * xinv.f;
xinv.i = xinv.i + m;
b = 2.0f - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f - xinv.f * ax;
xinv.f = xinv.f * b;
xinv.f = vdupq_m(xinv.f, F32_MAX, vcmpeqq(x, 0.0f));
/*
* restore sign
*/
xinv.f = vnegq_m(xinv.f, xinv.f, vcmpltq(x, 0.0f));
return xinv.f;
}
__STATIC_INLINE f32x4_t vdiv_f32(
f32x4_t num, f32x4_t den)
{
return vmulq(num, vrecip_hiprec_f32(den));
}
/**
@brief Single-precision taylor dev.
@param[in] x f32 quad vector input
@param[in] coeffs f32 quad vector coeffs
@return destination f32 quad vector
*/
__STATIC_INLINE f32x4_t vtaylor_polyq_f32(
f32x4_t x,
const float32_t * coeffs)
{
f32x4_t A = vfmasq(vdupq_n_f32(coeffs[4]), x, coeffs[0]);
f32x4_t B = vfmasq(vdupq_n_f32(coeffs[6]), x, coeffs[2]);
f32x4_t C = vfmasq(vdupq_n_f32(coeffs[5]), x, coeffs[1]);
f32x4_t D = vfmasq(vdupq_n_f32(coeffs[7]), x, coeffs[3]);
f32x4_t x2 = vmulq(x, x);
f32x4_t x4 = vmulq(x2, x2);
f32x4_t res = vfmaq(vfmaq_f32(A, B, x2), vfmaq_f32(C, D, x2), x4);
return res;
}
__STATIC_INLINE f32x4_t vmant_exp_f32(
f32x4_t x,
int32x4_t * e)
{
any32x4_t r;
int32x4_t n;
r.f = x;
n = r.i >> 23;
n = n - 127;
r.i = r.i - (n << 23);
*e = n;
return r.f;
}
__STATIC_INLINE f32x4_t vlogq_f32(f32x4_t vecIn)
{
q31x4_t vecExpUnBiased;
f32x4_t vecTmpFlt0, vecTmpFlt1;
f32x4_t vecAcc0, vecAcc1, vecAcc2, vecAcc3;
f32x4_t vecExpUnBiasedFlt;
/*
* extract exponent
*/
vecTmpFlt1 = vmant_exp_f32(vecIn, &vecExpUnBiased);
vecTmpFlt0 = vecTmpFlt1 * vecTmpFlt1;
/*
* a = (__logf_lut_f32[4] * r.f) + (__logf_lut_f32[0]);
*/
vecAcc0 = vdupq_n_f32(__logf_lut_f32[0]);
vecAcc0 = vfmaq(vecAcc0, vecTmpFlt1, __logf_lut_f32[4]);
/*
* b = (__logf_lut_f32[6] * r.f) + (__logf_lut_f32[2]);
*/
vecAcc1 = vdupq_n_f32(__logf_lut_f32[2]);
vecAcc1 = vfmaq(vecAcc1, vecTmpFlt1, __logf_lut_f32[6]);
/*
* c = (__logf_lut_f32[5] * r.f) + (__logf_lut_f32[1]);
*/
vecAcc2 = vdupq_n_f32(__logf_lut_f32[1]);
vecAcc2 = vfmaq(vecAcc2, vecTmpFlt1, __logf_lut_f32[5]);
/*
* d = (__logf_lut_f32[7] * r.f) + (__logf_lut_f32[3]);
*/
vecAcc3 = vdupq_n_f32(__logf_lut_f32[3]);
vecAcc3 = vfmaq(vecAcc3, vecTmpFlt1, __logf_lut_f32[7]);
/*
* a = a + b * xx;
*/
vecAcc0 = vfmaq(vecAcc0, vecAcc1, vecTmpFlt0);
/*
* c = c + d * xx;
*/
vecAcc2 = vfmaq(vecAcc2, vecAcc3, vecTmpFlt0);
/*
* xx = xx * xx;
*/
vecTmpFlt0 = vecTmpFlt0 * vecTmpFlt0;
vecExpUnBiasedFlt = vcvtq_f32_s32(vecExpUnBiased);
/*
* r.f = a + c * xx;
*/
vecAcc0 = vfmaq(vecAcc0, vecAcc2, vecTmpFlt0);
/*
* add exponent
* r.f = r.f + ((float32_t) m) * __logf_rng_f32;
*/
vecAcc0 = vfmaq(vecAcc0, vecExpUnBiasedFlt, __logf_rng_f32);
// set log0 down to -inf
vecAcc0 = vdupq_m(vecAcc0, -F32_MAX, vcmpeqq(vecIn, 0.0f));
return vecAcc0;
}
__STATIC_INLINE f32x4_t vexpq_f32(
f32x4_t x)
{
// Perform range reduction [-log(2),log(2)]
int32x4_t m = vcvtq_s32_f32(vmulq_n_f32(x, 1.4426950408f));
f32x4_t val = vfmsq_f32(x, vcvtq_f32_s32(m), vdupq_n_f32(0.6931471805f));
// Polynomial Approximation
f32x4_t poly = vtaylor_polyq_f32(val, exp_tab);
// Reconstruct
poly = (f32x4_t) (vqaddq_s32((q31x4_t) (poly), vqshlq_n_s32(m, 23)));
poly = vdupq_m(poly, 0.0f, vcmpltq_n_s32(m, -126));
return poly;
}
__STATIC_INLINE f32x4_t arm_vec_exponent_f32(f32x4_t x, int32_t nb)
{
f32x4_t r = x;
nb--;
while (nb > 0) {
r = vmulq(r, x);
nb--;
}
return (r);
}
__STATIC_INLINE f32x4_t vrecip_f32(f32x4_t vecIn)
{
f32x4_t vecSx, vecW, vecTmp;
any32x4_t v;
vecSx = vabsq(vecIn);
v.f = vecIn;
v.i = vsubq(vdupq_n_s32(INV_NEWTON_INIT_F32), v.i);
vecW = vmulq(vecSx, v.f);
// v.f = v.f * (8 + w * (-28 + w * (56 + w * (-70 + w *(56 + w * (-28 + w * (8 - w)))))));
vecTmp = vsubq(vdupq_n_f32(8.0f), vecW);
vecTmp = vfmasq(vecW, vecTmp, -28.0f);
vecTmp = vfmasq(vecW, vecTmp, 56.0f);
vecTmp = vfmasq(vecW, vecTmp, -70.0f);
vecTmp = vfmasq(vecW, vecTmp, 56.0f);
vecTmp = vfmasq(vecW, vecTmp, -28.0f);
vecTmp = vfmasq(vecW, vecTmp, 8.0f);
v.f = vmulq(v.f, vecTmp);
v.f = vdupq_m(v.f, F32_MAX, vcmpeqq(vecIn, 0.0f));
/*
* restore sign
*/
v.f = vnegq_m(v.f, v.f, vcmpltq(vecIn, 0.0f));
return v.f;
}
__STATIC_INLINE f32x4_t vtanhq_f32(
f32x4_t val)
{
f32x4_t x =
vminnmq_f32(vmaxnmq_f32(val, vdupq_n_f32(-10.f)), vdupq_n_f32(10.0f));
f32x4_t exp2x = vexpq_f32(vmulq_n_f32(x, 2.f));
f32x4_t num = vsubq_n_f32(exp2x, 1.f);
f32x4_t den = vaddq_n_f32(exp2x, 1.f);
f32x4_t tanh = vmulq_f32(num, vrecip_f32(den));
return tanh;
}
__STATIC_INLINE f32x4_t vpowq_f32(
f32x4_t val,
f32x4_t n)
{
return vexpq_f32(vmulq_f32(n, vlogq_f32(val)));
}
#endif /* (defined(ARM_MATH_MVEF) || defined(ARM_MATH_HELIUM)) && !defined(ARM_MATH_AUTOVECTORIZE)*/
#if (defined(ARM_MATH_MVEI) || defined(ARM_MATH_HELIUM)) && !defined(ARM_MATH_AUTOVECTORIZE)
#endif /* (defined(ARM_MATH_MVEI) || defined(ARM_MATH_HELIUM)) */
#if (defined(ARM_MATH_NEON) || defined(ARM_MATH_NEON_EXPERIMENTAL)) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "NEMath.h"
/**
* @brief Vectorized integer exponentiation
* @param[in] x value
* @param[in] nb integer exponent >= 1
* @return x^nb
*
*/
__STATIC_INLINE float32x4_t arm_vec_exponent_f32(float32x4_t x, int32_t nb)
{
float32x4_t r = x;
nb --;
while(nb > 0)
{
r = vmulq_f32(r , x);
nb--;
}
return(r);
}
__STATIC_INLINE float32x4_t __arm_vec_sqrt_f32_neon(float32x4_t x)
{
float32x4_t x1 = vmaxq_f32(x, vdupq_n_f32(FLT_MIN));
float32x4_t e = vrsqrteq_f32(x1);
e = vmulq_f32(vrsqrtsq_f32(vmulq_f32(x1, e), e), e);
e = vmulq_f32(vrsqrtsq_f32(vmulq_f32(x1, e), e), e);
return vmulq_f32(x, e);
}
__STATIC_INLINE int16x8_t __arm_vec_sqrt_q15_neon(int16x8_t vec)
{
float32x4_t tempF;
int32x4_t tempHI,tempLO;
tempLO = vmovl_s16(vget_low_s16(vec));
tempF = vcvtq_n_f32_s32(tempLO,15);
tempF = __arm_vec_sqrt_f32_neon(tempF);
tempLO = vcvtq_n_s32_f32(tempF,15);
tempHI = vmovl_s16(vget_high_s16(vec));
tempF = vcvtq_n_f32_s32(tempHI,15);
tempF = __arm_vec_sqrt_f32_neon(tempF);
tempHI = vcvtq_n_s32_f32(tempF,15);
return(vcombine_s16(vqmovn_s32(tempLO),vqmovn_s32(tempHI)));
}
__STATIC_INLINE int32x4_t __arm_vec_sqrt_q31_neon(int32x4_t vec)
{
float32x4_t temp;
temp = vcvtq_n_f32_s32(vec,31);
temp = __arm_vec_sqrt_f32_neon(temp);
return(vcvtq_n_s32_f32(temp,31));
}
#endif /* (defined(ARM_MATH_NEON) || defined(ARM_MATH_NEON_EXPERIMENTAL)) && !defined(ARM_MATH_AUTOVECTORIZE) */
#ifdef __cplusplus
}
#endif
#endif /* _ARM_VEC_MATH_H */
/**
*
* End of file.
*/

312
sdk/CMSIS/DSP/Include/arm_vec_math_f16.h Normal file
View file

@ -0,0 +1,312 @@
/******************************************************************************
* @file arm_vec_math_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2021 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_VEC_MATH_F16_H
#define ARM_VEC_MATH_F16_H
#include "arm_math_types_f16.h"
#include "arm_common_tables_f16.h"
#include "arm_helium_utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
static const float16_t __logf_rng_f16=0.693147180f16;
/* fast inverse approximation (3x newton) */
__STATIC_INLINE f16x8_t vrecip_medprec_f16(
f16x8_t x)
{
q15x8_t m;
f16x8_t b;
any16x8_t xinv;
f16x8_t ax = vabsq(x);
xinv.f = ax;
m = 0x03c00 - (xinv.i & 0x07c00);
xinv.i = xinv.i + m;
xinv.f = 1.41176471f16 - 0.47058824f16 * xinv.f;
xinv.i = xinv.i + m;
b = 2.0f16 - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f16 - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f16 - xinv.f * ax;
xinv.f = xinv.f * b;
xinv.f = vdupq_m_n_f16(xinv.f, F16INFINITY, vcmpeqq_n_f16(x, 0.0f));
/*
* restore sign
*/
xinv.f = vnegq_m(xinv.f, xinv.f, vcmpltq_n_f16(x, 0.0f));
return xinv.f;
}
/* fast inverse approximation (4x newton) */
__STATIC_INLINE f16x8_t vrecip_hiprec_f16(
f16x8_t x)
{
q15x8_t m;
f16x8_t b;
any16x8_t xinv;
f16x8_t ax = vabsq(x);
xinv.f = ax;
m = 0x03c00 - (xinv.i & 0x07c00);
xinv.i = xinv.i + m;
xinv.f = 1.41176471f16 - 0.47058824f16 * xinv.f;
xinv.i = xinv.i + m;
b = 2.0f16 - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f16 - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f16 - xinv.f * ax;
xinv.f = xinv.f * b;
b = 2.0f16 - xinv.f * ax;
xinv.f = xinv.f * b;
xinv.f = vdupq_m_n_f16(xinv.f, F16INFINITY, vcmpeqq_n_f16(x, 0.0f));
/*
* restore sign
*/
xinv.f = vnegq_m(xinv.f, xinv.f, vcmpltq_n_f16(x, 0.0f));
return xinv.f;
}
__STATIC_INLINE f16x8_t vdiv_f16(
f16x8_t num, f16x8_t den)
{
return vmulq(num, vrecip_hiprec_f16(den));
}
/**
@brief Single-precision taylor dev.
@param[in] x f16 vector input
@param[in] coeffs f16 vector coeffs
@return destination f16 vector
*/
__STATIC_INLINE float16x8_t vtaylor_polyq_f16(
float16x8_t x,
const float16_t * coeffs)
{
float16x8_t A = vfmasq(vdupq_n_f16(coeffs[4]), x, coeffs[0]);
float16x8_t B = vfmasq(vdupq_n_f16(coeffs[6]), x, coeffs[2]);
float16x8_t C = vfmasq(vdupq_n_f16(coeffs[5]), x, coeffs[1]);
float16x8_t D = vfmasq(vdupq_n_f16(coeffs[7]), x, coeffs[3]);
float16x8_t x2 = vmulq(x, x);
float16x8_t x4 = vmulq(x2, x2);
float16x8_t res = vfmaq(vfmaq_f16(A, B, x2), vfmaq_f16(C, D, x2), x4);
return res;
}
#define VMANT_EXP_F16(x) \
any16x8_t r; \
int16x8_t n; \
\
r.f = x; \
n = r.i >> 10; \
n = n - 15; \
r.i = r.i - (n << 10);\
\
vecExpUnBiased = n; \
vecTmpFlt1 = r.f;
__STATIC_INLINE float16x8_t vlogq_f16(float16x8_t vecIn)
{
q15x8_t vecExpUnBiased;
float16x8_t vecTmpFlt0, vecTmpFlt1;
float16x8_t vecAcc0, vecAcc1, vecAcc2, vecAcc3;
float16x8_t vecExpUnBiasedFlt;
/*
* extract exponent
*/
VMANT_EXP_F16(vecIn);
vecTmpFlt0 = vecTmpFlt1 * vecTmpFlt1;
/*
* a = (__logf_lut_f16[4] * r.f) + (__logf_lut_f16[0]);
*/
vecAcc0 = vdupq_n_f16(__logf_lut_f16[0]);
vecAcc0 = vfmaq(vecAcc0, vecTmpFlt1, __logf_lut_f16[4]);
/*
* b = (__logf_lut_f16[6] * r.f) + (__logf_lut_f16[2]);
*/
vecAcc1 = vdupq_n_f16(__logf_lut_f16[2]);
vecAcc1 = vfmaq(vecAcc1, vecTmpFlt1, __logf_lut_f16[6]);
/*
* c = (__logf_lut_f16[5] * r.f) + (__logf_lut_f16[1]);
*/
vecAcc2 = vdupq_n_f16(__logf_lut_f16[1]);
vecAcc2 = vfmaq(vecAcc2, vecTmpFlt1, __logf_lut_f16[5]);
/*
* d = (__logf_lut_f16[7] * r.f) + (__logf_lut_f16[3]);
*/
vecAcc3 = vdupq_n_f16(__logf_lut_f16[3]);
vecAcc3 = vfmaq(vecAcc3, vecTmpFlt1, __logf_lut_f16[7]);
/*
* a = a + b * xx;
*/
vecAcc0 = vfmaq(vecAcc0, vecAcc1, vecTmpFlt0);
/*
* c = c + d * xx;
*/
vecAcc2 = vfmaq(vecAcc2, vecAcc3, vecTmpFlt0);
/*
* xx = xx * xx;
*/
vecTmpFlt0 = vecTmpFlt0 * vecTmpFlt0;
vecExpUnBiasedFlt = vcvtq_f16_s16(vecExpUnBiased);
/*
* r.f = a + c * xx;
*/
vecAcc0 = vfmaq(vecAcc0, vecAcc2, vecTmpFlt0);
/*
* add exponent
* r.f = r.f + ((float32_t) m) * __logf_rng_f16;
*/
vecAcc0 = vfmaq(vecAcc0, vecExpUnBiasedFlt, __logf_rng_f16);
// set log0 down to -inf
vecAcc0 = vdupq_m_n_f16(vecAcc0, -(_Float16)F16INFINITY, vcmpeqq_n_f16(vecIn, 0.0f));
return vecAcc0;
}
__STATIC_INLINE float16x8_t vexpq_f16(
float16x8_t x)
{
// Perform range reduction [-log(2),log(2)]
int16x8_t m = vcvtq_s16_f16(vmulq_n_f16(x, 1.4426950408f16));
float16x8_t val = vfmsq_f16(x, vcvtq_f16_s16(m), vdupq_n_f16(0.6931471805f16));
// Polynomial Approximation
float16x8_t poly = vtaylor_polyq_f16(val, exp_tab_f16);
// Reconstruct
poly = (float16x8_t) (vqaddq_s16((int16x8_t) (poly), vqshlq_n_s16(m, 10)));
poly = vdupq_m_n_f16(poly, 0.0f16, vcmpltq_n_s16(m, -14));
return poly;
}
__STATIC_INLINE float16x8_t arm_vec_exponent_f16(float16x8_t x, int16_t nb)
{
float16x8_t r = x;
nb--;
while (nb > 0) {
r = vmulq(r, x);
nb--;
}
return (r);
}
__STATIC_INLINE f16x8_t vpowq_f16(
f16x8_t val,
f16x8_t n)
{
return vexpq_f16(vmulq_f16(n, vlogq_f16(val)));
}
#define INV_NEWTON_INIT_F16 0x7773
__STATIC_INLINE f16x8_t vrecip_f16(f16x8_t vecIn)
{
f16x8_t vecSx, vecW, vecTmp;
any16x8_t v;
vecSx = vabsq(vecIn);
v.f = vecIn;
v.i = vsubq(vdupq_n_s16(INV_NEWTON_INIT_F16), v.i);
vecW = vmulq(vecSx, v.f);
// v.f = v.f * (8 + w * (-28 + w * (56 + w * (-70 + w *(56 + w * (-28 + w * (8 - w)))))));
vecTmp = vsubq(vdupq_n_f16(8.0f16), vecW);
vecTmp = vfmasq_n_f16(vecW, vecTmp, -28.0f16);
vecTmp = vfmasq_n_f16(vecW, vecTmp, 56.0f16);
vecTmp = vfmasq_n_f16(vecW, vecTmp, -70.0f16);
vecTmp = vfmasq_n_f16(vecW, vecTmp, 56.0f16);
vecTmp = vfmasq_n_f16(vecW, vecTmp, -28.0f16);
vecTmp = vfmasq_n_f16(vecW, vecTmp, 8.0f16);
v.f = vmulq(v.f, vecTmp);
v.f = vdupq_m_n_f16(v.f, F16INFINITY, vcmpeqq_n_f16(vecIn, 0.0f));
/*
* restore sign
*/
v.f = vnegq_m(v.f, v.f, vcmpltq_n_f16(vecIn, 0.0f));
return v.f;
}
__STATIC_INLINE f16x8_t vtanhq_f16(
f16x8_t val)
{
f16x8_t x =
vminnmq_f16(vmaxnmq_f16(val, vdupq_n_f16(-10.f16)), vdupq_n_f16(10.0f16));
f16x8_t exp2x = vexpq_f16(vmulq_n_f16(x, 2.f16));
f16x8_t num = vsubq_n_f16(exp2x, 1.f16);
f16x8_t den = vaddq_n_f16(exp2x, 1.f16);
f16x8_t tanh = vmulq_f16(num, vrecip_f16(den));
return tanh;
}
#endif /* defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)*/
#ifdef __cplusplus
}
#endif
#endif /* ARM FLOAT16 SUPPORTED */
#endif /* _ARM_VEC_MATH_F16_H */
/**
*
* End of file.
*/

864
sdk/CMSIS/DSP/Include/dsp/basic_math_functions.h Normal file
View file

@ -0,0 +1,864 @@
/******************************************************************************
* @file basic_math_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef BASIC_MATH_FUNCTIONS_H_
#define BASIC_MATH_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupMath Basic Math Functions
*/
/**
* @brief Q7 vector multiplication.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_mult_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Q15 vector multiplication.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_mult_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Q31 vector multiplication.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_mult_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector multiplication.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_mult_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector multiplication.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_mult_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector addition.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_add_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector addition.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_add_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Q7 vector addition.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_add_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Q15 vector addition.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_add_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Q31 vector addition.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_add_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector subtraction.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_sub_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector subtraction.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_sub_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Q7 vector subtraction.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_sub_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Q15 vector subtraction.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_sub_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Q31 vector subtraction.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_sub_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Multiplies a floating-point vector by a scalar.
* @param[in] pSrc points to the input vector
* @param[in] scale scale factor to be applied
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_scale_f32(
const float32_t * pSrc,
float32_t scale,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Multiplies a floating-point vector by a scalar.
* @param[in] pSrc points to the input vector
* @param[in] scale scale factor to be applied
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_scale_f64(
const float64_t * pSrc,
float64_t scale,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Multiplies a Q7 vector by a scalar.
* @param[in] pSrc points to the input vector
* @param[in] scaleFract fractional portion of the scale value
* @param[in] shift number of bits to shift the result by
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_scale_q7(
const q7_t * pSrc,
q7_t scaleFract,
int8_t shift,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Multiplies a Q15 vector by a scalar.
* @param[in] pSrc points to the input vector
* @param[in] scaleFract fractional portion of the scale value
* @param[in] shift number of bits to shift the result by
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_scale_q15(
const q15_t * pSrc,
q15_t scaleFract,
int8_t shift,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Multiplies a Q31 vector by a scalar.
* @param[in] pSrc points to the input vector
* @param[in] scaleFract fractional portion of the scale value
* @param[in] shift number of bits to shift the result by
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_scale_q31(
const q31_t * pSrc,
q31_t scaleFract,
int8_t shift,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Q7 vector absolute value.
* @param[in] pSrc points to the input buffer
* @param[out] pDst points to the output buffer
* @param[in] blockSize number of samples in each vector
*/
void arm_abs_q7(
const q7_t * pSrc,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector absolute value.
* @param[in] pSrc points to the input buffer
* @param[out] pDst points to the output buffer
* @param[in] blockSize number of samples in each vector
*/
void arm_abs_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector absolute value.
* @param[in] pSrc points to the input buffer
* @param[out] pDst points to the output buffer
* @param[in] blockSize number of samples in each vector
*/
void arm_abs_f64(
const float64_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Q15 vector absolute value.
* @param[in] pSrc points to the input buffer
* @param[out] pDst points to the output buffer
* @param[in] blockSize number of samples in each vector
*/
void arm_abs_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Q31 vector absolute value.
* @param[in] pSrc points to the input buffer
* @param[out] pDst points to the output buffer
* @param[in] blockSize number of samples in each vector
*/
void arm_abs_q31(
const q31_t * pSrc,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Dot product of floating-point vectors.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] blockSize number of samples in each vector
* @param[out] result output result returned here
*/
void arm_dot_prod_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
uint32_t blockSize,
float32_t * result);
/**
* @brief Dot product of floating-point vectors.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] blockSize number of samples in each vector
* @param[out] result output result returned here
*/
void arm_dot_prod_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
uint32_t blockSize,
float64_t * result);
/**
* @brief Dot product of Q7 vectors.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] blockSize number of samples in each vector
* @param[out] result output result returned here
*/
void arm_dot_prod_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
uint32_t blockSize,
q31_t * result);
/**
* @brief Dot product of Q15 vectors.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] blockSize number of samples in each vector
* @param[out] result output result returned here
*/
void arm_dot_prod_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
uint32_t blockSize,
q63_t * result);
/**
* @brief Dot product of Q31 vectors.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] blockSize number of samples in each vector
* @param[out] result output result returned here
*/
void arm_dot_prod_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
uint32_t blockSize,
q63_t * result);
/**
* @brief Shifts the elements of a Q7 vector a specified number of bits.
* @param[in] pSrc points to the input vector
* @param[in] shiftBits number of bits to shift. A positive value shifts left; a negative value shifts right.
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_shift_q7(
const q7_t * pSrc,
int8_t shiftBits,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Shifts the elements of a Q15 vector a specified number of bits.
* @param[in] pSrc points to the input vector
* @param[in] shiftBits number of bits to shift. A positive value shifts left; a negative value shifts right.
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_shift_q15(
const q15_t * pSrc,
int8_t shiftBits,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Shifts the elements of a Q31 vector a specified number of bits.
* @param[in] pSrc points to the input vector
* @param[in] shiftBits number of bits to shift. A positive value shifts left; a negative value shifts right.
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_shift_q31(
const q31_t * pSrc,
int8_t shiftBits,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Adds a constant offset to a floating-point vector.
* @param[in] pSrc points to the input vector
* @param[in] offset is the offset to be added
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_offset_f64(
const float64_t * pSrc,
float64_t offset,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Adds a constant offset to a floating-point vector.
* @param[in] pSrc points to the input vector
* @param[in] offset is the offset to be added
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_offset_f32(
const float32_t * pSrc,
float32_t offset,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Adds a constant offset to a Q7 vector.
* @param[in] pSrc points to the input vector
* @param[in] offset is the offset to be added
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_offset_q7(
const q7_t * pSrc,
q7_t offset,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Adds a constant offset to a Q15 vector.
* @param[in] pSrc points to the input vector
* @param[in] offset is the offset to be added
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_offset_q15(
const q15_t * pSrc,
q15_t offset,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Adds a constant offset to a Q31 vector.
* @param[in] pSrc points to the input vector
* @param[in] offset is the offset to be added
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_offset_q31(
const q31_t * pSrc,
q31_t offset,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Negates the elements of a floating-point vector.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_negate_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Negates the elements of a floating-point vector.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_negate_f64(
const float64_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Negates the elements of a Q7 vector.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_negate_q7(
const q7_t * pSrc,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Negates the elements of a Q15 vector.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_negate_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Negates the elements of a Q31 vector.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_negate_q31(
const q31_t * pSrc,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise AND of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_and_u16(
const uint16_t * pSrcA,
const uint16_t * pSrcB,
uint16_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise AND of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_and_u32(
const uint32_t * pSrcA,
const uint32_t * pSrcB,
uint32_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise AND of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_and_u8(
const uint8_t * pSrcA,
const uint8_t * pSrcB,
uint8_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise OR of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_or_u16(
const uint16_t * pSrcA,
const uint16_t * pSrcB,
uint16_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise OR of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_or_u32(
const uint32_t * pSrcA,
const uint32_t * pSrcB,
uint32_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise OR of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_or_u8(
const uint8_t * pSrcA,
const uint8_t * pSrcB,
uint8_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise NOT of a fixed-point vector.
* @param[in] pSrc points to input vector
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_not_u16(
const uint16_t * pSrc,
uint16_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise NOT of a fixed-point vector.
* @param[in] pSrc points to input vector
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_not_u32(
const uint32_t * pSrc,
uint32_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise NOT of a fixed-point vector.
* @param[in] pSrc points to input vector
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_not_u8(
const uint8_t * pSrc,
uint8_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise XOR of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_xor_u16(
const uint16_t * pSrcA,
const uint16_t * pSrcB,
uint16_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise XOR of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_xor_u32(
const uint32_t * pSrcA,
const uint32_t * pSrcB,
uint32_t * pDst,
uint32_t blockSize);
/**
* @brief Compute the logical bitwise XOR of two fixed-point vectors.
* @param[in] pSrcA points to input vector A
* @param[in] pSrcB points to input vector B
* @param[out] pDst points to output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_xor_u8(
const uint8_t * pSrcA,
const uint8_t * pSrcB,
uint8_t * pDst,
uint32_t blockSize);
/**
@brief Elementwise floating-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
void arm_clip_f32(const float32_t * pSrc,
float32_t * pDst,
float32_t low,
float32_t high,
uint32_t numSamples);
/**
@brief Elementwise fixed-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
void arm_clip_q31(const q31_t * pSrc,
q31_t * pDst,
q31_t low,
q31_t high,
uint32_t numSamples);
/**
@brief Elementwise fixed-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
void arm_clip_q15(const q15_t * pSrc,
q15_t * pDst,
q15_t low,
q15_t high,
uint32_t numSamples);
/**
@brief Elementwise fixed-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
void arm_clip_q7(const q7_t * pSrc,
q7_t * pDst,
q7_t low,
q7_t high,
uint32_t numSamples);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _BASIC_MATH_FUNCTIONS_H_ */

165
sdk/CMSIS/DSP/Include/dsp/basic_math_functions_f16.h Normal file
View file

@ -0,0 +1,165 @@
/******************************************************************************
* @file basic_math_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef BASIC_MATH_FUNCTIONS_F16_H_
#define BASIC_MATH_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Floating-point vector addition.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_add_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector subtraction.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_sub_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Multiplies a floating-point vector by a scalar.
* @param[in] pSrc points to the input vector
* @param[in] scale scale factor to be applied
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_scale_f16(
const float16_t * pSrc,
float16_t scale,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Floating-point vector absolute value.
* @param[in] pSrc points to the input buffer
* @param[out] pDst points to the output buffer
* @param[in] blockSize number of samples in each vector
*/
void arm_abs_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Adds a constant offset to a floating-point vector.
* @param[in] pSrc points to the input vector
* @param[in] offset is the offset to be added
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_offset_f16(
const float16_t * pSrc,
float16_t offset,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Dot product of floating-point vectors.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] blockSize number of samples in each vector
* @param[out] result output result returned here
*/
void arm_dot_prod_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
uint32_t blockSize,
float16_t * result);
/**
* @brief Floating-point vector multiplication.
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in each vector
*/
void arm_mult_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Negates the elements of a floating-point vector.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] blockSize number of samples in the vector
*/
void arm_negate_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
/**
@brief Elementwise floating-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
void arm_clip_f16(const float16_t * pSrc,
float16_t * pDst,
float16_t low,
float16_t high,
uint32_t numSamples);
#endif /* defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _BASIC_MATH_FUNCTIONS_F16_H_ */

86
sdk/CMSIS/DSP/Include/dsp/bayes_functions.h Normal file
View file

@ -0,0 +1,86 @@
/******************************************************************************
* @file bayes_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef BAYES_FUNCTIONS_H_
#define BAYES_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/statistics_functions.h"
/**
* @defgroup groupBayes Bayesian estimators
*
* Implement the naive gaussian Bayes estimator.
* The training must be done from scikit-learn.
*
* The parameters can be easily
* generated from the scikit-learn object. Some examples are given in
* DSP/Testing/PatternGeneration/Bayes.py
*/
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @brief Instance structure for Naive Gaussian Bayesian estimator.
*/
typedef struct
{
uint32_t vectorDimension; /**< Dimension of vector space */
uint32_t numberOfClasses; /**< Number of different classes */
const float32_t *theta; /**< Mean values for the Gaussians */
const float32_t *sigma; /**< Variances for the Gaussians */
const float32_t *classPriors; /**< Class prior probabilities */
float32_t epsilon; /**< Additive value to variances */
} arm_gaussian_naive_bayes_instance_f32;
/**
* @brief Naive Gaussian Bayesian Estimator
*
* @param[in] S points to a naive bayes instance structure
* @param[in] in points to the elements of the input vector.
* @param[out] *pOutputProbabilities points to a buffer of length numberOfClasses containing estimated probabilities
* @param[out] *pBufferB points to a temporary buffer of length numberOfClasses
* @return The predicted class
*/
uint32_t arm_gaussian_naive_bayes_predict_f32(const arm_gaussian_naive_bayes_instance_f32 *S,
const float32_t * in,
float32_t *pOutputProbabilities,
float32_t *pBufferB);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _BAYES_FUNCTIONS_H_ */

77
sdk/CMSIS/DSP/Include/dsp/bayes_functions_f16.h Normal file
View file

@ -0,0 +1,77 @@
/******************************************************************************
* @file bayes_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef BAYES_FUNCTIONS_F16_H_
#define BAYES_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/statistics_functions_f16.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Instance structure for Naive Gaussian Bayesian estimator.
*/
typedef struct
{
uint32_t vectorDimension; /**< Dimension of vector space */
uint32_t numberOfClasses; /**< Number of different classes */
const float16_t *theta; /**< Mean values for the Gaussians */
const float16_t *sigma; /**< Variances for the Gaussians */
const float16_t *classPriors; /**< Class prior probabilities */
float16_t epsilon; /**< Additive value to variances */
} arm_gaussian_naive_bayes_instance_f16;
/**
* @brief Naive Gaussian Bayesian Estimator
*
* @param[in] S points to a naive bayes instance structure
* @param[in] in points to the elements of the input vector.
* @param[out] *pOutputProbabilities points to a buffer of length numberOfClasses containing estimated probabilities
* @param[out] *pBufferB points to a temporary buffer of length numberOfClasses
* @return The predicted class
*/
uint32_t arm_gaussian_naive_bayes_predict_f16(const arm_gaussian_naive_bayes_instance_f16 *S,
const float16_t * in,
float16_t *pOutputProbabilities,
float16_t *pBufferB);
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _BAYES_FUNCTIONS_F16_H_ */

345
sdk/CMSIS/DSP/Include/dsp/complex_math_functions.h Normal file
View file

@ -0,0 +1,345 @@
/******************************************************************************
* @file complex_math_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef COMPLEX_MATH_FUNCTIONS_H_
#define COMPLEX_MATH_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/fast_math_functions.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupCmplxMath Complex Math Functions
* This set of functions operates on complex data vectors.
* The data in the complex arrays is stored in an interleaved fashion
* (real, imag, real, imag, ...).
* In the API functions, the number of samples in a complex array refers
* to the number of complex values; the array contains twice this number of
* real values.
*/
/**
* @brief Floating-point complex conjugate.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_conj_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t numSamples);
/**
* @brief Q31 complex conjugate.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_conj_q31(
const q31_t * pSrc,
q31_t * pDst,
uint32_t numSamples);
/**
* @brief Q15 complex conjugate.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_conj_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex magnitude squared
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_squared_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex magnitude squared
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_squared_f64(
const float64_t * pSrc,
float64_t * pDst,
uint32_t numSamples);
/**
* @brief Q31 complex magnitude squared
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_squared_q31(
const q31_t * pSrc,
q31_t * pDst,
uint32_t numSamples);
/**
* @brief Q15 complex magnitude squared
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_squared_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex magnitude
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex magnitude
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_f64(
const float64_t * pSrc,
float64_t * pDst,
uint32_t numSamples);
/**
* @brief Q31 complex magnitude
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_q31(
const q31_t * pSrc,
q31_t * pDst,
uint32_t numSamples);
/**
* @brief Q15 complex magnitude
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t numSamples);
/**
* @brief Q15 complex magnitude
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_fast_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t numSamples);
/**
* @brief Q15 complex dot product
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] numSamples number of complex samples in each vector
* @param[out] realResult real part of the result returned here
* @param[out] imagResult imaginary part of the result returned here
*/
void arm_cmplx_dot_prod_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
uint32_t numSamples,
q31_t * realResult,
q31_t * imagResult);
/**
* @brief Q31 complex dot product
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] numSamples number of complex samples in each vector
* @param[out] realResult real part of the result returned here
* @param[out] imagResult imaginary part of the result returned here
*/
void arm_cmplx_dot_prod_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
uint32_t numSamples,
q63_t * realResult,
q63_t * imagResult);
/**
* @brief Floating-point complex dot product
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] numSamples number of complex samples in each vector
* @param[out] realResult real part of the result returned here
* @param[out] imagResult imaginary part of the result returned here
*/
void arm_cmplx_dot_prod_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
uint32_t numSamples,
float32_t * realResult,
float32_t * imagResult);
/**
* @brief Q15 complex-by-real multiplication
* @param[in] pSrcCmplx points to the complex input vector
* @param[in] pSrcReal points to the real input vector
* @param[out] pCmplxDst points to the complex output vector
* @param[in] numSamples number of samples in each vector
*/
void arm_cmplx_mult_real_q15(
const q15_t * pSrcCmplx,
const q15_t * pSrcReal,
q15_t * pCmplxDst,
uint32_t numSamples);
/**
* @brief Q31 complex-by-real multiplication
* @param[in] pSrcCmplx points to the complex input vector
* @param[in] pSrcReal points to the real input vector
* @param[out] pCmplxDst points to the complex output vector
* @param[in] numSamples number of samples in each vector
*/
void arm_cmplx_mult_real_q31(
const q31_t * pSrcCmplx,
const q31_t * pSrcReal,
q31_t * pCmplxDst,
uint32_t numSamples);
/**
* @brief Floating-point complex-by-real multiplication
* @param[in] pSrcCmplx points to the complex input vector
* @param[in] pSrcReal points to the real input vector
* @param[out] pCmplxDst points to the complex output vector
* @param[in] numSamples number of samples in each vector
*/
void arm_cmplx_mult_real_f32(
const float32_t * pSrcCmplx,
const float32_t * pSrcReal,
float32_t * pCmplxDst,
uint32_t numSamples);
/**
* @brief Q15 complex-by-complex multiplication
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_mult_cmplx_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
q15_t * pDst,
uint32_t numSamples);
/**
* @brief Q31 complex-by-complex multiplication
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_mult_cmplx_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex-by-complex multiplication
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_mult_cmplx_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
float32_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex-by-complex multiplication
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_mult_cmplx_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
float64_t * pDst,
uint32_t numSamples);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _COMPLEX_MATH_FUNCTIONS_H_ */

123
sdk/CMSIS/DSP/Include/dsp/complex_math_functions_f16.h Normal file
View file

@ -0,0 +1,123 @@
/******************************************************************************
* @file complex_math_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef COMPLEX_MATH_FUNCTIONS_F16_H_
#define COMPLEX_MATH_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/fast_math_functions_f16.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Floating-point complex conjugate.
* @param[in] pSrc points to the input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_conj_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex magnitude squared
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_squared_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex magnitude
* @param[in] pSrc points to the complex input vector
* @param[out] pDst points to the real output vector
* @param[in] numSamples number of complex samples in the input vector
*/
void arm_cmplx_mag_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t numSamples);
/**
* @brief Floating-point complex dot product
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] numSamples number of complex samples in each vector
* @param[out] realResult real part of the result returned here
* @param[out] imagResult imaginary part of the result returned here
*/
void arm_cmplx_dot_prod_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
uint32_t numSamples,
float16_t * realResult,
float16_t * imagResult);
/**
* @brief Floating-point complex-by-real multiplication
* @param[in] pSrcCmplx points to the complex input vector
* @param[in] pSrcReal points to the real input vector
* @param[out] pCmplxDst points to the complex output vector
* @param[in] numSamples number of samples in each vector
*/
void arm_cmplx_mult_real_f16(
const float16_t * pSrcCmplx,
const float16_t * pSrcReal,
float16_t * pCmplxDst,
uint32_t numSamples);
/**
* @brief Floating-point complex-by-complex multiplication
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[out] pDst points to the output vector
* @param[in] numSamples number of complex samples in each vector
*/
void arm_cmplx_mult_cmplx_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
float16_t * pDst,
uint32_t numSamples);
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _COMPLEX_MATH_FUNCTIONS_F16_H_ */

785
sdk/CMSIS/DSP/Include/dsp/controller_functions.h Normal file
View file

@ -0,0 +1,785 @@
/******************************************************************************
* @file controller_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef CONTROLLER_FUNCTIONS_H_
#define CONTROLLER_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @brief Macros required for SINE and COSINE Controller functions
*/
#define CONTROLLER_Q31_SHIFT (32 - 9)
/* 1.31(q31) Fixed value of 2/360 */
/* -1 to +1 is divided into 360 values so total spacing is (2/360) */
#define INPUT_SPACING 0xB60B61
/**
* @defgroup groupController Controller Functions
*/
/**
@ingroup groupController
*/
/**
@defgroup SinCos Sine Cosine
Computes the trigonometric sine and cosine values using a combination of table lookup
and linear interpolation.
There are separate functions for Q31 and floating-point data types.
The input to the floating-point version is in degrees while the
fixed-point Q31 have a scaled input with the range
[-1 0.9999] mapping to [-180 +180] degrees.
The floating point function also allows values that are out of the usual range. When this happens, the function will
take extra time to adjust the input value to the range of [-180 180].
The result is accurate to 5 digits after the decimal point.
The implementation is based on table lookup using 360 values together with linear interpolation.
The steps used are:
-# Calculation of the nearest integer table index.
-# Compute the fractional portion (fract) of the input.
-# Fetch the value corresponding to \c index from sine table to \c y0 and also value from \c index+1 to \c y1.
-# Sine value is computed as <code> *psinVal = y0 + (fract * (y1 - y0))</code>.
-# Fetch the value corresponding to \c index from cosine table to \c y0 and also value from \c index+1 to \c y1.
-# Cosine value is computed as <code> *pcosVal = y0 + (fract * (y1 - y0))</code>.
*/
/**
* @brief Floating-point sin_cos function.
* @param[in] theta input value in degrees
* @param[out] pSinVal points to the processed sine output.
* @param[out] pCosVal points to the processed cos output.
*/
void arm_sin_cos_f32(
float32_t theta,
float32_t * pSinVal,
float32_t * pCosVal);
/**
* @brief Q31 sin_cos function.
* @param[in] theta scaled input value in degrees
* @param[out] pSinVal points to the processed sine output.
* @param[out] pCosVal points to the processed cosine output.
*/
void arm_sin_cos_q31(
q31_t theta,
q31_t * pSinVal,
q31_t * pCosVal);
/**
@ingroup groupController
*/
/**
* @defgroup PID PID Motor Control
*
* A Proportional Integral Derivative (PID) controller is a generic feedback control
* loop mechanism widely used in industrial control systems.
* A PID controller is the most commonly used type of feedback controller.
*
* This set of functions implements (PID) controllers
* for Q15, Q31, and floating-point data types. The functions operate on a single sample
* of data and each call to the function returns a single processed value.
* <code>S</code> points to an instance of the PID control data structure. <code>in</code>
* is the input sample value. The functions return the output value.
*
* \par Algorithm:
* <pre>
* y[n] = y[n-1] + A0 * x[n] + A1 * x[n-1] + A2 * x[n-2]
* A0 = Kp + Ki + Kd
* A1 = (-Kp ) - (2 * Kd )
* A2 = Kd
* </pre>
*
* \par
* where \c Kp is proportional constant, \c Ki is Integral constant and \c Kd is Derivative constant
*
* \par
* \image html PID.gif "Proportional Integral Derivative Controller"
*
* \par
* The PID controller calculates an "error" value as the difference between
* the measured output and the reference input.
* The controller attempts to minimize the error by adjusting the process control inputs.
* The proportional value determines the reaction to the current error,
* the integral value determines the reaction based on the sum of recent errors,
* and the derivative value determines the reaction based on the rate at which the error has been changing.
*
* \par Instance Structure
* The Gains A0, A1, A2 and state variables for a PID controller are stored together in an instance data structure.
* A separate instance structure must be defined for each PID Controller.
* There are separate instance structure declarations for each of the 3 supported data types.
*
* \par Reset Functions
* There is also an associated reset function for each data type which clears the state array.
*
* \par Initialization Functions
* There is also an associated initialization function for each data type.
* The initialization function performs the following operations:
* - Initializes the Gains A0, A1, A2 from Kp,Ki, Kd gains.
* - Zeros out the values in the state buffer.
*
* \par
* Instance structure cannot be placed into a const data section and it is recommended to use the initialization function.
*
* \par Fixed-Point Behavior
* Care must be taken when using the fixed-point versions of the PID Controller functions.
* In particular, the overflow and saturation behavior of the accumulator used in each function must be considered.
* Refer to the function specific documentation below for usage guidelines.
*/
/**
* @ingroup PID
* @brief Instance structure for the Q15 PID Control.
*/
typedef struct
{
q15_t A0; /**< The derived gain, A0 = Kp + Ki + Kd . */
#if !defined (ARM_MATH_DSP)
q15_t A1; /**< The derived gain A1 = -Kp - 2Kd */
q15_t A2; /**< The derived gain A1 = Kd. */
#else
q31_t A1; /**< The derived gain A1 = -Kp - 2Kd | Kd.*/
#endif
q15_t state[3]; /**< The state array of length 3. */
q15_t Kp; /**< The proportional gain. */
q15_t Ki; /**< The integral gain. */
q15_t Kd; /**< The derivative gain. */
} arm_pid_instance_q15;
/**
* @ingroup PID
* @brief Instance structure for the Q31 PID Control.
*/
typedef struct
{
q31_t A0; /**< The derived gain, A0 = Kp + Ki + Kd . */
q31_t A1; /**< The derived gain, A1 = -Kp - 2Kd. */
q31_t A2; /**< The derived gain, A2 = Kd . */
q31_t state[3]; /**< The state array of length 3. */
q31_t Kp; /**< The proportional gain. */
q31_t Ki; /**< The integral gain. */
q31_t Kd; /**< The derivative gain. */
} arm_pid_instance_q31;
/**
* @ingroup PID
* @brief Instance structure for the floating-point PID Control.
*/
typedef struct
{
float32_t A0; /**< The derived gain, A0 = Kp + Ki + Kd . */
float32_t A1; /**< The derived gain, A1 = -Kp - 2Kd. */
float32_t A2; /**< The derived gain, A2 = Kd . */
float32_t state[3]; /**< The state array of length 3. */
float32_t Kp; /**< The proportional gain. */
float32_t Ki; /**< The integral gain. */
float32_t Kd; /**< The derivative gain. */
} arm_pid_instance_f32;
/**
* @brief Initialization function for the floating-point PID Control.
* @param[in,out] S points to an instance of the PID structure.
* @param[in] resetStateFlag flag to reset the state. 0 = no change in state 1 = reset the state.
*/
void arm_pid_init_f32(
arm_pid_instance_f32 * S,
int32_t resetStateFlag);
/**
* @brief Reset function for the floating-point PID Control.
* @param[in,out] S is an instance of the floating-point PID Control structure
*/
void arm_pid_reset_f32(
arm_pid_instance_f32 * S);
/**
* @brief Initialization function for the Q31 PID Control.
* @param[in,out] S points to an instance of the Q15 PID structure.
* @param[in] resetStateFlag flag to reset the state. 0 = no change in state 1 = reset the state.
*/
void arm_pid_init_q31(
arm_pid_instance_q31 * S,
int32_t resetStateFlag);
/**
* @brief Reset function for the Q31 PID Control.
* @param[in,out] S points to an instance of the Q31 PID Control structure
*/
void arm_pid_reset_q31(
arm_pid_instance_q31 * S);
/**
* @brief Initialization function for the Q15 PID Control.
* @param[in,out] S points to an instance of the Q15 PID structure.
* @param[in] resetStateFlag flag to reset the state. 0 = no change in state 1 = reset the state.
*/
void arm_pid_init_q15(
arm_pid_instance_q15 * S,
int32_t resetStateFlag);
/**
* @brief Reset function for the Q15 PID Control.
* @param[in,out] S points to an instance of the q15 PID Control structure
*/
void arm_pid_reset_q15(
arm_pid_instance_q15 * S);
/**
* @ingroup PID
* @brief Process function for the floating-point PID Control.
* @param[in,out] S is an instance of the floating-point PID Control structure
* @param[in] in input sample to process
* @return processed output sample.
*/
__STATIC_FORCEINLINE float32_t arm_pid_f32(
arm_pid_instance_f32 * S,
float32_t in)
{
float32_t out;
/* y[n] = y[n-1] + A0 * x[n] + A1 * x[n-1] + A2 * x[n-2] */
out = (S->A0 * in) +
(S->A1 * S->state[0]) + (S->A2 * S->state[1]) + (S->state[2]);
/* Update state */
S->state[1] = S->state[0];
S->state[0] = in;
S->state[2] = out;
/* return to application */
return (out);
}
/**
@ingroup PID
@brief Process function for the Q31 PID Control.
@param[in,out] S points to an instance of the Q31 PID Control structure
@param[in] in input sample to process
@return processed output sample.
\par Scaling and Overflow Behavior
The function is implemented using an internal 64-bit accumulator.
The accumulator has a 2.62 format and maintains full precision of the intermediate multiplication results but provides only a single guard bit.
Thus, if the accumulator result overflows it wraps around rather than clip.
In order to avoid overflows completely the input signal must be scaled down by 2 bits as there are four additions.
After all multiply-accumulates are performed, the 2.62 accumulator is truncated to 1.32 format and then saturated to 1.31 format.
*/
__STATIC_FORCEINLINE q31_t arm_pid_q31(
arm_pid_instance_q31 * S,
q31_t in)
{
q63_t acc;
q31_t out;
/* acc = A0 * x[n] */
acc = (q63_t) S->A0 * in;
/* acc += A1 * x[n-1] */
acc += (q63_t) S->A1 * S->state[0];
/* acc += A2 * x[n-2] */
acc += (q63_t) S->A2 * S->state[1];
/* convert output to 1.31 format to add y[n-1] */
out = (q31_t) (acc >> 31U);
/* out += y[n-1] */
out += S->state[2];
/* Update state */
S->state[1] = S->state[0];
S->state[0] = in;
S->state[2] = out;
/* return to application */
return (out);
}
/**
@ingroup PID
@brief Process function for the Q15 PID Control.
@param[in,out] S points to an instance of the Q15 PID Control structure
@param[in] in input sample to process
@return processed output sample.
\par Scaling and Overflow Behavior
The function is implemented using a 64-bit internal accumulator.
Both Gains and state variables are represented in 1.15 format and multiplications yield a 2.30 result.
The 2.30 intermediate results are accumulated in a 64-bit accumulator in 34.30 format.
There is no risk of internal overflow with this approach and the full precision of intermediate multiplications is preserved.
After all additions have been performed, the accumulator is truncated to 34.15 format by discarding low 15 bits.
Lastly, the accumulator is saturated to yield a result in 1.15 format.
*/
__STATIC_FORCEINLINE q15_t arm_pid_q15(
arm_pid_instance_q15 * S,
q15_t in)
{
q63_t acc;
q15_t out;
#if defined (ARM_MATH_DSP)
/* Implementation of PID controller */
/* acc = A0 * x[n] */
acc = (q31_t) __SMUAD((uint32_t)S->A0, (uint32_t)in);
/* acc += A1 * x[n-1] + A2 * x[n-2] */
acc = (q63_t)__SMLALD((uint32_t)S->A1, (uint32_t)read_q15x2 (S->state), (uint64_t)acc);
#else
/* acc = A0 * x[n] */
acc = ((q31_t) S->A0) * in;
/* acc += A1 * x[n-1] + A2 * x[n-2] */
acc += (q31_t) S->A1 * S->state[0];
acc += (q31_t) S->A2 * S->state[1];
#endif
/* acc += y[n-1] */
acc += (q31_t) S->state[2] << 15;
/* saturate the output */
out = (q15_t) (__SSAT((q31_t)(acc >> 15), 16));
/* Update state */
S->state[1] = S->state[0];
S->state[0] = in;
S->state[2] = out;
/* return to application */
return (out);
}
/**
* @ingroup groupController
*/
/**
* @defgroup park Vector Park Transform
*
* Forward Park transform converts the input two-coordinate vector to flux and torque components.
* The Park transform can be used to realize the transformation of the <code>Ialpha</code> and the <code>Ibeta</code> currents
* from the stationary to the moving reference frame and control the spatial relationship between
* the stator vector current and rotor flux vector.
* If we consider the d axis aligned with the rotor flux, the diagram below shows the
* current vector and the relationship from the two reference frames:
* \image html park.gif "Stator current space vector and its component in (a,b) and in the d,q rotating reference frame"
*
* The function operates on a single sample of data and each call to the function returns the processed output.
* The library provides separate functions for Q31 and floating-point data types.
* \par Algorithm
* \image html parkFormula.gif
* where <code>Ialpha</code> and <code>Ibeta</code> are the stator vector components,
* <code>pId</code> and <code>pIq</code> are rotor vector components and <code>cosVal</code> and <code>sinVal</code> are the
* cosine and sine values of theta (rotor flux position).
* \par Fixed-Point Behavior
* Care must be taken when using the Q31 version of the Park transform.
* In particular, the overflow and saturation behavior of the accumulator used must be considered.
* Refer to the function specific documentation below for usage guidelines.
*/
/**
* @ingroup park
* @brief Floating-point Park transform
* @param[in] Ialpha input two-phase vector coordinate alpha
* @param[in] Ibeta input two-phase vector coordinate beta
* @param[out] pId points to output rotor reference frame d
* @param[out] pIq points to output rotor reference frame q
* @param[in] sinVal sine value of rotation angle theta
* @param[in] cosVal cosine value of rotation angle theta
*
* The function implements the forward Park transform.
*
*/
__STATIC_FORCEINLINE void arm_park_f32(
float32_t Ialpha,
float32_t Ibeta,
float32_t * pId,
float32_t * pIq,
float32_t sinVal,
float32_t cosVal)
{
/* Calculate pId using the equation, pId = Ialpha * cosVal + Ibeta * sinVal */
*pId = Ialpha * cosVal + Ibeta * sinVal;
/* Calculate pIq using the equation, pIq = - Ialpha * sinVal + Ibeta * cosVal */
*pIq = -Ialpha * sinVal + Ibeta * cosVal;
}
/**
@ingroup park
@brief Park transform for Q31 version
@param[in] Ialpha input two-phase vector coordinate alpha
@param[in] Ibeta input two-phase vector coordinate beta
@param[out] pId points to output rotor reference frame d
@param[out] pIq points to output rotor reference frame q
@param[in] sinVal sine value of rotation angle theta
@param[in] cosVal cosine value of rotation angle theta
\par Scaling and Overflow Behavior
The function is implemented using an internal 32-bit accumulator.
The accumulator maintains 1.31 format by truncating lower 31 bits of the intermediate multiplication in 2.62 format.
There is saturation on the addition and subtraction, hence there is no risk of overflow.
*/
__STATIC_FORCEINLINE void arm_park_q31(
q31_t Ialpha,
q31_t Ibeta,
q31_t * pId,
q31_t * pIq,
q31_t sinVal,
q31_t cosVal)
{
q31_t product1, product2; /* Temporary variables used to store intermediate results */
q31_t product3, product4; /* Temporary variables used to store intermediate results */
/* Intermediate product is calculated by (Ialpha * cosVal) */
product1 = (q31_t) (((q63_t) (Ialpha) * (cosVal)) >> 31);
/* Intermediate product is calculated by (Ibeta * sinVal) */
product2 = (q31_t) (((q63_t) (Ibeta) * (sinVal)) >> 31);
/* Intermediate product is calculated by (Ialpha * sinVal) */
product3 = (q31_t) (((q63_t) (Ialpha) * (sinVal)) >> 31);
/* Intermediate product is calculated by (Ibeta * cosVal) */
product4 = (q31_t) (((q63_t) (Ibeta) * (cosVal)) >> 31);
/* Calculate pId by adding the two intermediate products 1 and 2 */
*pId = __QADD(product1, product2);
/* Calculate pIq by subtracting the two intermediate products 3 from 4 */
*pIq = __QSUB(product4, product3);
}
/**
* @ingroup groupController
*/
/**
* @defgroup inv_park Vector Inverse Park transform
* Inverse Park transform converts the input flux and torque components to two-coordinate vector.
*
* The function operates on a single sample of data and each call to the function returns the processed output.
* The library provides separate functions for Q31 and floating-point data types.
* \par Algorithm
* \image html parkInvFormula.gif
* where <code>pIalpha</code> and <code>pIbeta</code> are the stator vector components,
* <code>Id</code> and <code>Iq</code> are rotor vector components and <code>cosVal</code> and <code>sinVal</code> are the
* cosine and sine values of theta (rotor flux position).
* \par Fixed-Point Behavior
* Care must be taken when using the Q31 version of the Park transform.
* In particular, the overflow and saturation behavior of the accumulator used must be considered.
* Refer to the function specific documentation below for usage guidelines.
*/
/**
* @ingroup inv_park
* @brief Floating-point Inverse Park transform
* @param[in] Id input coordinate of rotor reference frame d
* @param[in] Iq input coordinate of rotor reference frame q
* @param[out] pIalpha points to output two-phase orthogonal vector axis alpha
* @param[out] pIbeta points to output two-phase orthogonal vector axis beta
* @param[in] sinVal sine value of rotation angle theta
* @param[in] cosVal cosine value of rotation angle theta
*/
__STATIC_FORCEINLINE void arm_inv_park_f32(
float32_t Id,
float32_t Iq,
float32_t * pIalpha,
float32_t * pIbeta,
float32_t sinVal,
float32_t cosVal)
{
/* Calculate pIalpha using the equation, pIalpha = Id * cosVal - Iq * sinVal */
*pIalpha = Id * cosVal - Iq * sinVal;
/* Calculate pIbeta using the equation, pIbeta = Id * sinVal + Iq * cosVal */
*pIbeta = Id * sinVal + Iq * cosVal;
}
/**
@ingroup inv_park
@brief Inverse Park transform for Q31 version
@param[in] Id input coordinate of rotor reference frame d
@param[in] Iq input coordinate of rotor reference frame q
@param[out] pIalpha points to output two-phase orthogonal vector axis alpha
@param[out] pIbeta points to output two-phase orthogonal vector axis beta
@param[in] sinVal sine value of rotation angle theta
@param[in] cosVal cosine value of rotation angle theta
@par Scaling and Overflow Behavior
The function is implemented using an internal 32-bit accumulator.
The accumulator maintains 1.31 format by truncating lower 31 bits of the intermediate multiplication in 2.62 format.
There is saturation on the addition, hence there is no risk of overflow.
*/
__STATIC_FORCEINLINE void arm_inv_park_q31(
q31_t Id,
q31_t Iq,
q31_t * pIalpha,
q31_t * pIbeta,
q31_t sinVal,
q31_t cosVal)
{
q31_t product1, product2; /* Temporary variables used to store intermediate results */
q31_t product3, product4; /* Temporary variables used to store intermediate results */
/* Intermediate product is calculated by (Id * cosVal) */
product1 = (q31_t) (((q63_t) (Id) * (cosVal)) >> 31);
/* Intermediate product is calculated by (Iq * sinVal) */
product2 = (q31_t) (((q63_t) (Iq) * (sinVal)) >> 31);
/* Intermediate product is calculated by (Id * sinVal) */
product3 = (q31_t) (((q63_t) (Id) * (sinVal)) >> 31);
/* Intermediate product is calculated by (Iq * cosVal) */
product4 = (q31_t) (((q63_t) (Iq) * (cosVal)) >> 31);
/* Calculate pIalpha by using the two intermediate products 1 and 2 */
*pIalpha = __QSUB(product1, product2);
/* Calculate pIbeta by using the two intermediate products 3 and 4 */
*pIbeta = __QADD(product4, product3);
}
/**
* @ingroup groupController
*/
/**
* @defgroup clarke Vector Clarke Transform
* Forward Clarke transform converts the instantaneous stator phases into a two-coordinate time invariant vector.
* Generally the Clarke transform uses three-phase currents <code>Ia, Ib and Ic</code> to calculate currents
* in the two-phase orthogonal stator axis <code>Ialpha</code> and <code>Ibeta</code>.
* When <code>Ialpha</code> is superposed with <code>Ia</code> as shown in the figure below
* \image html clarke.gif Stator current space vector and its components in (a,b).
* and <code>Ia + Ib + Ic = 0</code>, in this condition <code>Ialpha</code> and <code>Ibeta</code>
* can be calculated using only <code>Ia</code> and <code>Ib</code>.
*
* The function operates on a single sample of data and each call to the function returns the processed output.
* The library provides separate functions for Q31 and floating-point data types.
* \par Algorithm
* \image html clarkeFormula.gif
* where <code>Ia</code> and <code>Ib</code> are the instantaneous stator phases and
* <code>pIalpha</code> and <code>pIbeta</code> are the two coordinates of time invariant vector.
* \par Fixed-Point Behavior
* Care must be taken when using the Q31 version of the Clarke transform.
* In particular, the overflow and saturation behavior of the accumulator used must be considered.
* Refer to the function specific documentation below for usage guidelines.
*/
/**
*
* @ingroup clarke
* @brief Floating-point Clarke transform
* @param[in] Ia input three-phase coordinate <code>a</code>
* @param[in] Ib input three-phase coordinate <code>b</code>
* @param[out] pIalpha points to output two-phase orthogonal vector axis alpha
* @param[out] pIbeta points to output two-phase orthogonal vector axis beta
*/
__STATIC_FORCEINLINE void arm_clarke_f32(
float32_t Ia,
float32_t Ib,
float32_t * pIalpha,
float32_t * pIbeta)
{
/* Calculate pIalpha using the equation, pIalpha = Ia */
*pIalpha = Ia;
/* Calculate pIbeta using the equation, pIbeta = (1/sqrt(3)) * Ia + (2/sqrt(3)) * Ib */
*pIbeta = (0.57735026919f * Ia + 1.15470053838f * Ib);
}
/**
@ingroup clarke
@brief Clarke transform for Q31 version
@param[in] Ia input three-phase coordinate <code>a</code>
@param[in] Ib input three-phase coordinate <code>b</code>
@param[out] pIalpha points to output two-phase orthogonal vector axis alpha
@param[out] pIbeta points to output two-phase orthogonal vector axis beta
\par Scaling and Overflow Behavior
The function is implemented using an internal 32-bit accumulator.
The accumulator maintains 1.31 format by truncating lower 31 bits of the intermediate multiplication in 2.62 format.
There is saturation on the addition, hence there is no risk of overflow.
*/
__STATIC_FORCEINLINE void arm_clarke_q31(
q31_t Ia,
q31_t Ib,
q31_t * pIalpha,
q31_t * pIbeta)
{
q31_t product1, product2; /* Temporary variables used to store intermediate results */
/* Calculating pIalpha from Ia by equation pIalpha = Ia */
*pIalpha = Ia;
/* Intermediate product is calculated by (1/(sqrt(3)) * Ia) */
product1 = (q31_t) (((q63_t) Ia * 0x24F34E8B) >> 30);
/* Intermediate product is calculated by (2/sqrt(3) * Ib) */
product2 = (q31_t) (((q63_t) Ib * 0x49E69D16) >> 30);
/* pIbeta is calculated by adding the intermediate products */
*pIbeta = __QADD(product1, product2);
}
/**
* @ingroup groupController
*/
/**
* @defgroup inv_clarke Vector Inverse Clarke Transform
* Inverse Clarke transform converts the two-coordinate time invariant vector into instantaneous stator phases.
*
* The function operates on a single sample of data and each call to the function returns the processed output.
* The library provides separate functions for Q31 and floating-point data types.
* \par Algorithm
* \image html clarkeInvFormula.gif
* where <code>pIa</code> and <code>pIb</code> are the instantaneous stator phases and
* <code>Ialpha</code> and <code>Ibeta</code> are the two coordinates of time invariant vector.
* \par Fixed-Point Behavior
* Care must be taken when using the Q31 version of the Clarke transform.
* In particular, the overflow and saturation behavior of the accumulator used must be considered.
* Refer to the function specific documentation below for usage guidelines.
*/
/**
* @ingroup inv_clarke
* @brief Floating-point Inverse Clarke transform
* @param[in] Ialpha input two-phase orthogonal vector axis alpha
* @param[in] Ibeta input two-phase orthogonal vector axis beta
* @param[out] pIa points to output three-phase coordinate <code>a</code>
* @param[out] pIb points to output three-phase coordinate <code>b</code>
*/
__STATIC_FORCEINLINE void arm_inv_clarke_f32(
float32_t Ialpha,
float32_t Ibeta,
float32_t * pIa,
float32_t * pIb)
{
/* Calculating pIa from Ialpha by equation pIa = Ialpha */
*pIa = Ialpha;
/* Calculating pIb from Ialpha and Ibeta by equation pIb = -(1/2) * Ialpha + (sqrt(3)/2) * Ibeta */
*pIb = -0.5f * Ialpha + 0.8660254039f * Ibeta;
}
/**
@ingroup inv_clarke
@brief Inverse Clarke transform for Q31 version
@param[in] Ialpha input two-phase orthogonal vector axis alpha
@param[in] Ibeta input two-phase orthogonal vector axis beta
@param[out] pIa points to output three-phase coordinate <code>a</code>
@param[out] pIb points to output three-phase coordinate <code>b</code>
\par Scaling and Overflow Behavior
The function is implemented using an internal 32-bit accumulator.
The accumulator maintains 1.31 format by truncating lower 31 bits of the intermediate multiplication in 2.62 format.
There is saturation on the subtraction, hence there is no risk of overflow.
*/
__STATIC_FORCEINLINE void arm_inv_clarke_q31(
q31_t Ialpha,
q31_t Ibeta,
q31_t * pIa,
q31_t * pIb)
{
q31_t product1, product2; /* Temporary variables used to store intermediate results */
/* Calculating pIa from Ialpha by equation pIa = Ialpha */
*pIa = Ialpha;
/* Intermediate product is calculated by (1/(2*sqrt(3)) * Ia) */
product1 = (q31_t) (((q63_t) (Ialpha) * (0x40000000)) >> 31);
/* Intermediate product is calculated by (1/sqrt(3) * pIb) */
product2 = (q31_t) (((q63_t) (Ibeta) * (0x6ED9EBA1)) >> 31);
/* pIb is calculated by subtracting the products */
*pIb = __QSUB(product2, product1);
}
#ifdef __cplusplus
}
#endif
#endif /* ifndef _CONTROLLER_FUNCTIONS_H_ */

41
sdk/CMSIS/DSP/Include/dsp/controller_functions_f16.h Normal file
View file

@ -0,0 +1,41 @@
/******************************************************************************
* @file controller_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef CONTROLLER_FUNCTIONS_F16_H_
#define CONTROLLER_FUNCTIONS_F16_H_
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _CONTROLLER_FUNCTIONS_F16_H_ */

146
sdk/CMSIS/DSP/Include/dsp/debug.h Normal file
View file

@ -0,0 +1,146 @@
/******************************************************************************
* @file basic_math_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef DEBUG_FUNCTIONS_H_
#define DEBUG_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/matrix_functions.h"
#include "dsp/matrix_functions_f16.h"
#include <stdio.h>
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
#define PROW_f16(S,NB) \
{ \
printf("{%f",(double)(S)[0]); \
for(unsigned int i=1;i<(NB) ;i++) \
{ \
printf(",%f",(double)(S)[i]);\
} \
printf("}"); \
};
#define PV_f16(S,V,NB)\
{ \
printf("%s=",(S)); \
PROW_f16((V),(NB)); \
printf(";\n"); \
};
#define PM_f16(S,M) \
{ \
printf("%s={",(S)); \
for(unsigned int row=0;row<(M)->numRows;row++) \
{ \
if (row != 0) \
{ \
printf("\n,"); \
} \
PROW_f16((M)->pData + row * (M)->numCols, (M)->numCols);\
} \
printf("};\n"); \
}
#endif
#define PROW_f32(S,NB) \
{ \
printf("{%f",(double)(S)[0]); \
for(unsigned int i=1;i<(NB) ;i++) \
{ \
printf(",%f",(double)(S)[i]);\
} \
printf("}"); \
};
#define PV_f32(S,V,NB)\
{ \
printf("%s=",(S)); \
PROW_f32((V),(NB)); \
printf(";\n"); \
};
#define PM_f32(S,M) \
{ \
printf("%s={",(S)); \
for(unsigned int row=0;row<(M)->numRows;row++) \
{ \
if (row != 0) \
{ \
printf("\n,"); \
} \
PROW_f32((M)->pData + row * (M)->numCols, (M)->numCols);\
} \
printf("};\n"); \
}
#define PROW_f64(S,NB) \
{ \
printf("{%.20g",(double)(S)[0]); \
for(unsigned int i=1;i<(NB) ;i++) \
{ \
printf(",%.20g",(double)(S)[i]);\
} \
printf("}"); \
};
#define PV_f64(S,V,NB) \
{ \
printf("%s=",(S)); \
PROW_f64((V),(NB));\
printf(";\n"); \
};
#define PM_f64(S,M) \
{ \
printf("%s={",(S)); \
for(unsigned int row=0;row<(M)->numRows;row++) \
{ \
if (row != 0) \
{ \
printf("\n,"); \
} \
PROW_f64((M)->pData + row * (M)->numCols, (M)->numCols);\
} \
printf("};\n"); \
}
#ifdef __cplusplus
}
#endif
#endif /* ifndef _DEBUG_FUNCTIONS_H_ */

386
sdk/CMSIS/DSP/Include/dsp/distance_functions.h Normal file
View file

@ -0,0 +1,386 @@
/******************************************************************************
* @file distance_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef DISTANCE_FUNCTIONS_H_
#define DISTANCE_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/statistics_functions.h"
#include "dsp/basic_math_functions.h"
#include "dsp/fast_math_functions.h"
#include "dsp/matrix_functions.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupDistance Distance Functions
*
* Distance functions for use with clustering algorithms.
* There are distance functions for float vectors and boolean vectors.
*
*/
/* 6.14 bug */
#if defined (__ARMCC_VERSION) && (__ARMCC_VERSION >= 6100100) && (__ARMCC_VERSION < 6150001)
__attribute__((weak)) float __powisf2(float a, int b);
#endif
/**
* @brief Euclidean distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_euclidean_distance_f32(const float32_t *pA,const float32_t *pB, uint32_t blockSize);
/**
* @brief Euclidean distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float64_t arm_euclidean_distance_f64(const float64_t *pA,const float64_t *pB, uint32_t blockSize);
/**
* @brief Bray-Curtis distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_braycurtis_distance_f32(const float32_t *pA,const float32_t *pB, uint32_t blockSize);
/**
* @brief Canberra distance between two vectors
*
* This function may divide by zero when samples pA[i] and pB[i] are both zero.
* The result of the computation will be correct. So the division per zero may be
* ignored.
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_canberra_distance_f32(const float32_t *pA,const float32_t *pB, uint32_t blockSize);
/**
* @brief Chebyshev distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_chebyshev_distance_f32(const float32_t *pA,const float32_t *pB, uint32_t blockSize);
/**
* @brief Chebyshev distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float64_t arm_chebyshev_distance_f64(const float64_t *pA,const float64_t *pB, uint32_t blockSize);
/**
* @brief Cityblock (Manhattan) distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_cityblock_distance_f32(const float32_t *pA,const float32_t *pB, uint32_t blockSize);
/**
* @brief Cityblock (Manhattan) distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float64_t arm_cityblock_distance_f64(const float64_t *pA,const float64_t *pB, uint32_t blockSize);
/**
* @brief Correlation distance between two vectors
*
* The input vectors are modified in place !
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_correlation_distance_f32(float32_t *pA,float32_t *pB, uint32_t blockSize);
/**
* @brief Cosine distance between two vectors
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_cosine_distance_f32(const float32_t *pA,const float32_t *pB, uint32_t blockSize);
/**
* @brief Cosine distance between two vectors
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float64_t arm_cosine_distance_f64(const float64_t *pA,const float64_t *pB, uint32_t blockSize);
/**
* @brief Jensen-Shannon distance between two vectors
*
* This function is assuming that elements of second vector are > 0
* and 0 only when the corresponding element of first vector is 0.
* Otherwise the result of the computation does not make sense
* and for speed reasons, the cases returning NaN or Infinity are not
* managed.
*
* When the function is computing x log (x / y) with x 0 and y 0,
* it will compute the right value (0) but a division per zero will occur
* and shoudl be ignored in client code.
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_jensenshannon_distance_f32(const float32_t *pA,const float32_t *pB,uint32_t blockSize);
/**
* @brief Minkowski distance between two vectors
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] n Norm order (>= 2)
* @param[in] blockSize vector length
* @return distance
*
*/
float32_t arm_minkowski_distance_f32(const float32_t *pA,const float32_t *pB, int32_t order, uint32_t blockSize);
/**
* @brief Dice distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] order Distance order
* @param[in] blockSize Number of samples
* @return distance
*
*/
float32_t arm_dice_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
/**
* @brief Hamming distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] numberOfBools Number of booleans
* @return distance
*
*/
float32_t arm_hamming_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
/**
* @brief Jaccard distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] numberOfBools Number of booleans
* @return distance
*
*/
float32_t arm_jaccard_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
/**
* @brief Kulsinski distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] numberOfBools Number of booleans
* @return distance
*
*/
float32_t arm_kulsinski_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
/**
* @brief Roger Stanimoto distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] numberOfBools Number of booleans
* @return distance
*
*/
float32_t arm_rogerstanimoto_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
/**
* @brief Russell-Rao distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] numberOfBools Number of booleans
* @return distance
*
*/
float32_t arm_russellrao_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
/**
* @brief Sokal-Michener distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] numberOfBools Number of booleans
* @return distance
*
*/
float32_t arm_sokalmichener_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
/**
* @brief Sokal-Sneath distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] numberOfBools Number of booleans
* @return distance
*
*/
float32_t arm_sokalsneath_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
/**
* @brief Yule distance between two vectors
*
* @param[in] pA First vector of packed booleans
* @param[in] pB Second vector of packed booleans
* @param[in] numberOfBools Number of booleans
* @return distance
*
*/
float32_t arm_yule_distance(const uint32_t *pA, const uint32_t *pB, uint32_t numberOfBools);
typedef enum
{
ARM_DTW_SAKOE_CHIBA_WINDOW = 1,
/*ARM_DTW_ITAKURA_WINDOW = 2,*/
ARM_DTW_SLANTED_BAND_WINDOW = 3
} arm_dtw_window;
/**
* @brief Window for dynamic time warping computation
* @param[in] windowType Type of window
* @param[in] windowSize Window size
* @param[in,out] pWindow Window
* @return Error if window type not recognized
*
*/
arm_status arm_dtw_init_window_q7(const arm_dtw_window windowType,
const int32_t windowSize,
arm_matrix_instance_q7 *pWindow);
/**
* @brief Dynamic Time Warping distance
* @param[in] pDistance Distance matrix (Query rows * Template columns)
* @param[in] pWindow Windowing (can be NULL if no windowing used)
* @param[out] pDTW Temporary cost buffer (same size)
* @param[out] distance Distance
* @return Error in case no path can be found with window constraint
*
*/
arm_status arm_dtw_distance_f32(const arm_matrix_instance_f32 *pDistance,
const arm_matrix_instance_q7 *pWindow,
arm_matrix_instance_f32 *pDTW,
float32_t *distance);
/**
* @brief Mapping between query and template
* @param[in] pDTW Cost matrix (Query rows * Template columns)
* @param[out] pPath Warping path in cost matrix 2*(nb rows + nb columns)
* @param[out] pathLength Length of path in number of points
*
*/
void arm_dtw_path_f32(const arm_matrix_instance_f32 *pDTW,
int16_t *pPath,
uint32_t *pathLength);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _DISTANCE_FUNCTIONS_H_ */

170
sdk/CMSIS/DSP/Include/dsp/distance_functions_f16.h Normal file
View file

@ -0,0 +1,170 @@
/******************************************************************************
* @file distance_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef DISTANCE_FUNCTIONS_F16_H_
#define DISTANCE_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
/* 6.14 bug */
#if defined (__ARMCC_VERSION) && (__ARMCC_VERSION >= 6100100) && (__ARMCC_VERSION < 6150001)
/* Defined in minkowski_f32 */
__attribute__((weak)) float __powisf2(float a, int b);
#endif
#include "dsp/statistics_functions_f16.h"
#include "dsp/basic_math_functions_f16.h"
#include "dsp/fast_math_functions_f16.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Euclidean distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_euclidean_distance_f16(const float16_t *pA,const float16_t *pB, uint32_t blockSize);
/**
* @brief Bray-Curtis distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_braycurtis_distance_f16(const float16_t *pA,const float16_t *pB, uint32_t blockSize);
/**
* @brief Canberra distance between two vectors
*
* This function may divide by zero when samples pA[i] and pB[i] are both zero.
* The result of the computation will be correct. So the division per zero may be
* ignored.
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_canberra_distance_f16(const float16_t *pA,const float16_t *pB, uint32_t blockSize);
/**
* @brief Chebyshev distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_chebyshev_distance_f16(const float16_t *pA,const float16_t *pB, uint32_t blockSize);
/**
* @brief Cityblock (Manhattan) distance between two vectors
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_cityblock_distance_f16(const float16_t *pA,const float16_t *pB, uint32_t blockSize);
/**
* @brief Correlation distance between two vectors
*
* The input vectors are modified in place !
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_correlation_distance_f16(float16_t *pA,float16_t *pB, uint32_t blockSize);
/**
* @brief Cosine distance between two vectors
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_cosine_distance_f16(const float16_t *pA,const float16_t *pB, uint32_t blockSize);
/**
* @brief Jensen-Shannon distance between two vectors
*
* This function is assuming that elements of second vector are > 0
* and 0 only when the corresponding element of first vector is 0.
* Otherwise the result of the computation does not make sense
* and for speed reasons, the cases returning NaN or Infinity are not
* managed.
*
* When the function is computing x log (x / y) with x 0 and y 0,
* it will compute the right value (0) but a division per zero will occur
* and shoudl be ignored in client code.
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_jensenshannon_distance_f16(const float16_t *pA,const float16_t *pB,uint32_t blockSize);
/**
* @brief Minkowski distance between two vectors
*
* @param[in] pA First vector
* @param[in] pB Second vector
* @param[in] n Norm order (>= 2)
* @param[in] blockSize vector length
* @return distance
*/
float16_t arm_minkowski_distance_f16(const float16_t *pA,const float16_t *pB, int32_t order, uint32_t blockSize);
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _DISTANCE_FUNCTIONS_F16_H_ */

377
sdk/CMSIS/DSP/Include/dsp/fast_math_functions.h Normal file
View file

@ -0,0 +1,377 @@
/******************************************************************************
* @file fast_math_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef FAST_MATH_FUNCTIONS_H_
#define FAST_MATH_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/basic_math_functions.h"
#include <math.h>
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @brief Macros required for SINE and COSINE Fast math approximations
*/
#define FAST_MATH_TABLE_SIZE 512
#define FAST_MATH_Q31_SHIFT (32 - 10)
#define FAST_MATH_Q15_SHIFT (16 - 10)
#ifndef PI
#define PI 3.14159265358979f
#endif
#ifndef PI_F64
#define PI_F64 3.14159265358979323846
#endif
/**
* @defgroup groupFastMath Fast Math Functions
* This set of functions provides a fast approximation to sine, cosine, and square root.
* As compared to most of the other functions in the CMSIS math library, the fast math functions
* operate on individual values and not arrays.
* There are separate functions for Q15, Q31, and floating-point data.
*
*/
/**
* @brief Fast approximation to the trigonometric sine function for floating-point data.
* @param[in] x input value in radians.
* @return sin(x).
*/
float32_t arm_sin_f32(
float32_t x);
/**
* @brief Fast approximation to the trigonometric sine function for Q31 data.
* @param[in] x Scaled input value in radians.
* @return sin(x).
*/
q31_t arm_sin_q31(
q31_t x);
/**
* @brief Fast approximation to the trigonometric sine function for Q15 data.
* @param[in] x Scaled input value in radians.
* @return sin(x).
*/
q15_t arm_sin_q15(
q15_t x);
/**
* @brief Fast approximation to the trigonometric cosine function for floating-point data.
* @param[in] x input value in radians.
* @return cos(x).
*/
float32_t arm_cos_f32(
float32_t x);
/**
* @brief Fast approximation to the trigonometric cosine function for Q31 data.
* @param[in] x Scaled input value in radians.
* @return cos(x).
*/
q31_t arm_cos_q31(
q31_t x);
/**
* @brief Fast approximation to the trigonometric cosine function for Q15 data.
* @param[in] x Scaled input value in radians.
* @return cos(x).
*/
q15_t arm_cos_q15(
q15_t x);
/**
@brief Floating-point vector of log values.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
void arm_vlog_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
@brief Floating-point vector of log values.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
void arm_vlog_f64(
const float64_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief q31 vector of log values.
* @param[in] pSrc points to the input vector in q31
* @param[out] pDst points to the output vector in q5.26
* @param[in] blockSize number of samples in each vector
*/
void arm_vlog_q31(const q31_t * pSrc,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief q15 vector of log values.
* @param[in] pSrc points to the input vector in q15
* @param[out] pDst points to the output vector in q4.11
* @param[in] blockSize number of samples in each vector
*/
void arm_vlog_q15(const q15_t * pSrc,
q15_t * pDst,
uint32_t blockSize);
/**
@brief Floating-point vector of exp values.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
void arm_vexp_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
@brief Floating-point vector of exp values.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
void arm_vexp_f64(
const float64_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @defgroup SQRT Square Root
*
* Computes the square root of a number.
* There are separate functions for Q15, Q31, and floating-point data types.
* The square root function is computed using the Newton-Raphson algorithm.
* This is an iterative algorithm of the form:
* <pre>
* x1 = x0 - f(x0)/f'(x0)
* </pre>
* where <code>x1</code> is the current estimate,
* <code>x0</code> is the previous estimate, and
* <code>f'(x0)</code> is the derivative of <code>f()</code> evaluated at <code>x0</code>.
* For the square root function, the algorithm reduces to:
* <pre>
* x0 = in/2 [initial guess]
* x1 = 1/2 * ( x0 + in / x0) [each iteration]
* </pre>
*/
/**
* @addtogroup SQRT
* @{
*/
/**
@brief Floating-point square root function.
@param[in] in input value
@param[out] pOut square root of input value
@return execution status
- \ref ARM_MATH_SUCCESS : input value is positive
- \ref ARM_MATH_ARGUMENT_ERROR : input value is negative; *pOut is set to 0
*/
__STATIC_FORCEINLINE arm_status arm_sqrt_f32(
const float32_t in,
float32_t * pOut)
{
if (in >= 0.0f)
{
#if defined ( __CC_ARM )
#if defined __TARGET_FPU_VFP
*pOut = __sqrtf(in);
#else
*pOut = sqrtf(in);
#endif
#elif defined ( __ICCARM__ )
#if defined __ARMVFP__
__ASM("VSQRT.F32 %0,%1" : "=t"(*pOut) : "t"(in));
#else
*pOut = sqrtf(in);
#endif
#elif defined ( __ARMCC_VERSION ) && ( __ARMCC_VERSION >= 6010050 )
*pOut = _sqrtf(in);
#elif defined(__GNUC_PYTHON__)
*pOut = sqrtf(in);
#elif defined ( __GNUC__ )
#if defined (__VFP_FP__) && !defined(__SOFTFP__)
__ASM("VSQRT.F32 %0,%1" : "=t"(*pOut) : "t"(in));
#else
*pOut = sqrtf(in);
#endif
#else
*pOut = sqrtf(in);
#endif
return (ARM_MATH_SUCCESS);
}
else
{
*pOut = 0.0f;
return (ARM_MATH_ARGUMENT_ERROR);
}
}
/**
@brief Q31 square root function.
@param[in] in input value. The range of the input value is [0 +1) or 0x00000000 to 0x7FFFFFFF
@param[out] pOut points to square root of input value
@return execution status
- \ref ARM_MATH_SUCCESS : input value is positive
- \ref ARM_MATH_ARGUMENT_ERROR : input value is negative; *pOut is set to 0
*/
arm_status arm_sqrt_q31(
q31_t in,
q31_t * pOut);
/**
@brief Q15 square root function.
@param[in] in input value. The range of the input value is [0 +1) or 0x0000 to 0x7FFF
@param[out] pOut points to square root of input value
@return execution status
- \ref ARM_MATH_SUCCESS : input value is positive
- \ref ARM_MATH_ARGUMENT_ERROR : input value is negative; *pOut is set to 0
*/
arm_status arm_sqrt_q15(
q15_t in,
q15_t * pOut);
/**
* @} end of SQRT group
*/
/**
@brief Fixed point division
@param[in] numerator Numerator
@param[in] denominator Denominator
@param[out] quotient Quotient value normalized between -1.0 and 1.0
@param[out] shift Shift left value to get the unnormalized quotient
@return error status
When dividing by 0, an error ARM_MATH_NANINF is returned. And the quotient is forced
to the saturated negative or positive value.
*/
arm_status arm_divide_q15(q15_t numerator,
q15_t denominator,
q15_t *quotient,
int16_t *shift);
/**
@brief Fixed point division
@param[in] numerator Numerator
@param[in] denominator Denominator
@param[out] quotient Quotient value normalized between -1.0 and 1.0
@param[out] shift Shift left value to get the unnormalized quotient
@return error status
When dividing by 0, an error ARM_MATH_NANINF is returned. And the quotient is forced
to the saturated negative or positive value.
*/
arm_status arm_divide_q31(q31_t numerator,
q31_t denominator,
q31_t *quotient,
int16_t *shift);
/**
@brief Arc tangent in radian of y/x using sign of x and y to determine right quadrant.
@param[in] y y coordinate
@param[in] x x coordinate
@param[out] result Result
@return error status.
*/
arm_status arm_atan2_f32(float32_t y,float32_t x,float32_t *result);
/**
@brief Arc tangent in radian of y/x using sign of x and y to determine right quadrant.
@param[in] y y coordinate
@param[in] x x coordinate
@param[out] result Result in Q2.29
@return error status.
*/
arm_status arm_atan2_q31(q31_t y,q31_t x,q31_t *result);
/**
@brief Arc tangent in radian of y/x using sign of x and y to determine right quadrant.
@param[in] y y coordinate
@param[in] x x coordinate
@param[out] result Result in Q2.13
@return error status.
*/
arm_status arm_atan2_q15(q15_t y,q15_t x,q15_t *result);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _FAST_MATH_FUNCTIONS_H_ */

122
sdk/CMSIS/DSP/Include/dsp/fast_math_functions_f16.h Normal file
View file

@ -0,0 +1,122 @@
/******************************************************************************
* @file fast_math_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef FAST_MATH_FUNCTIONS_F16_H_
#define FAST_MATH_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
/* For sqrt_f32 */
#include "dsp/fast_math_functions.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @addtogroup SQRT
* @{
*/
/**
@brief Floating-point square root function.
@param[in] in input value
@param[out] pOut square root of input value
@return execution status
- \ref ARM_MATH_SUCCESS : input value is positive
- \ref ARM_MATH_ARGUMENT_ERROR : input value is negative; *pOut is set to 0
*/
__STATIC_FORCEINLINE arm_status arm_sqrt_f16(
float16_t in,
float16_t * pOut)
{
float32_t r;
arm_status status;
status=arm_sqrt_f32((float32_t)in,&r);
*pOut=(float16_t)r;
return(status);
}
/**
@} end of SQRT group
*/
/**
@brief Floating-point vector of log values.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
void arm_vlog_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
/**
@brief Floating-point vector of exp values.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
void arm_vexp_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
/**
@brief Floating-point vector of inverse values.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
void arm_vinverse_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
/**
@brief Arc tangent in radian of y/x using sign of x and y to determine right quadrant.
@param[in] y y coordinate
@param[in] x x coordinate
@param[out] result Result
@return error status.
*/
arm_status arm_atan2_f16(float16_t y,float16_t x,float16_t *result);
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _FAST_MATH_FUNCTIONS_F16_H_ */

2570
sdk/CMSIS/DSP/Include/dsp/filtering_functions.h Normal file
View file

File diff suppressed because it is too large Load diff

236
sdk/CMSIS/DSP/Include/dsp/filtering_functions_f16.h Normal file
View file

@ -0,0 +1,236 @@
/******************************************************************************
* @file filtering_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef FILTERING_FUNCTIONS_F16_H_
#define FILTERING_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Instance structure for the floating-point FIR filter.
*/
typedef struct
{
uint16_t numTaps; /**< number of filter coefficients in the filter. */
float16_t *pState; /**< points to the state variable array. The array is of length numTaps+blockSize-1. */
const float16_t *pCoeffs; /**< points to the coefficient array. The array is of length numTaps. */
} arm_fir_instance_f16;
/**
* @brief Initialization function for the floating-point FIR filter.
* @param[in,out] S points to an instance of the floating-point FIR filter structure.
* @param[in] numTaps Number of filter coefficients in the filter.
* @param[in] pCoeffs points to the filter coefficients.
* @param[in] pState points to the state buffer.
* @param[in] blockSize number of samples that are processed at a time.
*/
void arm_fir_init_f16(
arm_fir_instance_f16 * S,
uint16_t numTaps,
const float16_t * pCoeffs,
float16_t * pState,
uint32_t blockSize);
/**
* @brief Processing function for the floating-point FIR filter.
* @param[in] S points to an instance of the floating-point FIR structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data.
* @param[in] blockSize number of samples to process.
*/
void arm_fir_f16(
const arm_fir_instance_f16 * S,
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Instance structure for the floating-point Biquad cascade filter.
*/
typedef struct
{
uint32_t numStages; /**< number of 2nd order stages in the filter. Overall order is 2*numStages. */
float16_t *pState; /**< Points to the array of state coefficients. The array is of length 4*numStages. */
const float16_t *pCoeffs; /**< Points to the array of coefficients. The array is of length 5*numStages. */
} arm_biquad_casd_df1_inst_f16;
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
/**
* @brief Instance structure for the modified Biquad coefs required by vectorized code.
*/
typedef struct
{
float16_t coeffs[12][8]; /**< Points to the array of modified coefficients. The array is of length 32. There is one per stage */
} arm_biquad_mod_coef_f16;
#endif
/**
* @brief Processing function for the floating-point Biquad cascade filter.
* @param[in] S points to an instance of the floating-point Biquad cascade structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data.
* @param[in] blockSize number of samples to process.
*/
void arm_biquad_cascade_df1_f16(
const arm_biquad_casd_df1_inst_f16 * S,
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
void arm_biquad_cascade_df1_mve_init_f16(
arm_biquad_casd_df1_inst_f16 * S,
uint8_t numStages,
const float16_t * pCoeffs,
arm_biquad_mod_coef_f16 * pCoeffsMod,
float16_t * pState);
#endif
void arm_biquad_cascade_df1_init_f16(
arm_biquad_casd_df1_inst_f16 * S,
uint8_t numStages,
const float16_t * pCoeffs,
float16_t * pState);
/**
* @brief Instance structure for the floating-point transposed direct form II Biquad cascade filter.
*/
typedef struct
{
uint8_t numStages; /**< number of 2nd order stages in the filter. Overall order is 2*numStages. */
float16_t *pState; /**< points to the array of state coefficients. The array is of length 2*numStages. */
const float16_t *pCoeffs; /**< points to the array of coefficients. The array is of length 5*numStages. */
} arm_biquad_cascade_df2T_instance_f16;
/**
* @brief Instance structure for the floating-point transposed direct form II Biquad cascade filter.
*/
typedef struct
{
uint8_t numStages; /**< number of 2nd order stages in the filter. Overall order is 2*numStages. */
float16_t *pState; /**< points to the array of state coefficients. The array is of length 4*numStages. */
const float16_t *pCoeffs; /**< points to the array of coefficients. The array is of length 5*numStages. */
} arm_biquad_cascade_stereo_df2T_instance_f16;
/**
* @brief Processing function for the floating-point transposed direct form II Biquad cascade filter.
* @param[in] S points to an instance of the filter data structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data
* @param[in] blockSize number of samples to process.
*/
void arm_biquad_cascade_df2T_f16(
const arm_biquad_cascade_df2T_instance_f16 * S,
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Processing function for the floating-point transposed direct form II Biquad cascade filter. 2 channels
* @param[in] S points to an instance of the filter data structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data
* @param[in] blockSize number of samples to process.
*/
void arm_biquad_cascade_stereo_df2T_f16(
const arm_biquad_cascade_stereo_df2T_instance_f16 * S,
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize);
/**
* @brief Initialization function for the floating-point transposed direct form II Biquad cascade filter.
* @param[in,out] S points to an instance of the filter data structure.
* @param[in] numStages number of 2nd order stages in the filter.
* @param[in] pCoeffs points to the filter coefficients.
* @param[in] pState points to the state buffer.
*/
void arm_biquad_cascade_df2T_init_f16(
arm_biquad_cascade_df2T_instance_f16 * S,
uint8_t numStages,
const float16_t * pCoeffs,
float16_t * pState);
/**
* @brief Initialization function for the floating-point transposed direct form II Biquad cascade filter.
* @param[in,out] S points to an instance of the filter data structure.
* @param[in] numStages number of 2nd order stages in the filter.
* @param[in] pCoeffs points to the filter coefficients.
* @param[in] pState points to the state buffer.
*/
void arm_biquad_cascade_stereo_df2T_init_f16(
arm_biquad_cascade_stereo_df2T_instance_f16 * S,
uint8_t numStages,
const float16_t * pCoeffs,
float16_t * pState);
/**
* @brief Correlation of floating-point sequences.
* @param[in] pSrcA points to the first input sequence.
* @param[in] srcALen length of the first input sequence.
* @param[in] pSrcB points to the second input sequence.
* @param[in] srcBLen length of the second input sequence.
* @param[out] pDst points to the block of output data Length 2 * max(srcALen, srcBLen) - 1.
*/
void arm_correlate_f16(
const float16_t * pSrcA,
uint32_t srcALen,
const float16_t * pSrcB,
uint32_t srcBLen,
float16_t * pDst);
/**
@brief Levinson Durbin
@param[in] phi autocovariance vector starting with lag 0 (length is nbCoefs + 1)
@param[out] a autoregressive coefficients
@param[out] err prediction error (variance)
@param[in] nbCoefs number of autoregressive coefficients
*/
void arm_levinson_durbin_f16(const float16_t *phi,
float16_t *a,
float16_t *err,
int nbCoefs);
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _FILTERING_FUNCTIONS_F16_H_ */

275
sdk/CMSIS/DSP/Include/dsp/interpolation_functions.h Normal file
View file

@ -0,0 +1,275 @@
/******************************************************************************
* @file interpolation_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef INTERPOLATION_FUNCTIONS_H_
#define INTERPOLATION_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupInterpolation Interpolation Functions
* These functions perform 1- and 2-dimensional interpolation of data.
* Linear interpolation is used for 1-dimensional data and
* bilinear interpolation is used for 2-dimensional data.
*/
/**
* @brief Instance structure for the floating-point Linear Interpolate function.
*/
typedef struct
{
uint32_t nValues; /**< nValues */
float32_t x1; /**< x1 */
float32_t xSpacing; /**< xSpacing */
const float32_t *pYData; /**< pointer to the table of Y values */
} arm_linear_interp_instance_f32;
/**
* @brief Instance structure for the floating-point bilinear interpolation function.
*/
typedef struct
{
uint16_t numRows; /**< number of rows in the data table. */
uint16_t numCols; /**< number of columns in the data table. */
const float32_t *pData; /**< points to the data table. */
} arm_bilinear_interp_instance_f32;
/**
* @brief Instance structure for the Q31 bilinear interpolation function.
*/
typedef struct
{
uint16_t numRows; /**< number of rows in the data table. */
uint16_t numCols; /**< number of columns in the data table. */
const q31_t *pData; /**< points to the data table. */
} arm_bilinear_interp_instance_q31;
/**
* @brief Instance structure for the Q15 bilinear interpolation function.
*/
typedef struct
{
uint16_t numRows; /**< number of rows in the data table. */
uint16_t numCols; /**< number of columns in the data table. */
const q15_t *pData; /**< points to the data table. */
} arm_bilinear_interp_instance_q15;
/**
* @brief Instance structure for the Q15 bilinear interpolation function.
*/
typedef struct
{
uint16_t numRows; /**< number of rows in the data table. */
uint16_t numCols; /**< number of columns in the data table. */
const q7_t *pData; /**< points to the data table. */
} arm_bilinear_interp_instance_q7;
/**
* @brief Struct for specifying cubic spline type
*/
typedef enum
{
ARM_SPLINE_NATURAL = 0, /**< Natural spline */
ARM_SPLINE_PARABOLIC_RUNOUT = 1 /**< Parabolic runout spline */
} arm_spline_type;
/**
* @brief Instance structure for the floating-point cubic spline interpolation.
*/
typedef struct
{
arm_spline_type type; /**< Type (boundary conditions) */
const float32_t * x; /**< x values */
const float32_t * y; /**< y values */
uint32_t n_x; /**< Number of known data points */
float32_t * coeffs; /**< Coefficients buffer (b,c, and d) */
} arm_spline_instance_f32;
/**
* @brief Processing function for the floating-point cubic spline interpolation.
* @param[in] S points to an instance of the floating-point spline structure.
* @param[in] xq points to the x values ot the interpolated data points.
* @param[out] pDst points to the block of output data.
* @param[in] blockSize number of samples of output data.
*/
void arm_spline_f32(
arm_spline_instance_f32 * S,
const float32_t * xq,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Initialization function for the floating-point cubic spline interpolation.
* @param[in,out] S points to an instance of the floating-point spline structure.
* @param[in] type type of cubic spline interpolation (boundary conditions)
* @param[in] x points to the x values of the known data points.
* @param[in] y points to the y values of the known data points.
* @param[in] n number of known data points.
* @param[in] coeffs coefficients array for b, c, and d
* @param[in] tempBuffer buffer array for internal computations
*/
void arm_spline_init_f32(
arm_spline_instance_f32 * S,
arm_spline_type type,
const float32_t * x,
const float32_t * y,
uint32_t n,
float32_t * coeffs,
float32_t * tempBuffer);
/**
* @brief Process function for the floating-point Linear Interpolation Function.
* @param[in,out] S is an instance of the floating-point Linear Interpolation structure
* @param[in] x input sample to process
* @return y processed output sample.
*
*/
float32_t arm_linear_interp_f32(
const arm_linear_interp_instance_f32 * S,
float32_t x);
/**
*
* @brief Process function for the Q31 Linear Interpolation Function.
* @param[in] pYData pointer to Q31 Linear Interpolation table
* @param[in] x input sample to process
* @param[in] nValues number of table values
* @return y processed output sample.
*
* \par
* Input sample <code>x</code> is in 12.20 format which contains 12 bits for table index and 20 bits for fractional part.
* This function can support maximum of table size 2^12.
*
*/
q31_t arm_linear_interp_q31(
const q31_t * pYData,
q31_t x,
uint32_t nValues);
/**
*
* @brief Process function for the Q15 Linear Interpolation Function.
* @param[in] pYData pointer to Q15 Linear Interpolation table
* @param[in] x input sample to process
* @param[in] nValues number of table values
* @return y processed output sample.
*
* \par
* Input sample <code>x</code> is in 12.20 format which contains 12 bits for table index and 20 bits for fractional part.
* This function can support maximum of table size 2^12.
*
*/
q15_t arm_linear_interp_q15(
const q15_t * pYData,
q31_t x,
uint32_t nValues);
/**
*
* @brief Process function for the Q7 Linear Interpolation Function.
* @param[in] pYData pointer to Q7 Linear Interpolation table
* @param[in] x input sample to process
* @param[in] nValues number of table values
* @return y processed output sample.
*
* \par
* Input sample <code>x</code> is in 12.20 format which contains 12 bits for table index and 20 bits for fractional part.
* This function can support maximum of table size 2^12.
*/
q7_t arm_linear_interp_q7(
const q7_t * pYData,
q31_t x,
uint32_t nValues);
/**
* @brief Floating-point bilinear interpolation.
* @param[in,out] S points to an instance of the interpolation structure.
* @param[in] X interpolation coordinate.
* @param[in] Y interpolation coordinate.
* @return out interpolated value.
*/
float32_t arm_bilinear_interp_f32(
const arm_bilinear_interp_instance_f32 * S,
float32_t X,
float32_t Y);
/**
* @brief Q31 bilinear interpolation.
* @param[in,out] S points to an instance of the interpolation structure.
* @param[in] X interpolation coordinate in 12.20 format.
* @param[in] Y interpolation coordinate in 12.20 format.
* @return out interpolated value.
*/
q31_t arm_bilinear_interp_q31(
arm_bilinear_interp_instance_q31 * S,
q31_t X,
q31_t Y);
/**
* @brief Q15 bilinear interpolation.
* @param[in,out] S points to an instance of the interpolation structure.
* @param[in] X interpolation coordinate in 12.20 format.
* @param[in] Y interpolation coordinate in 12.20 format.
* @return out interpolated value.
*/
q15_t arm_bilinear_interp_q15(
arm_bilinear_interp_instance_q15 * S,
q31_t X,
q31_t Y);
/**
* @brief Q7 bilinear interpolation.
* @param[in,out] S points to an instance of the interpolation structure.
* @param[in] X interpolation coordinate in 12.20 format.
* @param[in] Y interpolation coordinate in 12.20 format.
* @return out interpolated value.
*/
q7_t arm_bilinear_interp_q7(
arm_bilinear_interp_instance_q7 * S,
q31_t X,
q31_t Y);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _INTERPOLATION_FUNCTIONS_H_ */

109
sdk/CMSIS/DSP/Include/dsp/interpolation_functions_f16.h Normal file
View file

@ -0,0 +1,109 @@
/******************************************************************************
* @file interpolation_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef INTERPOLATION_FUNCTIONS_F16_H_
#define INTERPOLATION_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Instance structure for the half floating-point Linear Interpolate function.
*/
typedef struct
{
uint32_t nValues; /**< nValues */
float16_t x1; /**< x1 */
float16_t xSpacing; /**< xSpacing */
const float16_t *pYData; /**< pointer to the table of Y values */
} arm_linear_interp_instance_f16;
/**
* @brief Instance structure for the floating-point bilinear interpolation function.
*/
typedef struct
{
uint16_t numRows;/**< number of rows in the data table. */
uint16_t numCols;/**< number of columns in the data table. */
const float16_t *pData; /**< points to the data table. */
} arm_bilinear_interp_instance_f16;
/**
* @addtogroup LinearInterpolate
* @{
*/
/**
* @brief Process function for the floating-point Linear Interpolation Function.
* @param[in,out] S is an instance of the floating-point Linear Interpolation structure
* @param[in] x input sample to process
* @return y processed output sample.
*/
float16_t arm_linear_interp_f16(
const arm_linear_interp_instance_f16 * S,
float16_t x);
/**
* @} end of LinearInterpolate group
*/
/**
* @addtogroup BilinearInterpolate
* @{
*/
/**
* @brief Floating-point bilinear interpolation.
* @param[in,out] S points to an instance of the interpolation structure.
* @param[in] X interpolation coordinate.
* @param[in] Y interpolation coordinate.
* @return out interpolated value.
*/
float16_t arm_bilinear_interp_f16(
const arm_bilinear_interp_instance_f16 * S,
float16_t X,
float16_t Y);
/**
* @} end of BilinearInterpolate group
*/
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _INTERPOLATION_FUNCTIONS_F16_H_ */

856
sdk/CMSIS/DSP/Include/dsp/matrix_functions.h Normal file
View file

@ -0,0 +1,856 @@
/******************************************************************************
* @file matrix_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.1
* @date 10 August 2022
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef MATRIX_FUNCTIONS_H_
#define MATRIX_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupMatrix Matrix Functions
*
* This set of functions provides basic matrix math operations.
* The functions operate on matrix data structures. For example,
* the type
* definition for the floating-point matrix structure is shown
* below:
* <pre>
* typedef struct
* {
* uint16_t numRows; // number of rows of the matrix.
* uint16_t numCols; // number of columns of the matrix.
* float32_t *pData; // points to the data of the matrix.
* } arm_matrix_instance_f32;
* </pre>
* There are similar definitions for Q15 and Q31 data types.
*
* The structure specifies the size of the matrix and then points to
* an array of data. The array is of size <code>numRows X numCols</code>
* and the values are arranged in row order. That is, the
* matrix element (i, j) is stored at:
* <pre>
* pData[i*numCols + j]
* </pre>
*
* \par Init Functions
* There is an associated initialization function for each type of matrix
* data structure.
* The initialization function sets the values of the internal structure fields.
* Refer to \ref arm_mat_init_f32(), \ref arm_mat_init_q31() and \ref arm_mat_init_q15()
* for floating-point, Q31 and Q15 types, respectively.
*
* \par
* Use of the initialization function is optional. However, if initialization function is used
* then the instance structure cannot be placed into a const data section.
* To place the instance structure in a const data
* section, manually initialize the data structure. For example:
* <pre>
* <code>arm_matrix_instance_f32 S = {nRows, nColumns, pData};</code>
* <code>arm_matrix_instance_q31 S = {nRows, nColumns, pData};</code>
* <code>arm_matrix_instance_q15 S = {nRows, nColumns, pData};</code>
* </pre>
* where <code>nRows</code> specifies the number of rows, <code>nColumns</code>
* specifies the number of columns, and <code>pData</code> points to the
* data array.
*
* \par Size Checking
* By default all of the matrix functions perform size checking on the input and
* output matrices. For example, the matrix addition function verifies that the
* two input matrices and the output matrix all have the same number of rows and
* columns. If the size check fails the functions return:
* <pre>
* ARM_MATH_SIZE_MISMATCH
* </pre>
* Otherwise the functions return
* <pre>
* ARM_MATH_SUCCESS
* </pre>
* There is some overhead associated with this matrix size checking.
* The matrix size checking is enabled via the \#define
* <pre>
* ARM_MATH_MATRIX_CHECK
* </pre>
* within the library project settings. By default this macro is defined
* and size checking is enabled. By changing the project settings and
* undefining this macro size checking is eliminated and the functions
* run a bit faster. With size checking disabled the functions always
* return <code>ARM_MATH_SUCCESS</code>.
*/
#define DEFAULT_HOUSEHOLDER_THRESHOLD_F64 (1.0e-16)
#define DEFAULT_HOUSEHOLDER_THRESHOLD_F32 (1.0e-12f)
/**
* @brief Instance structure for the floating-point matrix structure.
*/
typedef struct
{
uint16_t numRows; /**< number of rows of the matrix. */
uint16_t numCols; /**< number of columns of the matrix. */
float32_t *pData; /**< points to the data of the matrix. */
} arm_matrix_instance_f32;
/**
* @brief Instance structure for the floating-point matrix structure.
*/
typedef struct
{
uint16_t numRows; /**< number of rows of the matrix. */
uint16_t numCols; /**< number of columns of the matrix. */
float64_t *pData; /**< points to the data of the matrix. */
} arm_matrix_instance_f64;
/**
* @brief Instance structure for the Q7 matrix structure.
*/
typedef struct
{
uint16_t numRows; /**< number of rows of the matrix. */
uint16_t numCols; /**< number of columns of the matrix. */
q7_t *pData; /**< points to the data of the matrix. */
} arm_matrix_instance_q7;
/**
* @brief Instance structure for the Q15 matrix structure.
*/
typedef struct
{
uint16_t numRows; /**< number of rows of the matrix. */
uint16_t numCols; /**< number of columns of the matrix. */
q15_t *pData; /**< points to the data of the matrix. */
} arm_matrix_instance_q15;
/**
* @brief Instance structure for the Q31 matrix structure.
*/
typedef struct
{
uint16_t numRows; /**< number of rows of the matrix. */
uint16_t numCols; /**< number of columns of the matrix. */
q31_t *pData; /**< points to the data of the matrix. */
} arm_matrix_instance_q31;
/**
* @brief Floating-point matrix addition.
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_add_f32(
const arm_matrix_instance_f32 * pSrcA,
const arm_matrix_instance_f32 * pSrcB,
arm_matrix_instance_f32 * pDst);
/**
* @brief Q15 matrix addition.
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_add_q15(
const arm_matrix_instance_q15 * pSrcA,
const arm_matrix_instance_q15 * pSrcB,
arm_matrix_instance_q15 * pDst);
/**
* @brief Q31 matrix addition.
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_add_q31(
const arm_matrix_instance_q31 * pSrcA,
const arm_matrix_instance_q31 * pSrcB,
arm_matrix_instance_q31 * pDst);
/**
* @brief Floating-point, complex, matrix multiplication.
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_cmplx_mult_f32(
const arm_matrix_instance_f32 * pSrcA,
const arm_matrix_instance_f32 * pSrcB,
arm_matrix_instance_f32 * pDst);
/**
* @brief Q15, complex, matrix multiplication.
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_cmplx_mult_q15(
const arm_matrix_instance_q15 * pSrcA,
const arm_matrix_instance_q15 * pSrcB,
arm_matrix_instance_q15 * pDst,
q15_t * pScratch);
/**
* @brief Q31, complex, matrix multiplication.
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_cmplx_mult_q31(
const arm_matrix_instance_q31 * pSrcA,
const arm_matrix_instance_q31 * pSrcB,
arm_matrix_instance_q31 * pDst);
/**
* @brief Floating-point matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_trans_f32(
const arm_matrix_instance_f32 * pSrc,
arm_matrix_instance_f32 * pDst);
/**
* @brief Floating-point matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_trans_f64(
const arm_matrix_instance_f64 * pSrc,
arm_matrix_instance_f64 * pDst);
/**
* @brief Floating-point complex matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_cmplx_trans_f32(
const arm_matrix_instance_f32 * pSrc,
arm_matrix_instance_f32 * pDst);
/**
* @brief Q15 matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_trans_q15(
const arm_matrix_instance_q15 * pSrc,
arm_matrix_instance_q15 * pDst);
/**
* @brief Q15 complex matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_cmplx_trans_q15(
const arm_matrix_instance_q15 * pSrc,
arm_matrix_instance_q15 * pDst);
/**
* @brief Q7 matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_trans_q7(
const arm_matrix_instance_q7 * pSrc,
arm_matrix_instance_q7 * pDst);
/**
* @brief Q31 matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_trans_q31(
const arm_matrix_instance_q31 * pSrc,
arm_matrix_instance_q31 * pDst);
/**
* @brief Q31 complex matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_cmplx_trans_q31(
const arm_matrix_instance_q31 * pSrc,
arm_matrix_instance_q31 * pDst);
/**
* @brief Floating-point matrix multiplication
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_f32(
const arm_matrix_instance_f32 * pSrcA,
const arm_matrix_instance_f32 * pSrcB,
arm_matrix_instance_f32 * pDst);
/**
* @brief Floating-point matrix multiplication
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_f64(
const arm_matrix_instance_f64 * pSrcA,
const arm_matrix_instance_f64 * pSrcB,
arm_matrix_instance_f64 * pDst);
/**
* @brief Floating-point matrix and vector multiplication
* @param[in] pSrcMat points to the input matrix structure
* @param[in] pVec points to vector
* @param[out] pDst points to output vector
*/
void arm_mat_vec_mult_f32(
const arm_matrix_instance_f32 *pSrcMat,
const float32_t *pVec,
float32_t *pDst);
/**
* @brief Q7 matrix multiplication
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @param[in] pState points to the array for storing intermediate results
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_q7(
const arm_matrix_instance_q7 * pSrcA,
const arm_matrix_instance_q7 * pSrcB,
arm_matrix_instance_q7 * pDst,
q7_t * pState);
/**
* @brief Q7 matrix and vector multiplication
* @param[in] pSrcMat points to the input matrix structure
* @param[in] pVec points to vector
* @param[out] pDst points to output vector
*/
void arm_mat_vec_mult_q7(
const arm_matrix_instance_q7 *pSrcMat,
const q7_t *pVec,
q7_t *pDst);
/**
* @brief Q15 matrix multiplication
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @param[in] pState points to the array for storing intermediate results
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_q15(
const arm_matrix_instance_q15 * pSrcA,
const arm_matrix_instance_q15 * pSrcB,
arm_matrix_instance_q15 * pDst,
q15_t * pState);
/**
* @brief Q15 matrix and vector multiplication
* @param[in] pSrcMat points to the input matrix structure
* @param[in] pVec points to vector
* @param[out] pDst points to output vector
*/
void arm_mat_vec_mult_q15(
const arm_matrix_instance_q15 *pSrcMat,
const q15_t *pVec,
q15_t *pDst);
/**
* @brief Q15 matrix multiplication (fast variant) for Cortex-M3 and Cortex-M4
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @param[in] pState points to the array for storing intermediate results
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_fast_q15(
const arm_matrix_instance_q15 * pSrcA,
const arm_matrix_instance_q15 * pSrcB,
arm_matrix_instance_q15 * pDst,
q15_t * pState);
/**
* @brief Q31 matrix multiplication
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_q31(
const arm_matrix_instance_q31 * pSrcA,
const arm_matrix_instance_q31 * pSrcB,
arm_matrix_instance_q31 * pDst);
/**
* @brief Q31 matrix multiplication
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @param[in] pState points to the array for storing intermediate results
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_opt_q31(
const arm_matrix_instance_q31 * pSrcA,
const arm_matrix_instance_q31 * pSrcB,
arm_matrix_instance_q31 * pDst,
q31_t *pState);
/**
* @brief Q31 matrix and vector multiplication
* @param[in] pSrcMat points to the input matrix structure
* @param[in] pVec points to vector
* @param[out] pDst points to output vector
*/
void arm_mat_vec_mult_q31(
const arm_matrix_instance_q31 *pSrcMat,
const q31_t *pVec,
q31_t *pDst);
/**
* @brief Q31 matrix multiplication (fast variant) for Cortex-M3 and Cortex-M4
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_fast_q31(
const arm_matrix_instance_q31 * pSrcA,
const arm_matrix_instance_q31 * pSrcB,
arm_matrix_instance_q31 * pDst);
/**
* @brief Floating-point matrix subtraction
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_sub_f32(
const arm_matrix_instance_f32 * pSrcA,
const arm_matrix_instance_f32 * pSrcB,
arm_matrix_instance_f32 * pDst);
/**
* @brief Floating-point matrix subtraction
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_sub_f64(
const arm_matrix_instance_f64 * pSrcA,
const arm_matrix_instance_f64 * pSrcB,
arm_matrix_instance_f64 * pDst);
/**
* @brief Q15 matrix subtraction
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_sub_q15(
const arm_matrix_instance_q15 * pSrcA,
const arm_matrix_instance_q15 * pSrcB,
arm_matrix_instance_q15 * pDst);
/**
* @brief Q31 matrix subtraction
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_sub_q31(
const arm_matrix_instance_q31 * pSrcA,
const arm_matrix_instance_q31 * pSrcB,
arm_matrix_instance_q31 * pDst);
/**
* @brief Floating-point matrix scaling.
* @param[in] pSrc points to the input matrix
* @param[in] scale scale factor
* @param[out] pDst points to the output matrix
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_scale_f32(
const arm_matrix_instance_f32 * pSrc,
float32_t scale,
arm_matrix_instance_f32 * pDst);
/**
* @brief Q15 matrix scaling.
* @param[in] pSrc points to input matrix
* @param[in] scaleFract fractional portion of the scale factor
* @param[in] shift number of bits to shift the result by
* @param[out] pDst points to output matrix
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_scale_q15(
const arm_matrix_instance_q15 * pSrc,
q15_t scaleFract,
int32_t shift,
arm_matrix_instance_q15 * pDst);
/**
* @brief Q31 matrix scaling.
* @param[in] pSrc points to input matrix
* @param[in] scaleFract fractional portion of the scale factor
* @param[in] shift number of bits to shift the result by
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_scale_q31(
const arm_matrix_instance_q31 * pSrc,
q31_t scaleFract,
int32_t shift,
arm_matrix_instance_q31 * pDst);
/**
* @brief Q31 matrix initialization.
* @param[in,out] S points to an instance of the floating-point matrix structure.
* @param[in] nRows number of rows in the matrix.
* @param[in] nColumns number of columns in the matrix.
* @param[in] pData points to the matrix data array.
*/
void arm_mat_init_q31(
arm_matrix_instance_q31 * S,
uint16_t nRows,
uint16_t nColumns,
q31_t * pData);
/**
* @brief Q15 matrix initialization.
* @param[in,out] S points to an instance of the floating-point matrix structure.
* @param[in] nRows number of rows in the matrix.
* @param[in] nColumns number of columns in the matrix.
* @param[in] pData points to the matrix data array.
*/
void arm_mat_init_q15(
arm_matrix_instance_q15 * S,
uint16_t nRows,
uint16_t nColumns,
q15_t * pData);
/**
* @brief Floating-point matrix initialization.
* @param[in,out] S points to an instance of the floating-point matrix structure.
* @param[in] nRows number of rows in the matrix.
* @param[in] nColumns number of columns in the matrix.
* @param[in] pData points to the matrix data array.
*/
void arm_mat_init_f32(
arm_matrix_instance_f32 * S,
uint16_t nRows,
uint16_t nColumns,
float32_t * pData);
/**
* @brief Floating-point matrix initialization.
* @param[in,out] S points to an instance of the floating-point matrix structure.
* @param[in] nRows number of rows in the matrix.
* @param[in] nColumns number of columns in the matrix.
* @param[in] pData points to the matrix data array.
*/
void arm_mat_init_f64(
arm_matrix_instance_f64 * S,
uint16_t nRows,
uint16_t nColumns,
float64_t * pData);
/**
* @brief Floating-point matrix inverse.
* @param[in] src points to the instance of the input floating-point matrix structure.
* @param[out] dst points to the instance of the output floating-point matrix structure.
* @return The function returns ARM_MATH_SIZE_MISMATCH, if the dimensions do not match.
* If the input matrix is singular (does not have an inverse), then the algorithm terminates and returns error status ARM_MATH_SINGULAR.
*/
arm_status arm_mat_inverse_f32(
const arm_matrix_instance_f32 * src,
arm_matrix_instance_f32 * dst);
/**
* @brief Floating-point matrix inverse.
* @param[in] src points to the instance of the input floating-point matrix structure.
* @param[out] dst points to the instance of the output floating-point matrix structure.
* @return The function returns ARM_MATH_SIZE_MISMATCH, if the dimensions do not match.
* If the input matrix is singular (does not have an inverse), then the algorithm terminates and returns error status ARM_MATH_SINGULAR.
*/
arm_status arm_mat_inverse_f64(
const arm_matrix_instance_f64 * src,
arm_matrix_instance_f64 * dst);
/**
* @brief Floating-point Cholesky decomposition of Symmetric Positive Definite Matrix.
* @param[in] src points to the instance of the input floating-point matrix structure.
* @param[out] dst points to the instance of the output floating-point matrix structure.
* @return The function returns ARM_MATH_SIZE_MISMATCH, if the dimensions do not match.
* If the input matrix does not have a decomposition, then the algorithm terminates and returns error status ARM_MATH_DECOMPOSITION_FAILURE.
* If the matrix is ill conditioned or only semi-definite, then it is better using the LDL^t decomposition.
* The decomposition is returning a lower triangular matrix.
*/
arm_status arm_mat_cholesky_f64(
const arm_matrix_instance_f64 * src,
arm_matrix_instance_f64 * dst);
/**
* @brief Floating-point Cholesky decomposition of Symmetric Positive Definite Matrix.
* @param[in] src points to the instance of the input floating-point matrix structure.
* @param[out] dst points to the instance of the output floating-point matrix structure.
* @return The function returns ARM_MATH_SIZE_MISMATCH, if the dimensions do not match.
* If the input matrix does not have a decomposition, then the algorithm terminates and returns error status ARM_MATH_DECOMPOSITION_FAILURE.
* If the matrix is ill conditioned or only semi-definite, then it is better using the LDL^t decomposition.
* The decomposition is returning a lower triangular matrix.
*/
arm_status arm_mat_cholesky_f32(
const arm_matrix_instance_f32 * src,
arm_matrix_instance_f32 * dst);
/**
* @brief Solve UT . X = A where UT is an upper triangular matrix
* @param[in] ut The upper triangular matrix
* @param[in] a The matrix a
* @param[out] dst The solution X of UT . X = A
* @return The function returns ARM_MATH_SINGULAR, if the system can't be solved.
*/
arm_status arm_mat_solve_upper_triangular_f32(
const arm_matrix_instance_f32 * ut,
const arm_matrix_instance_f32 * a,
arm_matrix_instance_f32 * dst);
/**
* @brief Solve LT . X = A where LT is a lower triangular matrix
* @param[in] lt The lower triangular matrix
* @param[in] a The matrix a
* @param[out] dst The solution X of LT . X = A
* @return The function returns ARM_MATH_SINGULAR, if the system can't be solved.
*/
arm_status arm_mat_solve_lower_triangular_f32(
const arm_matrix_instance_f32 * lt,
const arm_matrix_instance_f32 * a,
arm_matrix_instance_f32 * dst);
/**
* @brief Solve UT . X = A where UT is an upper triangular matrix
* @param[in] ut The upper triangular matrix
* @param[in] a The matrix a
* @param[out] dst The solution X of UT . X = A
* @return The function returns ARM_MATH_SINGULAR, if the system can't be solved.
*/
arm_status arm_mat_solve_upper_triangular_f64(
const arm_matrix_instance_f64 * ut,
const arm_matrix_instance_f64 * a,
arm_matrix_instance_f64 * dst);
/**
* @brief Solve LT . X = A where LT is a lower triangular matrix
* @param[in] lt The lower triangular matrix
* @param[in] a The matrix a
* @param[out] dst The solution X of LT . X = A
* @return The function returns ARM_MATH_SINGULAR, if the system can't be solved.
*/
arm_status arm_mat_solve_lower_triangular_f64(
const arm_matrix_instance_f64 * lt,
const arm_matrix_instance_f64 * a,
arm_matrix_instance_f64 * dst);
/**
* @brief Floating-point LDL decomposition of Symmetric Positive Semi-Definite Matrix.
* @param[in] src points to the instance of the input floating-point matrix structure.
* @param[out] l points to the instance of the output floating-point triangular matrix structure.
* @param[out] d points to the instance of the output floating-point diagonal matrix structure.
* @param[out] p points to the instance of the output floating-point permutation vector.
* @return The function returns ARM_MATH_SIZE_MISMATCH, if the dimensions do not match.
* If the input matrix does not have a decomposition, then the algorithm terminates and returns error status ARM_MATH_DECOMPOSITION_FAILURE.
* The decomposition is returning a lower triangular matrix.
*/
arm_status arm_mat_ldlt_f32(
const arm_matrix_instance_f32 * src,
arm_matrix_instance_f32 * l,
arm_matrix_instance_f32 * d,
uint16_t * pp);
/**
* @brief Floating-point LDL decomposition of Symmetric Positive Semi-Definite Matrix.
* @param[in] src points to the instance of the input floating-point matrix structure.
* @param[out] l points to the instance of the output floating-point triangular matrix structure.
* @param[out] d points to the instance of the output floating-point diagonal matrix structure.
* @param[out] p points to the instance of the output floating-point permutation vector.
* @return The function returns ARM_MATH_SIZE_MISMATCH, if the dimensions do not match.
* If the input matrix does not have a decomposition, then the algorithm terminates and returns error status ARM_MATH_DECOMPOSITION_FAILURE.
* The decomposition is returning a lower triangular matrix.
*/
arm_status arm_mat_ldlt_f64(
const arm_matrix_instance_f64 * src,
arm_matrix_instance_f64 * l,
arm_matrix_instance_f64 * d,
uint16_t * pp);
/**
@brief QR decomposition of a m x n floating point matrix with m >= n.
@param[in] pSrc points to input matrix structure. The source matrix is modified by the function.
@param[in] threshold norm2 threshold.
@param[out] pOutR points to output R matrix structure of dimension m x n
@param[out] pOutQ points to output Q matrix structure of dimension m x m
@param[out] pOutTau points to Householder scaling factors of dimension n
@param[inout] pTmpA points to a temporary vector of dimension m.
@param[inout] pTmpB points to a temporary vector of dimension n.
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_SIZE_MISMATCH : Matrix size check failed
- \ref ARM_MATH_SINGULAR : Input matrix is found to be singular (non-invertible)
*/
arm_status arm_mat_qr_f32(
const arm_matrix_instance_f32 * pSrc,
const float32_t threshold,
arm_matrix_instance_f32 * pOutR,
arm_matrix_instance_f32 * pOutQ,
float32_t * pOutTau,
float32_t *pTmpA,
float32_t *pTmpB
);
/**
@brief QR decomposition of a m x n floating point matrix with m >= n.
@param[in] pSrc points to input matrix structure. The source matrix is modified by the function.
@param[in] threshold norm2 threshold.
@param[out] pOutR points to output R matrix structure of dimension m x n
@param[out] pOutQ points to output Q matrix structure of dimension m x m
@param[out] pOutTau points to Householder scaling factors of dimension n
@param[inout] pTmpA points to a temporary vector of dimension m.
@param[inout] pTmpB points to a temporary vector of dimension n.
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_SIZE_MISMATCH : Matrix size check failed
- \ref ARM_MATH_SINGULAR : Input matrix is found to be singular (non-invertible)
*/
arm_status arm_mat_qr_f64(
const arm_matrix_instance_f64 * pSrc,
const float64_t threshold,
arm_matrix_instance_f64 * pOutR,
arm_matrix_instance_f64 * pOutQ,
float64_t * pOutTau,
float64_t *pTmpA,
float64_t *pTmpB
);
/**
@brief Householder transform of a floating point vector.
@param[in] pSrc points to the input vector.
@param[in] threshold norm2 threshold.
@param[in] blockSize dimension of the vector space.
@param[outQ] pOut points to the output vector.
@return beta return the scaling factor beta
*/
float32_t arm_householder_f32(
const float32_t * pSrc,
const float32_t threshold,
uint32_t blockSize,
float32_t * pOut
);
/**
@brief Householder transform of a double floating point vector.
@param[in] pSrc points to the input vector.
@param[in] threshold norm2 threshold.
@param[in] blockSize dimension of the vector space.
@param[outQ] pOut points to the output vector.
@return beta return the scaling factor beta
*/
float64_t arm_householder_f64(
const float64_t * pSrc,
const float64_t threshold,
uint32_t blockSize,
float64_t * pOut
);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _MATRIX_FUNCTIONS_H_ */

261
sdk/CMSIS/DSP/Include/dsp/matrix_functions_f16.h Normal file
View file

@ -0,0 +1,261 @@
/******************************************************************************
* @file matrix_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef MATRIX_FUNCTIONS_F16_H_
#define MATRIX_FUNCTIONS_F16_H_
#ifdef __cplusplus
extern "C"
{
#endif
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#if defined(ARM_FLOAT16_SUPPORTED)
#define DEFAULT_HOUSEHOLDER_THRESHOLD_F16 (1.0e-3f)
/**
* @brief Instance structure for the floating-point matrix structure.
*/
typedef struct
{
uint16_t numRows; /**< number of rows of the matrix. */
uint16_t numCols; /**< number of columns of the matrix. */
float16_t *pData; /**< points to the data of the matrix. */
} arm_matrix_instance_f16;
/**
* @brief Floating-point matrix addition.
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_add_f16(
const arm_matrix_instance_f16 * pSrcA,
const arm_matrix_instance_f16 * pSrcB,
arm_matrix_instance_f16 * pDst);
/**
* @brief Floating-point, complex, matrix multiplication.
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_cmplx_mult_f16(
const arm_matrix_instance_f16 * pSrcA,
const arm_matrix_instance_f16 * pSrcB,
arm_matrix_instance_f16 * pDst);
/**
* @brief Floating-point matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_trans_f16(
const arm_matrix_instance_f16 * pSrc,
arm_matrix_instance_f16 * pDst);
/**
* @brief Floating-point complex matrix transpose.
* @param[in] pSrc points to the input matrix
* @param[out] pDst points to the output matrix
* @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
* or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_cmplx_trans_f16(
const arm_matrix_instance_f16 * pSrc,
arm_matrix_instance_f16 * pDst);
/**
* @brief Floating-point matrix multiplication
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_mult_f16(
const arm_matrix_instance_f16 * pSrcA,
const arm_matrix_instance_f16 * pSrcB,
arm_matrix_instance_f16 * pDst);
/**
* @brief Floating-point matrix and vector multiplication
* @param[in] pSrcMat points to the input matrix structure
* @param[in] pVec points to vector
* @param[out] pDst points to output vector
*/
void arm_mat_vec_mult_f16(
const arm_matrix_instance_f16 *pSrcMat,
const float16_t *pVec,
float16_t *pDst);
/**
* @brief Floating-point matrix subtraction
* @param[in] pSrcA points to the first input matrix structure
* @param[in] pSrcB points to the second input matrix structure
* @param[out] pDst points to output matrix structure
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_sub_f16(
const arm_matrix_instance_f16 * pSrcA,
const arm_matrix_instance_f16 * pSrcB,
arm_matrix_instance_f16 * pDst);
/**
* @brief Floating-point matrix scaling.
* @param[in] pSrc points to the input matrix
* @param[in] scale scale factor
* @param[out] pDst points to the output matrix
* @return The function returns either
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
*/
arm_status arm_mat_scale_f16(
const arm_matrix_instance_f16 * pSrc,
float16_t scale,
arm_matrix_instance_f16 * pDst);
/**
* @brief Floating-point matrix initialization.
* @param[in,out] S points to an instance of the floating-point matrix structure.
* @param[in] nRows number of rows in the matrix.
* @param[in] nColumns number of columns in the matrix.
* @param[in] pData points to the matrix data array.
*/
void arm_mat_init_f16(
arm_matrix_instance_f16 * S,
uint16_t nRows,
uint16_t nColumns,
float16_t * pData);
/**
* @brief Floating-point matrix inverse.
* @param[in] src points to the instance of the input floating-point matrix structure.
* @param[out] dst points to the instance of the output floating-point matrix structure.
* @return The function returns ARM_MATH_SIZE_MISMATCH, if the dimensions do not match.
* If the input matrix is singular (does not have an inverse), then the algorithm terminates and returns error status ARM_MATH_SINGULAR.
*/
arm_status arm_mat_inverse_f16(
const arm_matrix_instance_f16 * src,
arm_matrix_instance_f16 * dst);
/**
* @brief Floating-point Cholesky decomposition of Symmetric Positive Definite Matrix.
* @param[in] src points to the instance of the input floating-point matrix structure.
* @param[out] dst points to the instance of the output floating-point matrix structure.
* @return The function returns ARM_MATH_SIZE_MISMATCH, if the dimensions do not match.
* If the input matrix does not have a decomposition, then the algorithm terminates and returns error status ARM_MATH_DECOMPOSITION_FAILURE.
* If the matrix is ill conditioned or only semi-definite, then it is better using the LDL^t decomposition.
* The decomposition is returning a lower triangular matrix.
*/
arm_status arm_mat_cholesky_f16(
const arm_matrix_instance_f16 * src,
arm_matrix_instance_f16 * dst);
/**
* @brief Solve UT . X = A where UT is an upper triangular matrix
* @param[in] ut The upper triangular matrix
* @param[in] a The matrix a
* @param[out] dst The solution X of UT . X = A
* @return The function returns ARM_MATH_SINGULAR, if the system can't be solved.
*/
arm_status arm_mat_solve_upper_triangular_f16(
const arm_matrix_instance_f16 * ut,
const arm_matrix_instance_f16 * a,
arm_matrix_instance_f16 * dst);
/**
* @brief Solve LT . X = A where LT is a lower triangular matrix
* @param[in] lt The lower triangular matrix
* @param[in] a The matrix a
* @param[out] dst The solution X of LT . X = A
* @return The function returns ARM_MATH_SINGULAR, if the system can't be solved.
*/
arm_status arm_mat_solve_lower_triangular_f16(
const arm_matrix_instance_f16 * lt,
const arm_matrix_instance_f16 * a,
arm_matrix_instance_f16 * dst);
/**
@brief QR decomposition of a m x n floating point matrix with m >= n.
@param[in] pSrc points to input matrix structure. The source matrix is modified by the function.
@param[in] threshold norm2 threshold.
@param[out] pOutR points to output R matrix structure of dimension m x n
@param[out] pOutQ points to output Q matrix structure of dimension m x m
@param[out] pOutTau points to Householder scaling factors of dimension n
@param[inout] pTmpA points to a temporary vector of dimension m.
@param[inout] pTmpB points to a temporary vector of dimension n.
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_SIZE_MISMATCH : Matrix size check failed
- \ref ARM_MATH_SINGULAR : Input matrix is found to be singular (non-invertible)
*/
arm_status arm_mat_qr_f16(
const arm_matrix_instance_f16 * pSrc,
const float16_t threshold,
arm_matrix_instance_f16 * pOutR,
arm_matrix_instance_f16 * pOutQ,
float16_t * pOutTau,
float16_t *pTmpA,
float16_t *pTmpB
);
/**
@brief Householder transform of a half floating point vector.
@param[in] pSrc points to the input vector.
@param[in] threshold norm2 threshold.
@param[in] blockSize dimension of the vector space.
@param[outQ] pOut points to the output vector.
@return beta return the scaling factor beta
*/
float16_t arm_householder_f16(
const float16_t * pSrc,
const float16_t threshold,
uint32_t blockSize,
float16_t * pOut
);
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _MATRIX_FUNCTIONS_F16_H_ */

640
sdk/CMSIS/DSP/Include/dsp/matrix_utils.h Normal file
View file

@ -0,0 +1,640 @@
/******************************************************************************
* @file matrix_utils.h
* @brief Public header file for CMSIS DSP Library
* @version V1.11.0
* @date 30 May 2022
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2022 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef MATRIX_UTILS_H_
#define MATRIX_UTILS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#define ELEM(A,ROW,COL) &((A)->pData[(A)->numCols* (ROW) + (COL)])
#define SCALE_COL_T(T,CAST,A,ROW,v,i) \
{ \
int32_t _w; \
T *data = (A)->pData; \
const int32_t _numCols = (A)->numCols; \
const int32_t nb = (A)->numRows - ROW;\
\
data += i + _numCols * (ROW); \
\
for(_w=0;_w < nb; _w++) \
{ \
*data *= CAST v; \
data += _numCols; \
} \
}
#define COPY_COL_T(T,A,ROW,COL,DST) \
{ \
uint32_t _row; \
T *_pb=DST; \
T *_pa = (A)->pData + ROW * (A)->numCols + COL;\
for(_row = ROW; _row < (A)->numRows; _row ++) \
{ \
*_pb++ = *_pa; \
_pa += (A)->numCols; \
} \
}
#if defined(ARM_FLOAT16_SUPPORTED)
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
#define SWAP_ROWS_F16(A,COL,i,j) \
{ \
int cnt = ((A)->numCols)-(COL); \
int32_t _w; \
float16_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols; \
\
for(_w=(COL);_w < _numCols; _w+=8) \
{ \
f16x8_t tmpa,tmpb; \
mve_pred16_t p0 = vctp16q(cnt); \
\
tmpa=vldrhq_z_f16(&data[i*_numCols + _w],p0);\
tmpb=vldrhq_z_f16(&data[j*_numCols + _w],p0);\
\
vstrhq_p(&data[i*_numCols + _w], tmpb, p0); \
vstrhq_p(&data[j*_numCols + _w], tmpa, p0); \
\
cnt -= 8; \
} \
}
#define SCALE_ROW_F16(A,COL,v,i) \
{ \
int cnt = ((A)->numCols)-(COL); \
int32_t _w; \
float16_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols; \
\
for(_w=(COL);_w < _numCols; _w+=8) \
{ \
f16x8_t tmpa; \
mve_pred16_t p0 = vctp16q(cnt); \
tmpa = vldrhq_z_f16(&data[i*_numCols + _w],p0);\
tmpa = vmulq_n_f16(tmpa,(_Float16)v); \
vstrhq_p(&data[i*_numCols + _w], tmpa, p0); \
cnt -= 8; \
} \
\
}
#define MAC_ROW_F16(COL,A,i,v,B,j) \
{ \
int cnt = ((A)->numCols)-(COL); \
int32_t _w; \
float16_t *dataA = (A)->pData; \
float16_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols; \
\
for(_w=(COL);_w < _numCols; _w+=8) \
{ \
f16x8_t tmpa,tmpb; \
mve_pred16_t p0 = vctp16q(cnt); \
tmpa = vldrhq_z_f16(&dataA[i*_numCols + _w],p0);\
tmpb = vldrhq_z_f16(&dataB[j*_numCols + _w],p0);\
tmpa = vfmaq_n_f16(tmpa,tmpb,v); \
vstrhq_p(&dataA[i*_numCols + _w], tmpa, p0); \
cnt -= 8; \
} \
\
}
#define MAS_ROW_F16(COL,A,i,v,B,j) \
{ \
int cnt = ((A)->numCols)-(COL); \
int32_t _w; \
float16_t *dataA = (A)->pData; \
float16_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols; \
f16x8_t vec=vdupq_n_f16(v); \
\
for(_w=(COL);_w < _numCols; _w+=8) \
{ \
f16x8_t tmpa,tmpb; \
mve_pred16_t p0 = vctp16q(cnt); \
tmpa = vldrhq_z_f16(&dataA[i*_numCols + _w],p0);\
tmpb = vldrhq_z_f16(&dataB[j*_numCols + _w],p0);\
tmpa = vfmsq_f16(tmpa,tmpb,vec); \
vstrhq_p(&dataA[i*_numCols + _w], tmpa, p0); \
cnt -= 8; \
} \
\
}
#else
#define SWAP_ROWS_F16(A,COL,i,j) \
{ \
int32_t _w; \
float16_t *dataI = (A)->pData; \
float16_t *dataJ = (A)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols-(COL); \
\
dataI += i*_numCols + (COL); \
dataJ += j*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
float16_t tmp; \
tmp = *dataI; \
*dataI++ = *dataJ; \
*dataJ++ = tmp; \
} \
}
#define SCALE_ROW_F16(A,COL,v,i) \
{ \
int32_t _w; \
float16_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols-(COL); \
\
data += i*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*data++ *= (_Float16)v; \
} \
}
#define MAC_ROW_F16(COL,A,i,v,B,j) \
{ \
int32_t _w; \
float16_t *dataA = (A)->pData; \
float16_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols; \
const int32_t nb = _numCols-(COL); \
\
dataA += i*_numCols + (COL); \
dataB += j*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*dataA++ += (_Float16)v * (_Float16)*dataB++;\
} \
}
#define MAS_ROW_F16(COL,A,i,v,B,j) \
{ \
int32_t _w; \
float16_t *dataA = (A)->pData; \
float16_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols; \
const int32_t nb = _numCols-(COL); \
\
dataA += i*_numCols + (COL); \
dataB += j*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*dataA++ -= (_Float16)v * (_Float16)*dataB++;\
} \
}
#endif /*defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)*/
/* Functions with only a scalar version */
#define COPY_COL_F16(A,ROW,COL,DST) \
COPY_COL_T(float16_t,A,ROW,COL,DST)
#define SCALE_COL_F16(A,ROW,v,i) \
SCALE_COL_T(float16_t,(_Float16),A,ROW,v,i)
#endif /* defined(ARM_FLOAT16_SUPPORTED)*/
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
#define SWAP_ROWS_F32(A,COL,i,j) \
{ \
int cnt = ((A)->numCols)-(COL); \
float32_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols; \
int32_t _w; \
\
for(_w=(COL);_w < _numCols; _w+=4) \
{ \
f32x4_t tmpa,tmpb; \
mve_pred16_t p0 = vctp32q(cnt); \
\
tmpa=vldrwq_z_f32(&data[i*_numCols + _w],p0);\
tmpb=vldrwq_z_f32(&data[j*_numCols + _w],p0);\
\
vstrwq_p(&data[i*_numCols + _w], tmpb, p0); \
vstrwq_p(&data[j*_numCols + _w], tmpa, p0); \
\
cnt -= 4; \
} \
}
#define MAC_ROW_F32(COL,A,i,v,B,j) \
{ \
int cnt = ((A)->numCols)-(COL); \
float32_t *dataA = (A)->pData; \
float32_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols; \
int32_t _w; \
\
for(_w=(COL);_w < _numCols; _w+=4) \
{ \
f32x4_t tmpa,tmpb; \
mve_pred16_t p0 = vctp32q(cnt); \
tmpa = vldrwq_z_f32(&dataA[i*_numCols + _w],p0);\
tmpb = vldrwq_z_f32(&dataB[j*_numCols + _w],p0);\
tmpa = vfmaq_n_f32(tmpa,tmpb,v); \
vstrwq_p(&dataA[i*_numCols + _w], tmpa, p0); \
cnt -= 4; \
} \
\
}
#define MAS_ROW_F32(COL,A,i,v,B,j) \
{ \
int cnt = ((A)->numCols)-(COL); \
float32_t *dataA = (A)->pData; \
float32_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols; \
int32_t _w; \
f32x4_t vec=vdupq_n_f32(v); \
\
for(_w=(COL);_w < _numCols; _w+=4) \
{ \
f32x4_t tmpa,tmpb; \
mve_pred16_t p0 = vctp32q(cnt); \
tmpa = vldrwq_z_f32(&dataA[i*_numCols + _w],p0);\
tmpb = vldrwq_z_f32(&dataB[j*_numCols + _w],p0);\
tmpa = vfmsq_f32(tmpa,tmpb,vec); \
vstrwq_p(&dataA[i*_numCols + _w], tmpa, p0); \
cnt -= 4; \
} \
\
}
#define SCALE_ROW_F32(A,COL,v,i) \
{ \
int cnt = ((A)->numCols)-(COL); \
float32_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols; \
int32_t _w; \
\
for(_w=(COL);_w < _numCols; _w+=4) \
{ \
f32x4_t tmpa; \
mve_pred16_t p0 = vctp32q(cnt); \
tmpa = vldrwq_z_f32(&data[i*_numCols + _w],p0);\
tmpa = vmulq_n_f32(tmpa,v); \
vstrwq_p(&data[i*_numCols + _w], tmpa, p0); \
cnt -= 4; \
} \
\
}
#elif defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
#define SWAP_ROWS_F32(A,COL,i,j) \
{ \
int32_t _w; \
float32_t *dataI = (A)->pData; \
float32_t *dataJ = (A)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols - COL; \
\
dataI += i*_numCols + (COL); \
dataJ += j*_numCols + (COL); \
\
float32_t tmp; \
\
for(_w=0;_w < nb; _w++) \
{ \
tmp = *dataI; \
*dataI++ = *dataJ; \
*dataJ++ = tmp; \
} \
}
#define MAC_ROW_F32(COL,A,i,v,B,j) \
{ \
float32_t *dataA = (A)->pData; \
float32_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols - (COL); \
int32_t nbElems; \
f32x4_t vec = vdupq_n_f32(v); \
\
nbElems = nb >> 2; \
\
dataA += i*_numCols + (COL); \
dataB += j*_numCols + (COL); \
\
while(nbElems>0) \
{ \
f32x4_t tmpa,tmpb; \
tmpa = vld1q_f32(dataA,p0); \
tmpb = vld1q_f32(dataB,p0); \
tmpa = vmlaq_f32(tmpa,tmpb,vec);\
vst1q_f32(dataA, tmpa, p0); \
nbElems--; \
dataA += 4; \
dataB += 4; \
} \
\
nbElems = nb & 3; \
while(nbElems > 0) \
{ \
*dataA++ += v* *dataB++; \
nbElems--; \
} \
}
#define MAS_ROW_F32(COL,A,i,v,B,j) \
{ \
float32_t *dataA = (A)->pData; \
float32_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols - (COL); \
int32_t nbElems; \
f32x4_t vec = vdupq_n_f32(v); \
\
nbElems = nb >> 2; \
\
dataA += i*_numCols + (COL); \
dataB += j*_numCols + (COL); \
\
while(nbElems>0) \
{ \
f32x4_t tmpa,tmpb; \
tmpa = vld1q_f32(dataA); \
tmpb = vld1q_f32(dataB); \
tmpa = vmlsq_f32(tmpa,tmpb,vec);\
vst1q_f32(dataA, tmpa); \
nbElems--; \
dataA += 4; \
dataB += 4; \
} \
\
nbElems = nb & 3; \
while(nbElems > 0) \
{ \
*dataA++ -= v* *dataB++; \
nbElems--; \
} \
}
#define SCALE_ROW_F32(A,COL,v,i) \
{ \
float32_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols; \
const int32_t nb = _numCols - (COL); \
int32_t nbElems; \
f32x4_t vec = vdupq_n_f32(v); \
\
nbElems = nb >> 2; \
\
data += i*_numCols + (COL); \
while(nbElems>0) \
{ \
f32x4_t tmpa; \
tmpa = vld1q_f32(data); \
tmpa = vmulq_f32(tmpa,vec); \
vst1q_f32(data, tmpa); \
data += 4; \
nbElems --; \
} \
\
nbElems = nb & 3; \
while(nbElems > 0) \
{ \
*data++ *= v; \
nbElems--; \
} \
\
}
#else
#define SWAP_ROWS_F32(A,COL,i,j) \
{ \
int32_t _w; \
float32_t tmp; \
float32_t *dataI = (A)->pData; \
float32_t *dataJ = (A)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols - COL; \
\
dataI += i*_numCols + (COL); \
dataJ += j*_numCols + (COL); \
\
\
for(_w=0;_w < nb; _w++) \
{ \
tmp = *dataI; \
*dataI++ = *dataJ; \
*dataJ++ = tmp; \
} \
}
#define SCALE_ROW_F32(A,COL,v,i) \
{ \
int32_t _w; \
float32_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols - COL; \
\
data += i*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*data++ *= v; \
} \
}
#define MAC_ROW_F32(COL,A,i,v,B,j) \
{ \
int32_t _w; \
float32_t *dataA = (A)->pData; \
float32_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols-(COL); \
\
dataA = dataA + i*_numCols + (COL); \
dataB = dataB + j*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*dataA++ += v* *dataB++; \
} \
}
#define MAS_ROW_F32(COL,A,i,v,B,j) \
{ \
int32_t _w; \
float32_t *dataA = (A)->pData; \
float32_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols-(COL); \
\
dataA = dataA + i*_numCols + (COL); \
dataB = dataB + j*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*dataA++ -= v* *dataB++; \
} \
}
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/* Functions _with only a scalar version */
#define COPY_COL_F32(A,ROW,COL,DST) \
COPY_COL_T(float32_t,A,ROW,COL,DST)
#define COPY_COL_F64(A,ROW,COL,DST) \
COPY_COL_T(float64_t,A,ROW,COL,DST)
#define SWAP_COLS_F32(A,COL,i,j) \
{ \
int32_t _w; \
float32_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols; \
for(_w=(COL);_w < _numCols; _w++) \
{ \
float32_t tmp; \
tmp = data[_w*_numCols + i]; \
data[_w*_numCols + i] = data[_w*_numCols + j];\
data[_w*_numCols + j] = tmp; \
} \
}
#define SCALE_COL_F32(A,ROW,v,i) \
SCALE_COL_T(float32_t,,A,ROW,v,i)
#define SWAP_ROWS_F64(A,COL,i,j) \
{ \
int32_t _w; \
float64_t *dataI = (A)->pData; \
float64_t *dataJ = (A)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols-(COL); \
\
dataI += i*_numCols + (COL); \
dataJ += j*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
float64_t tmp; \
tmp = *dataI; \
*dataI++ = *dataJ; \
*dataJ++ = tmp; \
} \
}
#define SWAP_COLS_F64(A,COL,i,j) \
{ \
int32_t _w; \
float64_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols; \
for(_w=(COL);_w < _numCols; _w++) \
{ \
float64_t tmp; \
tmp = data[_w*_numCols + i]; \
data[_w*_numCols + i] = data[_w*_numCols + j];\
data[_w*_numCols + j] = tmp; \
} \
}
#define SCALE_ROW_F64(A,COL,v,i) \
{ \
int32_t _w; \
float64_t *data = (A)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols-(COL); \
\
data += i*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*data++ *= v; \
} \
}
#define SCALE_COL_F64(A,ROW,v,i) \
SCALE_COL_T(float64_t,,A,ROW,v,i)
#define MAC_ROW_F64(COL,A,i,v,B,j) \
{ \
int32_t _w; \
float64_t *dataA = (A)->pData; \
float64_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols-(COL); \
\
dataA += i*_numCols + (COL); \
dataB += j*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*dataA++ += v* *dataB++; \
} \
}
#define MAS_ROW_F64(COL,A,i,v,B,j) \
{ \
int32_t _w; \
float64_t *dataA = (A)->pData; \
float64_t *dataB = (B)->pData; \
const int32_t _numCols = (A)->numCols;\
const int32_t nb = _numCols-(COL); \
\
dataA += i*_numCols + (COL); \
dataB += j*_numCols + (COL); \
\
for(_w=0;_w < nb; _w++) \
{ \
*dataA++ -= v* *dataB++; \
} \
}
#ifdef __cplusplus
}
#endif
#endif /* ifndef _MATRIX_UTILS_H_ */

576
sdk/CMSIS/DSP/Include/dsp/none.h Normal file
View file

@ -0,0 +1,576 @@
/******************************************************************************
* @file none.h
* @brief Intrinsincs when no DSP extension available
* @version V1.9.0
* @date 20. July 2020
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
Definitions in this file are allowing to reuse some versions of the
CMSIS-DSP to build on a core (M0 for instance) or a host where
DSP extension are not available.
Ideally a pure C version should have been used instead.
But those are not always available or use a restricted set
of intrinsics.
*/
#ifndef NONE_H_
#define NONE_H_
#include "arm_math_types.h"
#ifdef __cplusplus
extern "C"
{
#endif
/*
Normally those kind of definitions are in a compiler file
in Core or Core_A.
But for MSVC compiler it is a bit special. The goal is very specific
to CMSIS-DSP and only to allow the use of this library from other
systems like Python or Matlab.
MSVC is not going to be used to cross-compile to ARM. So, having a MSVC
compiler file in Core or Core_A would not make sense.
*/
#if defined ( _MSC_VER ) || defined(__GNUC_PYTHON__) || defined(__APPLE_CC__)
__STATIC_FORCEINLINE uint8_t __CLZ(uint32_t data)
{
if (data == 0U) { return 32U; }
uint32_t count = 0U;
uint32_t mask = 0x80000000U;
while ((data & mask) == 0U)
{
count += 1U;
mask = mask >> 1U;
}
return count;
}
__STATIC_FORCEINLINE int32_t __SSAT(int32_t val, uint32_t sat)
{
if ((sat >= 1U) && (sat <= 32U))
{
const int32_t max = (int32_t)((1U << (sat - 1U)) - 1U);
const int32_t min = -1 - max ;
if (val > max)
{
return max;
}
else if (val < min)
{
return min;
}
}
return val;
}
__STATIC_FORCEINLINE uint32_t __USAT(int32_t val, uint32_t sat)
{
if (sat <= 31U)
{
const uint32_t max = ((1U << sat) - 1U);
if (val > (int32_t)max)
{
return max;
}
else if (val < 0)
{
return 0U;
}
}
return (uint32_t)val;
}
/**
\brief Rotate Right in unsigned value (32 bit)
\details Rotate Right (immediate) provides the value of the contents of a register rotated by a variable number of bits.
\param [in] op1 Value to rotate
\param [in] op2 Number of Bits to rotate
\return Rotated value
*/
__STATIC_FORCEINLINE uint32_t __ROR(uint32_t op1, uint32_t op2)
{
op2 %= 32U;
if (op2 == 0U)
{
return op1;
}
return (op1 >> op2) | (op1 << (32U - op2));
}
#endif
/**
* @brief Clips Q63 to Q31 values.
*/
__STATIC_FORCEINLINE q31_t clip_q63_to_q31(
q63_t x)
{
return ((q31_t) (x >> 32) != ((q31_t) x >> 31)) ?
((0x7FFFFFFF ^ ((q31_t) (x >> 63)))) : (q31_t) x;
}
/**
* @brief Clips Q63 to Q15 values.
*/
__STATIC_FORCEINLINE q15_t clip_q63_to_q15(
q63_t x)
{
return ((q31_t) (x >> 32) != ((q31_t) x >> 31)) ?
((0x7FFF ^ ((q15_t) (x >> 63)))) : (q15_t) (x >> 15);
}
/**
* @brief Clips Q31 to Q7 values.
*/
__STATIC_FORCEINLINE q7_t clip_q31_to_q7(
q31_t x)
{
return ((q31_t) (x >> 24) != ((q31_t) x >> 23)) ?
((0x7F ^ ((q7_t) (x >> 31)))) : (q7_t) x;
}
/**
* @brief Clips Q31 to Q15 values.
*/
__STATIC_FORCEINLINE q15_t clip_q31_to_q15(
q31_t x)
{
return ((q31_t) (x >> 16) != ((q31_t) x >> 15)) ?
((0x7FFF ^ ((q15_t) (x >> 31)))) : (q15_t) x;
}
/**
* @brief Multiplies 32 X 64 and returns 32 bit result in 2.30 format.
*/
__STATIC_FORCEINLINE q63_t mult32x64(
q63_t x,
q31_t y)
{
return ((((q63_t) (x & 0x00000000FFFFFFFF) * y) >> 32) +
(((q63_t) (x >> 32) * y) ) );
}
/* SMMLAR */
#define multAcc_32x32_keep32_R(a, x, y) \
a = (q31_t) (((((q63_t) a) << 32) + ((q63_t) x * y) + 0x80000000LL ) >> 32)
/* SMMLSR */
#define multSub_32x32_keep32_R(a, x, y) \
a = (q31_t) (((((q63_t) a) << 32) - ((q63_t) x * y) + 0x80000000LL ) >> 32)
/* SMMULR */
#define mult_32x32_keep32_R(a, x, y) \
a = (q31_t) (((q63_t) x * y + 0x80000000LL ) >> 32)
/* SMMLA */
#define multAcc_32x32_keep32(a, x, y) \
a += (q31_t) (((q63_t) x * y) >> 32)
/* SMMLS */
#define multSub_32x32_keep32(a, x, y) \
a -= (q31_t) (((q63_t) x * y) >> 32)
/* SMMUL */
#define mult_32x32_keep32(a, x, y) \
a = (q31_t) (((q63_t) x * y ) >> 32)
#ifndef ARM_MATH_DSP
/**
* @brief definition to pack two 16 bit values.
*/
#define __PKHBT(ARG1, ARG2, ARG3) ( (((int32_t)(ARG1) << 0) & (int32_t)0x0000FFFF) | \
(((int32_t)(ARG2) << ARG3) & (int32_t)0xFFFF0000) )
#define __PKHTB(ARG1, ARG2, ARG3) ( (((int32_t)(ARG1) << 0) & (int32_t)0xFFFF0000) | \
(((int32_t)(ARG2) >> ARG3) & (int32_t)0x0000FFFF) )
#endif
/**
* @brief definition to pack four 8 bit values.
*/
#ifndef ARM_MATH_BIG_ENDIAN
#define __PACKq7(v0,v1,v2,v3) ( (((int32_t)(v0) << 0) & (int32_t)0x000000FF) | \
(((int32_t)(v1) << 8) & (int32_t)0x0000FF00) | \
(((int32_t)(v2) << 16) & (int32_t)0x00FF0000) | \
(((int32_t)(v3) << 24) & (int32_t)0xFF000000) )
#else
#define __PACKq7(v0,v1,v2,v3) ( (((int32_t)(v3) << 0) & (int32_t)0x000000FF) | \
(((int32_t)(v2) << 8) & (int32_t)0x0000FF00) | \
(((int32_t)(v1) << 16) & (int32_t)0x00FF0000) | \
(((int32_t)(v0) << 24) & (int32_t)0xFF000000) )
#endif
/*
* @brief C custom defined intrinsic functions
*/
#if !defined (ARM_MATH_DSP)
/*
* @brief C custom defined QADD8
*/
__STATIC_FORCEINLINE uint32_t __QADD8(
uint32_t x,
uint32_t y)
{
q31_t r, s, t, u;
r = __SSAT(((((q31_t)x << 24) >> 24) + (((q31_t)y << 24) >> 24)), 8) & (int32_t)0x000000FF;
s = __SSAT(((((q31_t)x << 16) >> 24) + (((q31_t)y << 16) >> 24)), 8) & (int32_t)0x000000FF;
t = __SSAT(((((q31_t)x << 8) >> 24) + (((q31_t)y << 8) >> 24)), 8) & (int32_t)0x000000FF;
u = __SSAT(((((q31_t)x ) >> 24) + (((q31_t)y ) >> 24)), 8) & (int32_t)0x000000FF;
return ((uint32_t)((u << 24) | (t << 16) | (s << 8) | (r )));
}
/*
* @brief C custom defined QSUB8
*/
__STATIC_FORCEINLINE uint32_t __QSUB8(
uint32_t x,
uint32_t y)
{
q31_t r, s, t, u;
r = __SSAT(((((q31_t)x << 24) >> 24) - (((q31_t)y << 24) >> 24)), 8) & (int32_t)0x000000FF;
s = __SSAT(((((q31_t)x << 16) >> 24) - (((q31_t)y << 16) >> 24)), 8) & (int32_t)0x000000FF;
t = __SSAT(((((q31_t)x << 8) >> 24) - (((q31_t)y << 8) >> 24)), 8) & (int32_t)0x000000FF;
u = __SSAT(((((q31_t)x ) >> 24) - (((q31_t)y ) >> 24)), 8) & (int32_t)0x000000FF;
return ((uint32_t)((u << 24) | (t << 16) | (s << 8) | (r )));
}
/*
* @brief C custom defined QADD16
*/
__STATIC_FORCEINLINE uint32_t __QADD16(
uint32_t x,
uint32_t y)
{
/* q31_t r, s; without initialisation 'arm_offset_q15 test' fails but 'intrinsic' tests pass! for armCC */
q31_t r = 0, s = 0;
r = __SSAT(((((q31_t)x << 16) >> 16) + (((q31_t)y << 16) >> 16)), 16) & (int32_t)0x0000FFFF;
s = __SSAT(((((q31_t)x ) >> 16) + (((q31_t)y ) >> 16)), 16) & (int32_t)0x0000FFFF;
return ((uint32_t)((s << 16) | (r )));
}
/*
* @brief C custom defined SHADD16
*/
__STATIC_FORCEINLINE uint32_t __SHADD16(
uint32_t x,
uint32_t y)
{
q31_t r, s;
r = (((((q31_t)x << 16) >> 16) + (((q31_t)y << 16) >> 16)) >> 1) & (int32_t)0x0000FFFF;
s = (((((q31_t)x ) >> 16) + (((q31_t)y ) >> 16)) >> 1) & (int32_t)0x0000FFFF;
return ((uint32_t)((s << 16) | (r )));
}
/*
* @brief C custom defined QSUB16
*/
__STATIC_FORCEINLINE uint32_t __QSUB16(
uint32_t x,
uint32_t y)
{
q31_t r, s;
r = __SSAT(((((q31_t)x << 16) >> 16) - (((q31_t)y << 16) >> 16)), 16) & (int32_t)0x0000FFFF;
s = __SSAT(((((q31_t)x ) >> 16) - (((q31_t)y ) >> 16)), 16) & (int32_t)0x0000FFFF;
return ((uint32_t)((s << 16) | (r )));
}
/*
* @brief C custom defined SHSUB16
*/
__STATIC_FORCEINLINE uint32_t __SHSUB16(
uint32_t x,
uint32_t y)
{
q31_t r, s;
r = (((((q31_t)x << 16) >> 16) - (((q31_t)y << 16) >> 16)) >> 1) & (int32_t)0x0000FFFF;
s = (((((q31_t)x ) >> 16) - (((q31_t)y ) >> 16)) >> 1) & (int32_t)0x0000FFFF;
return ((uint32_t)((s << 16) | (r )));
}
/*
* @brief C custom defined QASX
*/
__STATIC_FORCEINLINE uint32_t __QASX(
uint32_t x,
uint32_t y)
{
q31_t r, s;
r = __SSAT(((((q31_t)x << 16) >> 16) - (((q31_t)y ) >> 16)), 16) & (int32_t)0x0000FFFF;
s = __SSAT(((((q31_t)x ) >> 16) + (((q31_t)y << 16) >> 16)), 16) & (int32_t)0x0000FFFF;
return ((uint32_t)((s << 16) | (r )));
}
/*
* @brief C custom defined SHASX
*/
__STATIC_FORCEINLINE uint32_t __SHASX(
uint32_t x,
uint32_t y)
{
q31_t r, s;
r = (((((q31_t)x << 16) >> 16) - (((q31_t)y ) >> 16)) >> 1) & (int32_t)0x0000FFFF;
s = (((((q31_t)x ) >> 16) + (((q31_t)y << 16) >> 16)) >> 1) & (int32_t)0x0000FFFF;
return ((uint32_t)((s << 16) | (r )));
}
/*
* @brief C custom defined QSAX
*/
__STATIC_FORCEINLINE uint32_t __QSAX(
uint32_t x,
uint32_t y)
{
q31_t r, s;
r = __SSAT(((((q31_t)x << 16) >> 16) + (((q31_t)y ) >> 16)), 16) & (int32_t)0x0000FFFF;
s = __SSAT(((((q31_t)x ) >> 16) - (((q31_t)y << 16) >> 16)), 16) & (int32_t)0x0000FFFF;
return ((uint32_t)((s << 16) | (r )));
}
/*
* @brief C custom defined SHSAX
*/
__STATIC_FORCEINLINE uint32_t __SHSAX(
uint32_t x,
uint32_t y)
{
q31_t r, s;
r = (((((q31_t)x << 16) >> 16) + (((q31_t)y ) >> 16)) >> 1) & (int32_t)0x0000FFFF;
s = (((((q31_t)x ) >> 16) - (((q31_t)y << 16) >> 16)) >> 1) & (int32_t)0x0000FFFF;
return ((uint32_t)((s << 16) | (r )));
}
/*
* @brief C custom defined SMUSDX
*/
__STATIC_FORCEINLINE uint32_t __SMUSDX(
uint32_t x,
uint32_t y)
{
return ((uint32_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y ) >> 16)) -
((((q31_t)x ) >> 16) * (((q31_t)y << 16) >> 16)) ));
}
/*
* @brief C custom defined SMUADX
*/
__STATIC_FORCEINLINE uint32_t __SMUADX(
uint32_t x,
uint32_t y)
{
return ((uint32_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y ) >> 16)) +
((((q31_t)x ) >> 16) * (((q31_t)y << 16) >> 16)) ));
}
/*
* @brief C custom defined QADD
*/
__STATIC_FORCEINLINE int32_t __QADD(
int32_t x,
int32_t y)
{
return ((int32_t)(clip_q63_to_q31((q63_t)x + (q31_t)y)));
}
/*
* @brief C custom defined QSUB
*/
__STATIC_FORCEINLINE int32_t __QSUB(
int32_t x,
int32_t y)
{
return ((int32_t)(clip_q63_to_q31((q63_t)x - (q31_t)y)));
}
/*
* @brief C custom defined SMLAD
*/
__STATIC_FORCEINLINE uint32_t __SMLAD(
uint32_t x,
uint32_t y,
uint32_t sum)
{
return ((uint32_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y << 16) >> 16)) +
((((q31_t)x ) >> 16) * (((q31_t)y ) >> 16)) +
( ((q31_t)sum ) ) ));
}
/*
* @brief C custom defined SMLADX
*/
__STATIC_FORCEINLINE uint32_t __SMLADX(
uint32_t x,
uint32_t y,
uint32_t sum)
{
return ((uint32_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y ) >> 16)) +
((((q31_t)x ) >> 16) * (((q31_t)y << 16) >> 16)) +
( ((q31_t)sum ) ) ));
}
/*
* @brief C custom defined SMLSDX
*/
__STATIC_FORCEINLINE uint32_t __SMLSDX(
uint32_t x,
uint32_t y,
uint32_t sum)
{
return ((uint32_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y ) >> 16)) -
((((q31_t)x ) >> 16) * (((q31_t)y << 16) >> 16)) +
( ((q31_t)sum ) ) ));
}
/*
* @brief C custom defined SMLALD
*/
__STATIC_FORCEINLINE uint64_t __SMLALD(
uint32_t x,
uint32_t y,
uint64_t sum)
{
/* return (sum + ((q15_t) (x >> 16) * (q15_t) (y >> 16)) + ((q15_t) x * (q15_t) y)); */
return ((uint64_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y << 16) >> 16)) +
((((q31_t)x ) >> 16) * (((q31_t)y ) >> 16)) +
( ((q63_t)sum ) ) ));
}
/*
* @brief C custom defined SMLALDX
*/
__STATIC_FORCEINLINE uint64_t __SMLALDX(
uint32_t x,
uint32_t y,
uint64_t sum)
{
/* return (sum + ((q15_t) (x >> 16) * (q15_t) y)) + ((q15_t) x * (q15_t) (y >> 16)); */
return ((uint64_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y ) >> 16)) +
((((q31_t)x ) >> 16) * (((q31_t)y << 16) >> 16)) +
( ((q63_t)sum ) ) ));
}
/*
* @brief C custom defined SMUAD
*/
__STATIC_FORCEINLINE uint32_t __SMUAD(
uint32_t x,
uint32_t y)
{
return ((uint32_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y << 16) >> 16)) +
((((q31_t)x ) >> 16) * (((q31_t)y ) >> 16)) ));
}
/*
* @brief C custom defined SMUSD
*/
__STATIC_FORCEINLINE uint32_t __SMUSD(
uint32_t x,
uint32_t y)
{
return ((uint32_t)(((((q31_t)x << 16) >> 16) * (((q31_t)y << 16) >> 16)) -
((((q31_t)x ) >> 16) * (((q31_t)y ) >> 16)) ));
}
/*
* @brief C custom defined SXTB16
*/
__STATIC_FORCEINLINE uint32_t __SXTB16(
uint32_t x)
{
return ((uint32_t)(((((q31_t)x << 24) >> 24) & (q31_t)0x0000FFFF) |
((((q31_t)x << 8) >> 8) & (q31_t)0xFFFF0000) ));
}
/*
* @brief C custom defined SMMLA
*/
__STATIC_FORCEINLINE int32_t __SMMLA(
int32_t x,
int32_t y,
int32_t sum)
{
return (sum + (int32_t) (((int64_t) x * y) >> 32));
}
#endif /* !defined (ARM_MATH_DSP) */
#ifdef __cplusplus
}
#endif
#endif /* ifndef _TRANSFORM_FUNCTIONS_H_ */

153
sdk/CMSIS/DSP/Include/dsp/quaternion_math_functions.h Normal file
View file

@ -0,0 +1,153 @@
/******************************************************************************
* @file quaternion_math_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
*
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2021 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef QUATERNION_MATH_FUNCTIONS_H_
#define QUATERNION_MATH_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupQuaternionMath Quaternion Math Functions
* Functions to operates on quaternions and convert between a
* rotation and quaternion representation.
*/
/**
@brief Floating-point quaternion Norm.
@param[in] pInputQuaternions points to the input vector of quaternions
@param[out] pNorms points to the output vector of norms
@param[in] nbQuaternions number of quaternions in each vector
*/
void arm_quaternion_norm_f32(const float32_t *pInputQuaternions,
float32_t *pNorms,
uint32_t nbQuaternions);
/**
@brief Floating-point quaternion inverse.
@param[in] pInputQuaternions points to the input vector of quaternions
@param[out] pInverseQuaternions points to the output vector of inverse quaternions
@param[in] nbQuaternions number of quaternions in each vector
*/
void arm_quaternion_inverse_f32(const float32_t *pInputQuaternions,
float32_t *pInverseQuaternions,
uint32_t nbQuaternions);
/**
@brief Floating-point quaternion conjugates.
@param[in] pInputQuaternions points to the input vector of quaternions
@param[out] pConjugateQuaternions points to the output vector of conjugate quaternions
@param[in] nbQuaternions number of quaternions in each vector
*/
void arm_quaternion_conjugate_f32(const float32_t *inputQuaternions,
float32_t *pConjugateQuaternions,
uint32_t nbQuaternions);
/**
@brief Floating-point normalization of quaternions.
@param[in] pInputQuaternions points to the input vector of quaternions
@param[out] pNormalizedQuaternions points to the output vector of normalized quaternions
@param[in] nbQuaternions number of quaternions in each vector
*/
void arm_quaternion_normalize_f32(const float32_t *inputQuaternions,
float32_t *pNormalizedQuaternions,
uint32_t nbQuaternions);
/**
@brief Floating-point product of two quaternions.
@param[in] qa First quaternion
@param[in] qb Second quaternion
@param[out] r Product of two quaternions
*/
void arm_quaternion_product_single_f32(const float32_t *qa,
const float32_t *qb,
float32_t *r);
/**
@brief Floating-point elementwise product two quaternions.
@param[in] qa First array of quaternions
@param[in] qb Second array of quaternions
@param[out] r Elementwise product of quaternions
@param[in] nbQuaternions Number of quaternions in the array
*/
void arm_quaternion_product_f32(const float32_t *qa,
const float32_t *qb,
float32_t *r,
uint32_t nbQuaternions);
/**
* @brief Conversion of quaternion to equivalent rotation matrix.
* @param[in] pInputQuaternions points to an array of normalized quaternions
* @param[out] pOutputRotations points to an array of 3x3 rotations (in row order)
* @param[in] nbQuaternions in the array
*
* <b>Format of rotation matrix</b>
* \par
* The quaternion a + ib + jc + kd is converted into rotation matrix:
* a^2 + b^2 - c^2 - d^2 2bc - 2ad 2bd + 2ac
* 2bc + 2ad a^2 - b^2 + c^2 - d^2 2cd - 2ab
* 2bd - 2ac 2cd + 2ab a^2 - b^2 - c^2 + d^2
*
* Rotation matrix is saved in row order : R00 R01 R02 R10 R11 R12 R20 R21 R22
*/
void arm_quaternion2rotation_f32(const float32_t *pInputQuaternions,
float32_t *pOutputRotations,
uint32_t nbQuaternions);
/**
* @brief Conversion of a rotation matrix to equivalent quaternion.
* @param[in] pInputRotations points to an array 3x3 rotation matrix (in row order)
* @param[out] pOutputQuaternions points to an array of quaternions
* @param[in] nbQuaternions in the array
*/
void arm_rotation2quaternion_f32(const float32_t *pInputRotations,
float32_t *pOutputQuaternions,
uint32_t nbQuaternions);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _QUATERNION_MATH_FUNCTIONS_H_ */

988
sdk/CMSIS/DSP/Include/dsp/statistics_functions.h Normal file
View file

@ -0,0 +1,988 @@
/******************************************************************************
* @file statistics_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.1
* @date 14 July 2022
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef STATISTICS_FUNCTIONS_H_
#define STATISTICS_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/basic_math_functions.h"
#include "dsp/fast_math_functions.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupStats Statistics Functions
*/
/**
* @brief Computation of the LogSumExp
*
* In probabilistic computations, the dynamic of the probability values can be very
* wide because they come from gaussian functions.
* To avoid underflow and overflow issues, the values are represented by their log.
* In this representation, multiplying the original exp values is easy : their logs are added.
* But adding the original exp values is requiring some special handling and it is the
* goal of the LogSumExp function.
*
* If the values are x1...xn, the function is computing:
*
* ln(exp(x1) + ... + exp(xn)) and the computation is done in such a way that
* rounding issues are minimised.
*
* The max xm of the values is extracted and the function is computing:
* xm + ln(exp(x1 - xm) + ... + exp(xn - xm))
*
* @param[in] *in Pointer to an array of input values.
* @param[in] blockSize Number of samples in the input array.
* @return LogSumExp
*
*/
float32_t arm_logsumexp_f32(const float32_t *in, uint32_t blockSize);
/**
* @brief Dot product with log arithmetic
*
* Vectors are containing the log of the samples
*
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] blockSize number of samples in each vector
* @param[in] pTmpBuffer temporary buffer of length blockSize
* @return The log of the dot product .
*
*/
float32_t arm_logsumexp_dot_prod_f32(const float32_t * pSrcA,
const float32_t * pSrcB,
uint32_t blockSize,
float32_t *pTmpBuffer);
/**
* @brief Entropy
*
* @param[in] pSrcA Array of input values.
* @param[in] blockSize Number of samples in the input array.
* @return Entropy -Sum(p ln p)
*
*/
float32_t arm_entropy_f32(const float32_t * pSrcA,uint32_t blockSize);
/**
* @brief Entropy
*
* @param[in] pSrcA Array of input values.
* @param[in] blockSize Number of samples in the input array.
* @return Entropy -Sum(p ln p)
*
*/
float64_t arm_entropy_f64(const float64_t * pSrcA, uint32_t blockSize);
/**
* @brief Kullback-Leibler
*
* @param[in] pSrcA Pointer to an array of input values for probability distribution A.
* @param[in] pSrcB Pointer to an array of input values for probability distribution B.
* @param[in] blockSize Number of samples in the input array.
* @return Kullback-Leibler Divergence D(A || B)
*
*/
float32_t arm_kullback_leibler_f32(const float32_t * pSrcA
,const float32_t * pSrcB
,uint32_t blockSize);
/**
* @brief Kullback-Leibler
*
* @param[in] pSrcA Pointer to an array of input values for probability distribution A.
* @param[in] pSrcB Pointer to an array of input values for probability distribution B.
* @param[in] blockSize Number of samples in the input array.
* @return Kullback-Leibler Divergence D(A || B)
*
*/
float64_t arm_kullback_leibler_f64(const float64_t * pSrcA,
const float64_t * pSrcB,
uint32_t blockSize);
/**
* @brief Sum of the squares of the elements of a Q31 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_power_q31(
const q31_t * pSrc,
uint32_t blockSize,
q63_t * pResult);
/**
* @brief Sum of the squares of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_power_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult);
/**
* @brief Sum of the squares of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_power_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult);
/**
* @brief Sum of the squares of the elements of a Q15 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_power_q15(
const q15_t * pSrc,
uint32_t blockSize,
q63_t * pResult);
/**
* @brief Sum of the squares of the elements of a Q7 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_power_q7(
const q7_t * pSrc,
uint32_t blockSize,
q31_t * pResult);
/**
* @brief Mean value of a Q7 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_mean_q7(
const q7_t * pSrc,
uint32_t blockSize,
q7_t * pResult);
/**
* @brief Mean value of a Q15 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_mean_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult);
/**
* @brief Mean value of a Q31 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_mean_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult);
/**
* @brief Mean value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_mean_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult);
/**
* @brief Mean value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_mean_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult);
/**
* @brief Variance of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_var_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult);
/**
* @brief Variance of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_var_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult);
/**
* @brief Variance of the elements of a Q31 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_var_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult);
/**
* @brief Variance of the elements of a Q15 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_var_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult);
/**
* @brief Root Mean Square of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_rms_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult);
/**
* @brief Root Mean Square of the elements of a Q31 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_rms_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult);
/**
* @brief Root Mean Square of the elements of a Q15 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_rms_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult);
/**
* @brief Standard deviation of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_std_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult);
/**
* @brief Standard deviation of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_std_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult);
/**
* @brief Standard deviation of the elements of a Q31 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_std_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult);
/**
* @brief Standard deviation of the elements of a Q15 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_std_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult);
/**
* @brief Minimum value of a Q7 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[in] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_min_q7(
const q7_t * pSrc,
uint32_t blockSize,
q7_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a Q7 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[in] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_absmin_q7(
const q7_t * pSrc,
uint32_t blockSize,
q7_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a Q7 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
*/
void arm_absmin_no_idx_q7(
const q7_t * pSrc,
uint32_t blockSize,
q7_t * pResult);
/**
* @brief Minimum value of a Q15 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[in] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_min_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a Q15 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[in] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_absmin_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a Q15 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
*/
void arm_absmin_no_idx_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult);
/**
* @brief Minimum value of a Q31 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[out] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_min_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a Q31 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[out] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_absmin_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a Q31 vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
*/
void arm_absmin_no_idx_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult);
/**
* @brief Minimum value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[out] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_min_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[out] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_absmin_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
*/
void arm_absmin_no_idx_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult);
/**
* @brief Minimum value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[out] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_min_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[out] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_absmin_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
*/
void arm_absmin_no_idx_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult);
/**
* @brief Maximum value of a Q7 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_max_q7(
const q7_t * pSrc,
uint32_t blockSize,
q7_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a Q7 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_absmax_q7(
const q7_t * pSrc,
uint32_t blockSize,
q7_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a Q7 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
*/
void arm_absmax_no_idx_q7(
const q7_t * pSrc,
uint32_t blockSize,
q7_t * pResult);
/**
* @brief Maximum value of a Q15 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_max_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a Q15 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_absmax_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a Q15 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
*/
void arm_absmax_no_idx_q15(
const q15_t * pSrc,
uint32_t blockSize,
q15_t * pResult);
/**
* @brief Maximum value of a Q31 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_max_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a Q31 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_absmax_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a Q31 vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
*/
void arm_absmax_no_idx_q31(
const q31_t * pSrc,
uint32_t blockSize,
q31_t * pResult);
/**
* @brief Maximum value of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_max_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_absmax_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
*/
void arm_absmax_no_idx_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult);
/**
* @brief Maximum value of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_max_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_absmax_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
*/
void arm_absmax_no_idx_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult);
/**
@brief Maximum value of a floating-point vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult maximum value returned here
*/
void arm_max_no_idx_f32(
const float32_t *pSrc,
uint32_t blockSize,
float32_t *pResult);
/**
@brief Minimum value of a floating-point vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult minimum value returned here
*/
void arm_min_no_idx_f32(
const float32_t *pSrc,
uint32_t blockSize,
float32_t *pResult);
/**
@brief Maximum value of a floating-point vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult maximum value returned here
*/
void arm_max_no_idx_f64(
const float64_t *pSrc,
uint32_t blockSize,
float64_t *pResult);
/**
@brief Maximum value of a q31 vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult maximum value returned here
*/
void arm_max_no_idx_q31(
const q31_t *pSrc,
uint32_t blockSize,
q31_t *pResult);
/**
@brief Maximum value of a q15 vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult maximum value returned here
*/
void arm_max_no_idx_q15(
const q15_t *pSrc,
uint32_t blockSize,
q15_t *pResult);
/**
@brief Maximum value of a q7 vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult maximum value returned here
*/
void arm_max_no_idx_q7(
const q7_t *pSrc,
uint32_t blockSize,
q7_t *pResult);
/**
@brief Minimum value of a floating-point vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult minimum value returned here
*/
void arm_min_no_idx_f64(
const float64_t *pSrc,
uint32_t blockSize,
float64_t *pResult);
/**
@brief Minimum value of a q31 vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult minimum value returned here
*/
void arm_min_no_idx_q31(
const q31_t *pSrc,
uint32_t blockSize,
q31_t *pResult);
/**
@brief Minimum value of a q15 vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult minimum value returned here
*/
void arm_min_no_idx_q15(
const q15_t *pSrc,
uint32_t blockSize,
q15_t *pResult);
/**
@brief Minimum value of a q7 vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult minimum value returned here
*/
void arm_min_no_idx_q7(
const q7_t *pSrc,
uint32_t blockSize,
q7_t *pResult);
/**
@brief Mean square error between two Q7 vectors.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult mean square error
*/
void arm_mse_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
uint32_t blockSize,
q7_t * pResult);
/**
@brief Mean square error between two Q15 vectors.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult mean square error
*/
void arm_mse_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
uint32_t blockSize,
q15_t * pResult);
/**
@brief Mean square error between two Q31 vectors.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult mean square error
*/
void arm_mse_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
uint32_t blockSize,
q31_t * pResult);
/**
@brief Mean square error between two single precision float vectors.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult mean square error
*/
void arm_mse_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
uint32_t blockSize,
float32_t * pResult);
/**
@brief Mean square error between two double precision float vectors.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult mean square error
*/
void arm_mse_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
uint32_t blockSize,
float64_t * pResult);
/**
* @brief Accumulation value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_accumulate_f32(
const float32_t * pSrc,
uint32_t blockSize,
float32_t * pResult);
/**
* @brief Accumulation value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_accumulate_f64(
const float64_t * pSrc,
uint32_t blockSize,
float64_t * pResult);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _STATISTICS_FUNCTIONS_H_ */

274
sdk/CMSIS/DSP/Include/dsp/statistics_functions_f16.h Normal file
View file

@ -0,0 +1,274 @@
/******************************************************************************
* @file statistics_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.1
* @date 14 July 2022
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef STATISTICS_FUNCTIONS_F16_H_
#define STATISTICS_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/basic_math_functions_f16.h"
#include "dsp/fast_math_functions_f16.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Sum of the squares of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_power_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult);
/**
* @brief Mean value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_mean_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult);
/**
* @brief Variance of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_var_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult);
/**
* @brief Root Mean Square of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_rms_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult);
/**
* @brief Standard deviation of the elements of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_std_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult);
/**
* @brief Minimum value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[out] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_min_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
* @param[out] pIndex is the array index of the minimum value in the input buffer.
*/
void arm_absmin_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_max_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult,
uint32_t * pIndex);
/**
* @brief Maximum value of absolute values of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
* @param[out] pIndex index of maximum value returned here
*/
void arm_absmax_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult,
uint32_t * pIndex);
/**
* @brief Minimum value of absolute values of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output pointer
*/
void arm_absmin_no_idx_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult);
/**
* @brief Maximum value of a floating-point vector.
* @param[in] pSrc points to the input buffer
* @param[in] blockSize length of the input vector
* @param[out] pResult maximum value returned here
*/
void arm_absmax_no_idx_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult);
/**
* @brief Entropy
*
* @param[in] pSrcA Array of input values.
* @param[in] blockSize Number of samples in the input array.
* @return Entropy -Sum(p ln p)
*/
float16_t arm_entropy_f16(const float16_t * pSrcA,uint32_t blockSize);
float16_t arm_logsumexp_f16(const float16_t *in, uint32_t blockSize);
/**
* @brief Dot product with log arithmetic
*
* Vectors are containing the log of the samples
*
* @param[in] pSrcA points to the first input vector
* @param[in] pSrcB points to the second input vector
* @param[in] blockSize number of samples in each vector
* @param[in] pTmpBuffer temporary buffer of length blockSize
* @return The log of the dot product .
*/
float16_t arm_logsumexp_dot_prod_f16(const float16_t * pSrcA,
const float16_t * pSrcB,
uint32_t blockSize,
float16_t *pTmpBuffer);
/**
* @brief Kullback-Leibler
*
* @param[in] pSrcA Pointer to an array of input values for probability distribution A.
* @param[in] pSrcB Pointer to an array of input values for probability distribution B.
* @param[in] blockSize Number of samples in the input array.
* @return Kullback-Leibler Divergence D(A || B)
*/
float16_t arm_kullback_leibler_f16(const float16_t * pSrcA
,const float16_t * pSrcB
,uint32_t blockSize);
/**
@brief Maximum value of a floating-point vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult maximum value returned here
*/
void arm_max_no_idx_f16(
const float16_t *pSrc,
uint32_t blockSize,
float16_t *pResult);
/**
@brief Minimum value of a floating-point vector.
@param[in] pSrc points to the input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult minimum value returned here
*/
void arm_min_no_idx_f16(
const float16_t *pSrc,
uint32_t blockSize,
float16_t *pResult);
/**
@brief Mean square error between two half precision float vectors.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[in] blockSize number of samples in input vector
@param[out] pResult mean square error
*/
void arm_mse_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
uint32_t blockSize,
float16_t * pResult);
/**
* @brief Sum value of a floating-point vector.
* @param[in] pSrc is input pointer
* @param[in] blockSize is the number of samples to process
* @param[out] pResult is output value.
*/
void arm_accumulate_f16(
const float16_t * pSrc,
uint32_t blockSize,
float16_t * pResult);
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _STATISTICS_FUNCTIONS_F16_H_ */

539
sdk/CMSIS/DSP/Include/dsp/support_functions.h Normal file
View file

@ -0,0 +1,539 @@
/******************************************************************************
* @file support_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.1
* @date 18 August 2022
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef SUPPORT_FUNCTIONS_H_
#define SUPPORT_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupSupport Support Functions
*/
/**
* @brief Converts the elements of the 64 bit floating-point vector to floating-point vector.
* @param[in] pSrc points to the floating-point 64 input vector
* @param[out] pDst points to the floating-point output vector
* @param[in] blockSize length of the input vector
*/
void arm_f64_to_float(
const float64_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the 64 bit floating-point vector to Q31 vector.
* @param[in] pSrc points to the floating-point 64 input vector
* @param[out] pDst points to the Q31 output vector
* @param[in] blockSize length of the input vector
*/
void arm_f64_to_q31(
const float64_t * pSrc,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the 64 bit floating-point vector to Q15 vector.
* @param[in] pSrc points to the floating-point 64 input vector
* @param[out] pDst points to the Q15 output vector
* @param[in] blockSize length of the input vector
*/
void arm_f64_to_q15(
const float64_t * pSrc,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the 64 bit floating-point vector to Q7 vector.
* @param[in] pSrc points to the floating-point 64 input vector
* @param[out] pDst points to the Q7 output vector
* @param[in] blockSize length of the input vector
*/
void arm_f64_to_q7(
const float64_t * pSrc,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the floating-point vector to 64 bit floating-point vector.
* @param[in] pSrc points to the floating-point input vector
* @param[out] pDst points to the 64 bit floating-point output vector
* @param[in] blockSize length of the input vector
*/
void arm_float_to_f64(
const float32_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the floating-point vector to Q31 vector.
* @param[in] pSrc points to the floating-point input vector
* @param[out] pDst points to the Q31 output vector
* @param[in] blockSize length of the input vector
*/
void arm_float_to_q31(
const float32_t * pSrc,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the floating-point vector to Q15 vector.
* @param[in] pSrc points to the floating-point input vector
* @param[out] pDst points to the Q15 output vector
* @param[in] blockSize length of the input vector
*/
void arm_float_to_q15(
const float32_t * pSrc,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the floating-point vector to Q7 vector.
* @param[in] pSrc points to the floating-point input vector
* @param[out] pDst points to the Q7 output vector
* @param[in] blockSize length of the input vector
*/
void arm_float_to_q7(
const float32_t * pSrc,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q31 vector to 64 bit floating-point vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q31_to_f64(
const q31_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q31 vector to floating-point vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q31_to_float(
const q31_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q31 vector to Q15 vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q31_to_q15(
const q31_t * pSrc,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q31 vector to Q7 vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q31_to_q7(
const q31_t * pSrc,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q15 vector to 64 bit floating-point vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q15_to_f64(
const q15_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q15 vector to floating-point vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q15_to_float(
const q15_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q15 vector to Q31 vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q15_to_q31(
const q15_t * pSrc,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q15 vector to Q7 vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q15_to_q7(
const q15_t * pSrc,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q7 vector to 64 bit floating-point vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q7_to_f64(
const q7_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q7 vector to floating-point vector.
* @param[in] pSrc is input pointer
* @param[out] pDst is output pointer
* @param[in] blockSize is the number of samples to process
*/
void arm_q7_to_float(
const q7_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q7 vector to Q31 vector.
* @param[in] pSrc input pointer
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_q7_to_q31(
const q7_t * pSrc,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Converts the elements of the Q7 vector to Q15 vector.
* @param[in] pSrc input pointer
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_q7_to_q15(
const q7_t * pSrc,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Struct for specifying sorting algorithm
*/
typedef enum
{
ARM_SORT_BITONIC = 0,
/**< Bitonic sort */
ARM_SORT_BUBBLE = 1,
/**< Bubble sort */
ARM_SORT_HEAP = 2,
/**< Heap sort */
ARM_SORT_INSERTION = 3,
/**< Insertion sort */
ARM_SORT_QUICK = 4,
/**< Quick sort */
ARM_SORT_SELECTION = 5
/**< Selection sort */
} arm_sort_alg;
/**
* @brief Struct for specifying sorting algorithm
*/
typedef enum
{
ARM_SORT_DESCENDING = 0,
/**< Descending order (9 to 0) */
ARM_SORT_ASCENDING = 1
/**< Ascending order (0 to 9) */
} arm_sort_dir;
/**
* @brief Instance structure for the sorting algorithms.
*/
typedef struct
{
arm_sort_alg alg; /**< Sorting algorithm selected */
arm_sort_dir dir; /**< Sorting order (direction) */
} arm_sort_instance_f32;
/**
* @param[in] S points to an instance of the sorting structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data.
* @param[in] blockSize number of samples to process.
*/
void arm_sort_f32(
const arm_sort_instance_f32 * S,
float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @param[in,out] S points to an instance of the sorting structure.
* @param[in] alg Selected algorithm.
* @param[in] dir Sorting order.
*/
void arm_sort_init_f32(
arm_sort_instance_f32 * S,
arm_sort_alg alg,
arm_sort_dir dir);
/**
* @brief Instance structure for the sorting algorithms.
*/
typedef struct
{
arm_sort_dir dir; /**< Sorting order (direction) */
float32_t * buffer; /**< Working buffer */
} arm_merge_sort_instance_f32;
/**
* @param[in] S points to an instance of the sorting structure.
* @param[in,out] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data
* @param[in] blockSize number of samples to process.
*/
void arm_merge_sort_f32(
const arm_merge_sort_instance_f32 * S,
float32_t *pSrc,
float32_t *pDst,
uint32_t blockSize);
/**
* @param[in,out] S points to an instance of the sorting structure.
* @param[in] dir Sorting order.
* @param[in] buffer Working buffer.
*/
void arm_merge_sort_init_f32(
arm_merge_sort_instance_f32 * S,
arm_sort_dir dir,
float32_t * buffer);
/**
* @brief Copies the elements of a floating-point vector.
* @param[in] pSrc input pointer
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_copy_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Copies the elements of a floating-point vector.
* @param[in] pSrc input pointer
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_copy_f64(
const float64_t * pSrc,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Copies the elements of a Q7 vector.
* @param[in] pSrc input pointer
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_copy_q7(
const q7_t * pSrc,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Copies the elements of a Q15 vector.
* @param[in] pSrc input pointer
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_copy_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Copies the elements of a Q31 vector.
* @param[in] pSrc input pointer
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_copy_q31(
const q31_t * pSrc,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Fills a constant value into a floating-point vector.
* @param[in] value input value to be filled
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_fill_f32(
float32_t value,
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Fills a constant value into a floating-point vector.
* @param[in] value input value to be filled
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_fill_f64(
float64_t value,
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Fills a constant value into a Q7 vector.
* @param[in] value input value to be filled
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_fill_q7(
q7_t value,
q7_t * pDst,
uint32_t blockSize);
/**
* @brief Fills a constant value into a Q15 vector.
* @param[in] value input value to be filled
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_fill_q15(
q15_t value,
q15_t * pDst,
uint32_t blockSize);
/**
* @brief Fills a constant value into a Q31 vector.
* @param[in] value input value to be filled
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_fill_q31(
q31_t value,
q31_t * pDst,
uint32_t blockSize);
/**
* @brief Weighted average
*
*
* @param[in] *in Array of input values.
* @param[in] *weigths Weights
* @param[in] blockSize Number of samples in the input array.
* @return Weighted average
*
*/
float32_t arm_weighted_average_f32(const float32_t *in
, const float32_t *weigths
, uint32_t blockSize);
/**
* @brief Barycenter
*
*
* @param[in] in List of vectors
* @param[in] weights Weights of the vectors
* @param[out] out Barycenter
* @param[in] nbVectors Number of vectors
* @param[in] vecDim Dimension of space (vector dimension)
*
*/
void arm_barycenter_f32(const float32_t *in
, const float32_t *weights
, float32_t *out
, uint32_t nbVectors
, uint32_t vecDim);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _SUPPORT_FUNCTIONS_H_ */

204
sdk/CMSIS/DSP/Include/dsp/support_functions_f16.h Normal file
View file

@ -0,0 +1,204 @@
/******************************************************************************
* @file support_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.1
* @date 18 August 2022
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef SUPPORT_FUNCTIONS_F16_H_
#define SUPPORT_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Copies the elements of a floating-point vector.
* @param[in] pSrc input pointer
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_copy_f16(const float16_t * pSrc, float16_t * pDst, uint32_t blockSize);
/**
* @brief Fills a constant value into a floating-point vector.
* @param[in] value input value to be filled
* @param[out] pDst output pointer
* @param[in] blockSize number of samples to process
*/
void arm_fill_f16(float16_t value, float16_t * pDst, uint32_t blockSize);
/**
* @brief Converts the elements of the floating-point vector to Q31 vector.
* @param[in] pSrc points to the f16 input vector
* @param[out] pDst points to the q15 output vector
* @param[in] blockSize length of the input vector
*/
void arm_f16_to_q15(const float16_t * pSrc, q15_t * pDst, uint32_t blockSize);
/**
* @brief Converts the elements of the floating-point vector to Q31 vector.
* @param[in] pSrc points to the q15 input vector
* @param[out] pDst points to the f16 output vector
* @param[in] blockSize length of the input vector
*/
void arm_q15_to_f16(const q15_t * pSrc, float16_t * pDst, uint32_t blockSize);
/**
* @brief Converts the elements of the 64 bit floating-point vector to 16 bit floating-point vector.
* @param[in] pSrc points to the f64 input vector
* @param[out] pDst points to the f16 output vector
* @param[in] blockSize length of the input vector
*/
void arm_f64_to_f16(const float64_t * pSrc, float16_t * pDst, uint32_t blockSize);
/**
* @brief Converts the elements of the 16 bit floating-point vector to 64 bit floating-point vector.
* @param[in] pSrc points to the f16 input vector
* @param[out] pDst points to the f64 output vector
* @param[in] blockSize length of the input vector
*/
void arm_f16_to_f64(const float16_t * pSrc, float64_t * pDst, uint32_t blockSize);
/**
* @brief Converts the elements of the floating-point vector to Q31 vector.
* @param[in] pSrc points to the f32 input vector
* @param[out] pDst points to the f16 output vector
* @param[in] blockSize length of the input vector
*/
void arm_float_to_f16(const float32_t * pSrc, float16_t * pDst, uint32_t blockSize);
/**
* @brief Converts the elements of the floating-point vector to Q31 vector.
* @param[in] pSrc points to the f16 input vector
* @param[out] pDst points to the f32 output vector
* @param[in] blockSize length of the input vector
*/
void arm_f16_to_float(const float16_t * pSrc, float32_t * pDst, uint32_t blockSize);
/**
* @brief Weighted average
* @param[in] *in Array of input values.
* @param[in] *weigths Weights
* @param[in] blockSize Number of samples in the input array.
* @return Weighted average
*/
float16_t arm_weighted_average_f16(const float16_t *in
, const float16_t *weigths
, uint32_t blockSize);
/**
* @brief Barycenter
* @param[in] in List of vectors
* @param[in] weights Weights of the vectors
* @param[out] out Barycenter
* @param[in] nbVectors Number of vectors
* @param[in] vecDim Dimension of space (vector dimension)
*/
void arm_barycenter_f16(const float16_t *in
, const float16_t *weights
, float16_t *out
, uint32_t nbVectors
, uint32_t vecDim);
/**
@ingroup groupSupport
*/
/**
* @defgroup typecast Typecasting
*/
/**
@addtogroup typecast
@{
*/
/**
* @brief Interpret a f16 as an s16 value
* @param[in] x input value.
* @return return value.
*
* @par Description
* It is a typecast. No conversion of the float to int is done.
* The memcpy will be optimized out by the compiler.
* memcpy is used to prevent type punning issues.
* With gcc, -fno-builtins MUST not be used or the
* memcpy will not be optimized out.
*/
__STATIC_INLINE int16_t arm_typecast_s16_f16(float16_t x)
{
int16_t res;
res=*(int16_t*)memcpy((char*)&res,(char*)&x,sizeof(float16_t));
return(res);
}
/**
* @brief Interpret an s16 as an f16 value
* @param[in] x input value.
* @return return value.
*
* @par Description
* It is a typecast. No conversion of the int to float is done.
* The memcpy will be optimized out by the compiler.
* memcpy is used to prevent type punning issues.
* With gcc, -fno-builtins MUST not be used or the
* memcpy will not be optimized out.
*/
__STATIC_INLINE float16_t arm_typecast_f16_s16(int16_t x)
{
float16_t res;
res=*(float16_t*)memcpy((char*)&res,(char*)&x,sizeof(int16_t));
return(res);
}
/**
@} end of typecast group
*/
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _SUPPORT_FUNCTIONS_F16_H_ */

46
sdk/CMSIS/DSP/Include/dsp/svm_defines.h Normal file
View file

@ -0,0 +1,46 @@
/******************************************************************************
* @file svm_defines.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
*
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef SVM_DEFINES_H_
#define SVM_DEFINES_H_
/**
* @brief Struct for specifying SVM Kernel
*/
typedef enum
{
ARM_ML_KERNEL_LINEAR = 0,
/**< Linear kernel */
ARM_ML_KERNEL_POLYNOMIAL = 1,
/**< Polynomial kernel */
ARM_ML_KERNEL_RBF = 2,
/**< Radial Basis Function kernel */
ARM_ML_KERNEL_SIGMOID = 3
/**< Sigmoid kernel */
} arm_ml_kernel_type;
#endif

280
sdk/CMSIS/DSP/Include/dsp/svm_functions.h Normal file
View file

@ -0,0 +1,280 @@
/******************************************************************************
* @file svm_functions.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef SVM_FUNCTIONS_H_
#define SVM_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/svm_defines.h"
#ifdef __cplusplus
extern "C"
{
#endif
#define STEP(x) (x) <= 0 ? 0 : 1
/**
* @defgroup groupSVM SVM Functions
* This set of functions is implementing SVM classification on 2 classes.
* The training must be done from scikit-learn. The parameters can be easily
* generated from the scikit-learn object. Some examples are given in
* DSP/Testing/PatternGeneration/SVM.py
*
* If more than 2 classes are needed, the functions in this folder
* will have to be used, as building blocks, to do multi-class classification.
*
* No multi-class classification is provided in this SVM folder.
*
*/
/**
* @brief Integer exponentiation
* @param[in] x value
* @param[in] nb integer exponent >= 1
* @return x^nb
*/
__STATIC_INLINE float32_t arm_exponent_f32(float32_t x, int32_t nb)
{
float32_t r = x;
nb --;
while(nb > 0)
{
r = r * x;
nb--;
}
return(r);
}
/**
* @brief Instance structure for linear SVM prediction function.
*/
typedef struct
{
uint32_t nbOfSupportVectors; /**< Number of support vectors */
uint32_t vectorDimension; /**< Dimension of vector space */
float32_t intercept; /**< Intercept */
const float32_t *dualCoefficients; /**< Dual coefficients */
const float32_t *supportVectors; /**< Support vectors */
const int32_t *classes; /**< The two SVM classes */
} arm_svm_linear_instance_f32;
/**
* @brief Instance structure for polynomial SVM prediction function.
*/
typedef struct
{
uint32_t nbOfSupportVectors; /**< Number of support vectors */
uint32_t vectorDimension; /**< Dimension of vector space */
float32_t intercept; /**< Intercept */
const float32_t *dualCoefficients; /**< Dual coefficients */
const float32_t *supportVectors; /**< Support vectors */
const int32_t *classes; /**< The two SVM classes */
int32_t degree; /**< Polynomial degree */
float32_t coef0; /**< Polynomial constant */
float32_t gamma; /**< Gamma factor */
} arm_svm_polynomial_instance_f32;
/**
* @brief Instance structure for rbf SVM prediction function.
*/
typedef struct
{
uint32_t nbOfSupportVectors; /**< Number of support vectors */
uint32_t vectorDimension; /**< Dimension of vector space */
float32_t intercept; /**< Intercept */
const float32_t *dualCoefficients; /**< Dual coefficients */
const float32_t *supportVectors; /**< Support vectors */
const int32_t *classes; /**< The two SVM classes */
float32_t gamma; /**< Gamma factor */
} arm_svm_rbf_instance_f32;
/**
* @brief Instance structure for sigmoid SVM prediction function.
*/
typedef struct
{
uint32_t nbOfSupportVectors; /**< Number of support vectors */
uint32_t vectorDimension; /**< Dimension of vector space */
float32_t intercept; /**< Intercept */
const float32_t *dualCoefficients; /**< Dual coefficients */
const float32_t *supportVectors; /**< Support vectors */
const int32_t *classes; /**< The two SVM classes */
float32_t coef0; /**< Independent constant */
float32_t gamma; /**< Gamma factor */
} arm_svm_sigmoid_instance_f32;
/**
* @brief SVM linear instance init function
* @param[in] S Parameters for SVM functions
* @param[in] nbOfSupportVectors Number of support vectors
* @param[in] vectorDimension Dimension of vector space
* @param[in] intercept Intercept
* @param[in] dualCoefficients Array of dual coefficients
* @param[in] supportVectors Array of support vectors
* @param[in] classes Array of 2 classes ID
*/
void arm_svm_linear_init_f32(arm_svm_linear_instance_f32 *S,
uint32_t nbOfSupportVectors,
uint32_t vectorDimension,
float32_t intercept,
const float32_t *dualCoefficients,
const float32_t *supportVectors,
const int32_t *classes);
/**
* @brief SVM linear prediction
* @param[in] S Pointer to an instance of the linear SVM structure.
* @param[in] in Pointer to input vector
* @param[out] pResult Decision value
*/
void arm_svm_linear_predict_f32(const arm_svm_linear_instance_f32 *S,
const float32_t * in,
int32_t * pResult);
/**
* @brief SVM polynomial instance init function
* @param[in] S points to an instance of the polynomial SVM structure.
* @param[in] nbOfSupportVectors Number of support vectors
* @param[in] vectorDimension Dimension of vector space
* @param[in] intercept Intercept
* @param[in] dualCoefficients Array of dual coefficients
* @param[in] supportVectors Array of support vectors
* @param[in] classes Array of 2 classes ID
* @param[in] degree Polynomial degree
* @param[in] coef0 coeff0 (scikit-learn terminology)
* @param[in] gamma gamma (scikit-learn terminology)
*/
void arm_svm_polynomial_init_f32(arm_svm_polynomial_instance_f32 *S,
uint32_t nbOfSupportVectors,
uint32_t vectorDimension,
float32_t intercept,
const float32_t *dualCoefficients,
const float32_t *supportVectors,
const int32_t *classes,
int32_t degree,
float32_t coef0,
float32_t gamma
);
/**
* @brief SVM polynomial prediction
* @param[in] S Pointer to an instance of the polynomial SVM structure.
* @param[in] in Pointer to input vector
* @param[out] pResult Decision value
*/
void arm_svm_polynomial_predict_f32(const arm_svm_polynomial_instance_f32 *S,
const float32_t * in,
int32_t * pResult);
/**
* @brief SVM radial basis function instance init function
* @param[in] S points to an instance of the polynomial SVM structure.
* @param[in] nbOfSupportVectors Number of support vectors
* @param[in] vectorDimension Dimension of vector space
* @param[in] intercept Intercept
* @param[in] dualCoefficients Array of dual coefficients
* @param[in] supportVectors Array of support vectors
* @param[in] classes Array of 2 classes ID
* @param[in] gamma gamma (scikit-learn terminology)
*/
void arm_svm_rbf_init_f32(arm_svm_rbf_instance_f32 *S,
uint32_t nbOfSupportVectors,
uint32_t vectorDimension,
float32_t intercept,
const float32_t *dualCoefficients,
const float32_t *supportVectors,
const int32_t *classes,
float32_t gamma
);
/**
* @brief SVM rbf prediction
* @param[in] S Pointer to an instance of the rbf SVM structure.
* @param[in] in Pointer to input vector
* @param[out] pResult decision value
*/
void arm_svm_rbf_predict_f32(const arm_svm_rbf_instance_f32 *S,
const float32_t * in,
int32_t * pResult);
/**
* @brief SVM sigmoid instance init function
* @param[in] S points to an instance of the rbf SVM structure.
* @param[in] nbOfSupportVectors Number of support vectors
* @param[in] vectorDimension Dimension of vector space
* @param[in] intercept Intercept
* @param[in] dualCoefficients Array of dual coefficients
* @param[in] supportVectors Array of support vectors
* @param[in] classes Array of 2 classes ID
* @param[in] coef0 coeff0 (scikit-learn terminology)
* @param[in] gamma gamma (scikit-learn terminology)
*/
void arm_svm_sigmoid_init_f32(arm_svm_sigmoid_instance_f32 *S,
uint32_t nbOfSupportVectors,
uint32_t vectorDimension,
float32_t intercept,
const float32_t *dualCoefficients,
const float32_t *supportVectors,
const int32_t *classes,
float32_t coef0,
float32_t gamma
);
/**
* @brief SVM sigmoid prediction
* @param[in] S Pointer to an instance of the rbf SVM structure.
* @param[in] in Pointer to input vector
* @param[out] pResult Decision value
*/
void arm_svm_sigmoid_predict_f32(const arm_svm_sigmoid_instance_f32 *S,
const float32_t * in,
int32_t * pResult);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _SVM_FUNCTIONS_H_ */

265
sdk/CMSIS/DSP/Include/dsp/svm_functions_f16.h Normal file
View file

@ -0,0 +1,265 @@
/******************************************************************************
* @file svm_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef SVM_FUNCTIONS_F16_H_
#define SVM_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#include "dsp/svm_defines.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
#define STEP(x) (x) <= 0 ? 0 : 1
/**
* @defgroup groupSVM SVM Functions
* This set of functions is implementing SVM classification on 2 classes.
* The training must be done from scikit-learn. The parameters can be easily
* generated from the scikit-learn object. Some examples are given in
* DSP/Testing/PatternGeneration/SVM.py
*
* If more than 2 classes are needed, the functions in this folder
* will have to be used, as building blocks, to do multi-class classification.
*
* No multi-class classification is provided in this SVM folder.
*
*/
/**
* @brief Instance structure for linear SVM prediction function.
*/
typedef struct
{
uint32_t nbOfSupportVectors; /**< Number of support vectors */
uint32_t vectorDimension; /**< Dimension of vector space */
float16_t intercept; /**< Intercept */
const float16_t *dualCoefficients; /**< Dual coefficients */
const float16_t *supportVectors; /**< Support vectors */
const int32_t *classes; /**< The two SVM classes */
} arm_svm_linear_instance_f16;
/**
* @brief Instance structure for polynomial SVM prediction function.
*/
typedef struct
{
uint32_t nbOfSupportVectors; /**< Number of support vectors */
uint32_t vectorDimension; /**< Dimension of vector space */
float16_t intercept; /**< Intercept */
const float16_t *dualCoefficients; /**< Dual coefficients */
const float16_t *supportVectors; /**< Support vectors */
const int32_t *classes; /**< The two SVM classes */
int32_t degree; /**< Polynomial degree */
float16_t coef0; /**< Polynomial constant */
float16_t gamma; /**< Gamma factor */
} arm_svm_polynomial_instance_f16;
/**
* @brief Instance structure for rbf SVM prediction function.
*/
typedef struct
{
uint32_t nbOfSupportVectors; /**< Number of support vectors */
uint32_t vectorDimension; /**< Dimension of vector space */
float16_t intercept; /**< Intercept */
const float16_t *dualCoefficients; /**< Dual coefficients */
const float16_t *supportVectors; /**< Support vectors */
const int32_t *classes; /**< The two SVM classes */
float16_t gamma; /**< Gamma factor */
} arm_svm_rbf_instance_f16;
/**
* @brief Instance structure for sigmoid SVM prediction function.
*/
typedef struct
{
uint32_t nbOfSupportVectors; /**< Number of support vectors */
uint32_t vectorDimension; /**< Dimension of vector space */
float16_t intercept; /**< Intercept */
const float16_t *dualCoefficients; /**< Dual coefficients */
const float16_t *supportVectors; /**< Support vectors */
const int32_t *classes; /**< The two SVM classes */
float16_t coef0; /**< Independent constant */
float16_t gamma; /**< Gamma factor */
} arm_svm_sigmoid_instance_f16;
/**
* @brief SVM linear instance init function
* @param[in] S Parameters for SVM functions
* @param[in] nbOfSupportVectors Number of support vectors
* @param[in] vectorDimension Dimension of vector space
* @param[in] intercept Intercept
* @param[in] dualCoefficients Array of dual coefficients
* @param[in] supportVectors Array of support vectors
* @param[in] classes Array of 2 classes ID
*/
void arm_svm_linear_init_f16(arm_svm_linear_instance_f16 *S,
uint32_t nbOfSupportVectors,
uint32_t vectorDimension,
float16_t intercept,
const float16_t *dualCoefficients,
const float16_t *supportVectors,
const int32_t *classes);
/**
* @brief SVM linear prediction
* @param[in] S Pointer to an instance of the linear SVM structure.
* @param[in] in Pointer to input vector
* @param[out] pResult Decision value
*/
void arm_svm_linear_predict_f16(const arm_svm_linear_instance_f16 *S,
const float16_t * in,
int32_t * pResult);
/**
* @brief SVM polynomial instance init function
* @param[in] S points to an instance of the polynomial SVM structure.
* @param[in] nbOfSupportVectors Number of support vectors
* @param[in] vectorDimension Dimension of vector space
* @param[in] intercept Intercept
* @param[in] dualCoefficients Array of dual coefficients
* @param[in] supportVectors Array of support vectors
* @param[in] classes Array of 2 classes ID
* @param[in] degree Polynomial degree
* @param[in] coef0 coeff0 (scikit-learn terminology)
* @param[in] gamma gamma (scikit-learn terminology)
*/
void arm_svm_polynomial_init_f16(arm_svm_polynomial_instance_f16 *S,
uint32_t nbOfSupportVectors,
uint32_t vectorDimension,
float16_t intercept,
const float16_t *dualCoefficients,
const float16_t *supportVectors,
const int32_t *classes,
int32_t degree,
float16_t coef0,
float16_t gamma
);
/**
* @brief SVM polynomial prediction
* @param[in] S Pointer to an instance of the polynomial SVM structure.
* @param[in] in Pointer to input vector
* @param[out] pResult Decision value
*/
void arm_svm_polynomial_predict_f16(const arm_svm_polynomial_instance_f16 *S,
const float16_t * in,
int32_t * pResult);
/**
* @brief SVM radial basis function instance init function
* @param[in] S points to an instance of the polynomial SVM structure.
* @param[in] nbOfSupportVectors Number of support vectors
* @param[in] vectorDimension Dimension of vector space
* @param[in] intercept Intercept
* @param[in] dualCoefficients Array of dual coefficients
* @param[in] supportVectors Array of support vectors
* @param[in] classes Array of 2 classes ID
* @param[in] gamma gamma (scikit-learn terminology)
*/
void arm_svm_rbf_init_f16(arm_svm_rbf_instance_f16 *S,
uint32_t nbOfSupportVectors,
uint32_t vectorDimension,
float16_t intercept,
const float16_t *dualCoefficients,
const float16_t *supportVectors,
const int32_t *classes,
float16_t gamma
);
/**
* @brief SVM rbf prediction
* @param[in] S Pointer to an instance of the rbf SVM structure.
* @param[in] in Pointer to input vector
* @param[out] pResult decision value
*/
void arm_svm_rbf_predict_f16(const arm_svm_rbf_instance_f16 *S,
const float16_t * in,
int32_t * pResult);
/**
* @brief SVM sigmoid instance init function
* @param[in] S points to an instance of the rbf SVM structure.
* @param[in] nbOfSupportVectors Number of support vectors
* @param[in] vectorDimension Dimension of vector space
* @param[in] intercept Intercept
* @param[in] dualCoefficients Array of dual coefficients
* @param[in] supportVectors Array of support vectors
* @param[in] classes Array of 2 classes ID
* @param[in] coef0 coeff0 (scikit-learn terminology)
* @param[in] gamma gamma (scikit-learn terminology)
*/
void arm_svm_sigmoid_init_f16(arm_svm_sigmoid_instance_f16 *S,
uint32_t nbOfSupportVectors,
uint32_t vectorDimension,
float16_t intercept,
const float16_t *dualCoefficients,
const float16_t *supportVectors,
const int32_t *classes,
float16_t coef0,
float16_t gamma
);
/**
* @brief SVM sigmoid prediction
* @param[in] S Pointer to an instance of the rbf SVM structure.
* @param[in] in Pointer to input vector
* @param[out] pResult Decision value
*/
void arm_svm_sigmoid_predict_f16(const arm_svm_sigmoid_instance_f16 *S,
const float16_t * in,
int32_t * pResult);
#endif /*defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _SVM_FUNCTIONS_F16_H_ */

1151
sdk/CMSIS/DSP/Include/dsp/transform_functions.h Normal file
View file

File diff suppressed because it is too large Load diff

316
sdk/CMSIS/DSP/Include/dsp/transform_functions_f16.h Normal file
View file

@ -0,0 +1,316 @@
/******************************************************************************
* @file transform_functions_f16.h
* @brief Public header file for CMSIS DSP Library
* @version V1.10.0
* @date 08 July 2021
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef TRANSFORM_FUNCTIONS_F16_H_
#define TRANSFORM_FUNCTIONS_F16_H_
#include "arm_math_types_f16.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if defined(ARM_FLOAT16_SUPPORTED)
/**
* @brief Instance structure for the floating-point CFFT/CIFFT function.
*/
typedef struct
{
uint16_t fftLen; /**< length of the FFT. */
uint8_t ifftFlag; /**< flag that selects forward (ifftFlag=0) or inverse (ifftFlag=1) transform. */
uint8_t bitReverseFlag; /**< flag that enables (bitReverseFlag=1) or disables (bitReverseFlag=0) bit reversal of output. */
const float16_t *pTwiddle; /**< points to the Twiddle factor table. */
const uint16_t *pBitRevTable; /**< points to the bit reversal table. */
uint16_t twidCoefModifier; /**< twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table. */
uint16_t bitRevFactor; /**< bit reversal modifier that supports different size FFTs with the same bit reversal table. */
float16_t onebyfftLen; /**< value of 1/fftLen. */
} arm_cfft_radix2_instance_f16;
/**
* @brief Instance structure for the floating-point CFFT/CIFFT function.
*/
typedef struct
{
uint16_t fftLen; /**< length of the FFT. */
uint8_t ifftFlag; /**< flag that selects forward (ifftFlag=0) or inverse (ifftFlag=1) transform. */
uint8_t bitReverseFlag; /**< flag that enables (bitReverseFlag=1) or disables (bitReverseFlag=0) bit reversal of output. */
const float16_t *pTwiddle; /**< points to the Twiddle factor table. */
const uint16_t *pBitRevTable; /**< points to the bit reversal table. */
uint16_t twidCoefModifier; /**< twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table. */
uint16_t bitRevFactor; /**< bit reversal modifier that supports different size FFTs with the same bit reversal table. */
float16_t onebyfftLen; /**< value of 1/fftLen. */
} arm_cfft_radix4_instance_f16;
/**
* @brief Instance structure for the floating-point CFFT/CIFFT function.
*/
typedef struct
{
uint16_t fftLen; /**< length of the FFT. */
const float16_t *pTwiddle; /**< points to the Twiddle factor table. */
const uint16_t *pBitRevTable; /**< points to the bit reversal table. */
uint16_t bitRevLength; /**< bit reversal table length. */
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
const uint32_t *rearranged_twiddle_tab_stride1_arr; /**< Per stage reordered twiddle pointer (offset 1) */ \
const uint32_t *rearranged_twiddle_tab_stride2_arr; /**< Per stage reordered twiddle pointer (offset 2) */ \
const uint32_t *rearranged_twiddle_tab_stride3_arr; /**< Per stage reordered twiddle pointer (offset 3) */ \
const float16_t *rearranged_twiddle_stride1; /**< reordered twiddle offset 1 storage */ \
const float16_t *rearranged_twiddle_stride2; /**< reordered twiddle offset 2 storage */ \
const float16_t *rearranged_twiddle_stride3;
#endif
} arm_cfft_instance_f16;
arm_status arm_cfft_init_4096_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_2048_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_1024_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_512_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_256_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_128_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_64_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_32_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_16_f16(arm_cfft_instance_f16 * S);
arm_status arm_cfft_init_f16(
arm_cfft_instance_f16 * S,
uint16_t fftLen);
void arm_cfft_f16(
const arm_cfft_instance_f16 * S,
float16_t * p1,
uint8_t ifftFlag,
uint8_t bitReverseFlag);
/**
* @brief Instance structure for the floating-point RFFT/RIFFT function.
*/
typedef struct
{
arm_cfft_instance_f16 Sint; /**< Internal CFFT structure. */
uint16_t fftLenRFFT; /**< length of the real sequence */
const float16_t * pTwiddleRFFT; /**< Twiddle factors real stage */
} arm_rfft_fast_instance_f16 ;
arm_status arm_rfft_fast_init_32_f16( arm_rfft_fast_instance_f16 * S );
arm_status arm_rfft_fast_init_64_f16( arm_rfft_fast_instance_f16 * S );
arm_status arm_rfft_fast_init_128_f16( arm_rfft_fast_instance_f16 * S );
arm_status arm_rfft_fast_init_256_f16( arm_rfft_fast_instance_f16 * S );
arm_status arm_rfft_fast_init_512_f16( arm_rfft_fast_instance_f16 * S );
arm_status arm_rfft_fast_init_1024_f16( arm_rfft_fast_instance_f16 * S );
arm_status arm_rfft_fast_init_2048_f16( arm_rfft_fast_instance_f16 * S );
arm_status arm_rfft_fast_init_4096_f16( arm_rfft_fast_instance_f16 * S );
arm_status arm_rfft_fast_init_f16 (
arm_rfft_fast_instance_f16 * S,
uint16_t fftLen);
void arm_rfft_fast_f16(
const arm_rfft_fast_instance_f16 * S,
float16_t * p, float16_t * pOut,
uint8_t ifftFlag);
/* Deprecated */
arm_status arm_cfft_radix4_init_f16(
arm_cfft_radix4_instance_f16 * S,
uint16_t fftLen,
uint8_t ifftFlag,
uint8_t bitReverseFlag);
/* Deprecated */
void arm_cfft_radix4_f16(
const arm_cfft_radix4_instance_f16 * S,
float16_t * pSrc);
/* Deprecated */
arm_status arm_cfft_radix2_init_f16(
arm_cfft_radix2_instance_f16 * S,
uint16_t fftLen,
uint8_t ifftFlag,
uint8_t bitReverseFlag);
/* Deprecated */
void arm_cfft_radix2_f16(
const arm_cfft_radix2_instance_f16 * S,
float16_t * pSrc);
/**
* @brief Instance structure for the Floating-point MFCC function.
*/
typedef struct
{
const float16_t *dctCoefs; /**< Internal DCT coefficients */
const float16_t *filterCoefs; /**< Internal Mel filter coefficients */
const float16_t *windowCoefs; /**< Windowing coefficients */
const uint32_t *filterPos; /**< Internal Mel filter positions in spectrum */
const uint32_t *filterLengths; /**< Internal Mel filter lengths */
uint32_t fftLen; /**< FFT length */
uint32_t nbMelFilters; /**< Number of Mel filters */
uint32_t nbDctOutputs; /**< Number of DCT outputs */
#if defined(ARM_MFCC_CFFT_BASED)
/* Implementation of the MFCC is using a CFFT */
arm_cfft_instance_f16 cfft; /**< Internal CFFT instance */
#else
/* Implementation of the MFCC is using a RFFT (default) */
arm_rfft_fast_instance_f16 rfft;
#endif
} arm_mfcc_instance_f16 ;
arm_status arm_mfcc_init_32_f16(
arm_mfcc_instance_f16 * S,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
arm_status arm_mfcc_init_64_f16(
arm_mfcc_instance_f16 * S,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
arm_status arm_mfcc_init_128_f16(
arm_mfcc_instance_f16 * S,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
arm_status arm_mfcc_init_256_f16(
arm_mfcc_instance_f16 * S,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
arm_status arm_mfcc_init_512_f16(
arm_mfcc_instance_f16 * S,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
arm_status arm_mfcc_init_1024_f16(
arm_mfcc_instance_f16 * S,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
arm_status arm_mfcc_init_2048_f16(
arm_mfcc_instance_f16 * S,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
arm_status arm_mfcc_init_4096_f16(
arm_mfcc_instance_f16 * S,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
arm_status arm_mfcc_init_f16(
arm_mfcc_instance_f16 * S,
uint32_t fftLen,
uint32_t nbMelFilters,
uint32_t nbDctOutputs,
const float16_t *dctCoefs,
const uint32_t *filterPos,
const uint32_t *filterLengths,
const float16_t *filterCoefs,
const float16_t *windowCoefs
);
/**
@brief MFCC F16
@param[in] S points to the mfcc instance structure
@param[in] pSrc points to the input samples
@param[out] pDst points to the output MFCC values
@param[inout] pTmp points to a temporary buffer of complex
*/
void arm_mfcc_f16(
const arm_mfcc_instance_f16 * S,
float16_t *pSrc,
float16_t *pDst,
float16_t *pTmp
);
#endif /* defined(ARM_FLOAT16_SUPPORTED)*/
#ifdef __cplusplus
}
#endif
#endif /* ifndef _TRANSFORM_FUNCTIONS_F16_H_ */

262
sdk/CMSIS/DSP/Include/dsp/utils.h Normal file
View file

@ -0,0 +1,262 @@
/******************************************************************************
* @file arm_math_utils.h
* @brief Public header file for CMSIS DSP Library
* @version V1.9.0
* @date 20. July 2020
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_MATH_UTILS_H_
#define ARM_MATH_UTILS_H_
#include "arm_math_types.h"
#include <limits.h>
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @brief Macros required for reciprocal calculation in Normalized LMS
*/
#define INDEX_MASK 0x0000003F
#ifndef MIN
#define MIN(x,y) ((x) < (y) ? (x) : (y))
#endif
#ifndef MAX
#define MAX(x,y) ((x) > (y) ? (x) : (y))
#endif
#ifndef ARM_SQ
#define ARM_SQ(x) ((x) * (x))
#endif
#ifndef ARM_ROUND_UP
#define ARM_ROUND_UP(N, S) ((((N) + (S) - 1) / (S)) * (S))
#endif
/**
* @brief Function to Calculates 1/in (reciprocal) value of Q31 Data type.
It should not be used with negative values.
*/
__STATIC_FORCEINLINE uint32_t arm_recip_q31(
q31_t in,
q31_t * dst,
const q31_t * pRecipTable)
{
q31_t out;
uint32_t tempVal;
uint32_t index, i;
uint32_t signBits;
if (in > 0)
{
signBits = ((uint32_t) (__CLZ( (uint32_t)in) - 1));
}
else
{
signBits = ((uint32_t) (__CLZ((uint32_t)(-in)) - 1));
}
/* Convert input sample to 1.31 format */
in = (in << signBits);
/* calculation of index for initial approximated Val */
index = (uint32_t)(in >> 24);
index = (index & INDEX_MASK);
/* 1.31 with exp 1 */
out = pRecipTable[index];
/* calculation of reciprocal value */
/* running approximation for two iterations */
for (i = 0U; i < 2U; i++)
{
tempVal = (uint32_t) (((q63_t) in * out) >> 31);
tempVal = 0x7FFFFFFFu - tempVal;
/* 1.31 with exp 1 */
/* out = (q31_t) (((q63_t) out * tempVal) >> 30); */
out = clip_q63_to_q31(((q63_t) out * tempVal) >> 30);
}
/* write output */
*dst = out;
/* return num of signbits of out = 1/in value */
return (signBits + 1U);
}
/**
* @brief Function to Calculates 1/in (reciprocal) value of Q15 Data type.
It should not be used with negative values.
*/
__STATIC_FORCEINLINE uint32_t arm_recip_q15(
q15_t in,
q15_t * dst,
const q15_t * pRecipTable)
{
q15_t out = 0;
int32_t tempVal = 0;
uint32_t index = 0, i = 0;
uint32_t signBits = 0;
if (in > 0)
{
signBits = ((uint32_t)(__CLZ( (uint32_t)in) - 17));
}
else
{
signBits = ((uint32_t)(__CLZ((uint32_t)(-in)) - 17));
}
/* Convert input sample to 1.15 format */
in = (q15_t)(in << signBits);
/* calculation of index for initial approximated Val */
index = (uint32_t)(in >> 8);
index = (index & INDEX_MASK);
/* 1.15 with exp 1 */
out = pRecipTable[index];
/* calculation of reciprocal value */
/* running approximation for two iterations */
for (i = 0U; i < 2U; i++)
{
tempVal = (((q31_t) in * out) >> 15);
tempVal = 0x7FFF - tempVal;
/* 1.15 with exp 1 */
out = (q15_t) (((q31_t) out * tempVal) >> 14);
/* out = clip_q31_to_q15(((q31_t) out * tempVal) >> 14); */
}
/* write output */
*dst = out;
/* return num of signbits of out = 1/in value */
return (signBits + 1);
}
/**
* @brief 64-bit to 32-bit unsigned normalization
* @param[in] in is input unsigned long long value
* @param[out] normalized is the 32-bit normalized value
* @param[out] norm is norm scale
*/
__STATIC_INLINE void arm_norm_64_to_32u(uint64_t in, int32_t * normalized, int32_t *norm)
{
int32_t n1;
int32_t hi = (int32_t) (in >> 32);
int32_t lo = (int32_t) ((in << 32) >> 32);
n1 = __CLZ((uint32_t)hi) - 32;
if (!n1)
{
/*
* input fits in 32-bit
*/
n1 = __CLZ((uint32_t)lo);
if (!n1)
{
/*
* MSB set, need to scale down by 1
*/
*norm = -1;
*normalized = (((uint32_t) lo) >> 1);
} else
{
if (n1 == 32)
{
/*
* input is zero
*/
*norm = 0;
*normalized = 0;
} else
{
/*
* 32-bit normalization
*/
*norm = n1 - 1;
*normalized = lo << *norm;
}
}
} else
{
/*
* input fits in 64-bit
*/
n1 = 1 - n1;
*norm = -n1;
/*
* 64 bit normalization
*/
*normalized = (int32_t)(((uint32_t)lo) >> n1) | (hi << (32 - n1));
}
}
__STATIC_INLINE int32_t arm_div_int64_to_int32(int64_t num, int32_t den)
{
int32_t result;
uint64_t absNum;
int32_t normalized;
int32_t norm;
/*
* if sum fits in 32bits
* avoid costly 64-bit division
*/
if (num == (int64_t)LONG_MIN)
{
absNum = LONG_MAX;
}
else
{
absNum = (uint64_t) (num > 0 ? num : -num);
}
arm_norm_64_to_32u(absNum, &normalized, &norm);
if (norm > 0)
/*
* 32-bit division
*/
result = (int32_t) num / den;
else
/*
* 64-bit division
*/
result = (int32_t) (num / den);
return result;
}
#undef INDEX_MASK
#ifdef __cplusplus
}
#endif
#endif /*ifndef _ARM_MATH_UTILS_H_ */

812
sdk/CMSIS/DSP/Include/dsp/window_functions.h Normal file
View file

@ -0,0 +1,812 @@
/******************************************************************************
* @file window_functions.h
* @brief Public header file for CMSIS DSP Library
* @version v1.15.0
* @date 15 December 2022
* Target Processor: Cortex-M and Cortex-A cores
******************************************************************************/
/*
* Copyright (c) 2010-2022 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef WINDOW_FUNCTIONS_H_
#define WINDOW_FUNCTIONS_H_
#include "arm_math_types.h"
#include "arm_math_memory.h"
#include "dsp/none.h"
#include "dsp/utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @defgroup groupWindow Window Functions
*/
/**
* @brief Welch window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 21.3 dB |
* | Normalized equivalent noise bandwidth | 1.2 bins |
* | Flatness | -2.2248 dB |
* | Recommended overlap | 29.3 % |
*
*/
void arm_welch_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Welch window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 21.3 dB |
* | Normalized equivalent noise bandwidth | 1.2 bins |
* | Flatness | -2.2248 dB |
* | Recommended overlap | 29.3 % |
*
*
*/
void arm_welch_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Bartlett window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 26.5 dB |
* | Normalized equivalent noise bandwidth | 1.3333 bins |
* | Flatness | -1.8242 dB |
* | Recommended overlap | 50.0 % |
*
*/
void arm_bartlett_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Bartlett window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 26.5 dB |
* | Normalized equivalent noise bandwidth | 1.3333 bins |
* | Flatness | -1.8242 dB |
* | Recommended overlap | 50.0 % |
*
*
*/
void arm_bartlett_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hamming window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 42.7 dB |
* | Normalized equivalent noise bandwidth | 1.3628 bins |
* | Flatness | -1.7514 dB |
* | Recommended overlap | 50 % |
*
*/
void arm_hamming_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hamming window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 42.7 dB |
* | Normalized equivalent noise bandwidth | 1.3628 bins |
* | Flatness | -1.7514 dB |
* | Recommended overlap | 50 % |
*
*
*/
void arm_hamming_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hanning window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 31.5 dB |
* | Normalized equivalent noise bandwidth | 1.5 bins |
* | Flatness | -1.4236 dB |
* | Recommended overlap | 50 % |
*
*/
void arm_hanning_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hanning window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 31.5 dB |
* | Normalized equivalent noise bandwidth | 1.5 bins |
* | Flatness | -1.4236 dB |
* | Recommended overlap | 50 % |
*
*
*/
void arm_hanning_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall3 window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 46.7 dB |
* | Normalized equivalent noise bandwidth | 1.9444 bins |
* | Flatness | -0.863 dB |
* | Recommended overlap | 64.7 % |
*
*/
void arm_nuttall3_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall3 window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 46.7 dB |
* | Normalized equivalent noise bandwidth | 1.9444 bins |
* | Flatness | -0.863 dB |
* | Recommended overlap | 64.7 % |
*
*
*/
void arm_nuttall3_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall4 window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 60.9 dB |
* | Normalized equivalent noise bandwidth | 2.31 bins |
* | Flatness | -0.6184 dB |
* | Recommended overlap | 70.5 % |
*
*/
void arm_nuttall4_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall4 window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 60.9 dB |
* | Normalized equivalent noise bandwidth | 2.31 bins |
* | Flatness | -0.6184 dB |
* | Recommended overlap | 70.5 % |
*
*
*/
void arm_nuttall4_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall3a window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 64.2 dB |
* | Normalized equivalent noise bandwidth | 1.7721 bins |
* | Flatness | -1.0453 dB |
* | Recommended overlap | 61.2 % |
*
*/
void arm_nuttall3a_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall3a window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 64.2 dB |
* | Normalized equivalent noise bandwidth | 1.7721 bins |
* | Flatness | -1.0453 dB |
* | Recommended overlap | 61.2 % |
*
*
*/
void arm_nuttall3a_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall3b window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 71.5 dB |
* | Normalized equivalent noise bandwidth | 1.7037 bins |
* | Flatness | -1.1352 dB |
* | Recommended overlap | 59.8 % |
*
*/
void arm_nuttall3b_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall3b window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 71.5 dB |
* | Normalized equivalent noise bandwidth | 1.7037 bins |
* | Flatness | -1.1352 dB |
* | Recommended overlap | 59.8 % |
*
*
*/
void arm_nuttall3b_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall4a window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 82.6 dB |
* | Normalized equivalent noise bandwidth | 2.1253 bins |
* | Flatness | -0.7321 dB |
* | Recommended overlap | 68.0 % |
*
*/
void arm_nuttall4a_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall4a window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 82.6 dB |
* | Normalized equivalent noise bandwidth | 2.1253 bins |
* | Flatness | -0.7321 dB |
* | Recommended overlap | 68.0 % |
*
*
*/
void arm_nuttall4a_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief 92 db blackman harris window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 92.0 dB |
* | Normalized equivalent noise bandwidth | 2.0044 bins |
* | Flatness | -0.8256 dB |
* | Recommended overlap | 66.1 % |
*
*/
void arm_blackman_harris_92db_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief 92 db blackman harris window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 92.0 dB |
* | Normalized equivalent noise bandwidth | 2.0044 bins |
* | Flatness | -0.8256 dB |
* | Recommended overlap | 66.1 % |
*
*
*/
void arm_blackman_harris_92db_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall4b window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 93.3 dB |
* | Normalized equivalent noise bandwidth | 2.0212 bins |
* | Flatness | -0.8118 dB |
* | Recommended overlap | 66.3 % |
*
*/
void arm_nuttall4b_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall4b window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 93.3 dB |
* | Normalized equivalent noise bandwidth | 2.0212 bins |
* | Flatness | -0.8118 dB |
* | Recommended overlap | 66.3 % |
*
*
*/
void arm_nuttall4b_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall4c window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 98.1 dB |
* | Normalized equivalent noise bandwidth | 1.9761 bins |
* | Flatness | -0.8506 dB |
* | Recommended overlap | 65.6 % |
*
*/
void arm_nuttall4c_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Nuttall4c window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 98.1 dB |
* | Normalized equivalent noise bandwidth | 1.9761 bins |
* | Flatness | -0.8506 dB |
* | Recommended overlap | 65.6 % |
*
*
*/
void arm_nuttall4c_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hft90d window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 90.2 dB |
* | Normalized equivalent noise bandwidth | 3.8832 bins |
* | Flatness | -0.0039 dB |
* | Recommended overlap | 76.0 % |
*
*/
void arm_hft90d_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hft90d window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 90.2 dB |
* | Normalized equivalent noise bandwidth | 3.8832 bins |
* | Flatness | -0.0039 dB |
* | Recommended overlap | 76.0 % |
*
*
*/
void arm_hft90d_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hft95 window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 95.0 dB |
* | Normalized equivalent noise bandwidth | 3.8112 bins |
* | Flatness | 0.0044 dB |
* | Recommended overlap | 75.6 % |
*
*/
void arm_hft95_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hft95 window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 95.0 dB |
* | Normalized equivalent noise bandwidth | 3.8112 bins |
* | Flatness | 0.0044 dB |
* | Recommended overlap | 75.6 % |
*
*
*/
void arm_hft95_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hft116d window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 116.8 dB |
* | Normalized equivalent noise bandwidth | 4.2186 bins |
* | Flatness | -0.0028 dB |
* | Recommended overlap | 78.2 % |
*
*/
void arm_hft116d_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hft116d window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 116.8 dB |
* | Normalized equivalent noise bandwidth | 4.2186 bins |
* | Flatness | -0.0028 dB |
* | Recommended overlap | 78.2 % |
*
*
*/
void arm_hft116d_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hft144d window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 144.1 dB |
* | Normalized equivalent noise bandwidth | 4.5386 bins |
* | Flatness | 0.0021 dB |
* | Recommended overlap | 79.9 % |
*
*/
void arm_hft144d_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hft144d window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 144.1 dB |
* | Normalized equivalent noise bandwidth | 4.5386 bins |
* | Flatness | 0.0021 dB |
* | Recommended overlap | 79.9 % |
*
*
*/
void arm_hft144d_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hft169d window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 169.5 dB |
* | Normalized equivalent noise bandwidth | 4.8347 bins |
* | Flatness | 0.0017 dB |
* | Recommended overlap | 81.2 % |
*
*/
void arm_hft169d_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hft169d window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 169.5 dB |
* | Normalized equivalent noise bandwidth | 4.8347 bins |
* | Flatness | 0.0017 dB |
* | Recommended overlap | 81.2 % |
*
*
*/
void arm_hft169d_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hft196d window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 196.2 dB |
* | Normalized equivalent noise bandwidth | 5.1134 bins |
* | Flatness | 0.0013 dB |
* | Recommended overlap | 82.3 % |
*
*/
void arm_hft196d_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hft196d window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 196.2 dB |
* | Normalized equivalent noise bandwidth | 5.1134 bins |
* | Flatness | 0.0013 dB |
* | Recommended overlap | 82.3 % |
*
*
*/
void arm_hft196d_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hft223d window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 223.0 dB |
* | Normalized equivalent noise bandwidth | 5.3888 bins |
* | Flatness | 0.0011 dB |
* | Recommended overlap | 83.3 % |
*
*/
void arm_hft223d_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hft223d window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 223.0 dB |
* | Normalized equivalent noise bandwidth | 5.3888 bins |
* | Flatness | 0.0011 dB |
* | Recommended overlap | 83.3 % |
*
*
*/
void arm_hft223d_f32(
float32_t * pDst,
uint32_t blockSize);
/**
* @brief Hft248d window (double).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 248.4 dB |
* | Normalized equivalent noise bandwidth | 5.6512 bins |
* | Flatness | 0.0009 dB |
* | Recommended overlap | 84.1 % |
*
*/
void arm_hft248d_f64(
float64_t * pDst,
uint32_t blockSize);
/**
* @brief Hft248d window (float).
* @param[out] pDst points to the output generated window
* @param[in] blockSize number of samples in the window
*
* @par Parameters of the window
*
* | Parameter | Value |
* | ------------------------------------: | -----------------: |
* | Peak sidelobe level | 248.4 dB |
* | Normalized equivalent noise bandwidth | 5.6512 bins |
* | Flatness | 0.0009 dB |
* | Recommended overlap | 84.1 % |
*
*
*/
void arm_hft248d_f32(
float32_t * pDst,
uint32_t blockSize);
#ifdef __cplusplus
}
#endif
#endif /* ifndef _BASIC_MATH_FUNCTIONS_H_ */

201
sdk/CMSIS/DSP/LICENSE Normal file
View file

@ -0,0 +1,201 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "{}"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright {yyyy} {name of copyright owner}
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

200
sdk/CMSIS/DSP/PrivateInclude/arm_sorting.h Normal file
View file

@ -0,0 +1,200 @@
/******************************************************************************
* @file arm_sorting.h
* @brief Private header file for CMSIS DSP Library
* @version V1.7.0
* @date 2019
******************************************************************************/
/*
* Copyright (c) 2010-2019 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_SORTING_H_
#define ARM_SORTING_H_
#include "arm_math.h"
#ifdef __cplusplus
extern "C"
{
#endif
/**
* @param[in] S points to an instance of the sorting structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data.
* @param[in] blockSize number of samples to process.
*/
void arm_bubble_sort_f32(
const arm_sort_instance_f32 * S,
float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @param[in] S points to an instance of the sorting structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data.
* @param[in] blockSize number of samples to process.
*/
void arm_heap_sort_f32(
const arm_sort_instance_f32 * S,
float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @param[in] S points to an instance of the sorting structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data.
* @param[in] blockSize number of samples to process.
*/
void arm_insertion_sort_f32(
const arm_sort_instance_f32 * S,
float32_t *pSrc,
float32_t* pDst,
uint32_t blockSize);
/**
* @param[in] S points to an instance of the sorting structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data
* @param[in] blockSize number of samples to process.
*/
void arm_quick_sort_f32(
const arm_sort_instance_f32 * S,
float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @param[in] S points to an instance of the sorting structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data
* @param[in] blockSize number of samples to process.
*/
void arm_selection_sort_f32(
const arm_sort_instance_f32 * S,
float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
/**
* @param[in] S points to an instance of the sorting structure.
* @param[in] pSrc points to the block of input data.
* @param[out] pDst points to the block of output data
* @param[in] blockSize number of samples to process.
*/
void arm_bitonic_sort_f32(
const arm_sort_instance_f32 * S,
float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize);
#if defined(ARM_MATH_NEON)
#define vtrn256_128q(a, b) \
do { \
float32x4_t vtrn128_temp = a.val[1]; \
a.val[1] = b.val[0]; \
b.val[0] = vtrn128_temp ; \
} while (0)
#define vtrn128_64q(a, b) \
do { \
float32x2_t ab, cd, ef, gh; \
ab = vget_low_f32(a); \
ef = vget_low_f32(b); \
cd = vget_high_f32(a); \
gh = vget_high_f32(b); \
a = vcombine_f32(ab, ef); \
b = vcombine_f32(cd, gh); \
} while (0)
#define vtrn256_64q(a, b) \
do { \
float32x2_t a_0, a_1, a_2, a_3; \
float32x2_t b_0, b_1, b_2, b_3; \
a_0 = vget_low_f32(a.val[0]); \
a_1 = vget_high_f32(a.val[0]); \
a_2 = vget_low_f32(a.val[1]); \
a_3 = vget_high_f32(a.val[1]); \
b_0 = vget_low_f32(b.val[0]); \
b_1 = vget_high_f32(b.val[0]); \
b_2 = vget_low_f32(b.val[1]); \
b_3 = vget_high_f32(b.val[1]); \
a.val[0] = vcombine_f32(a_0, b_0); \
a.val[1] = vcombine_f32(a_2, b_2); \
b.val[0] = vcombine_f32(a_1, b_1); \
b.val[1] = vcombine_f32(a_3, b_3); \
} while (0)
#define vtrn128_32q(a, b) \
do { \
float32x4x2_t vtrn32_tmp = vtrnq_f32((a), (b)); \
(a) = vtrn32_tmp.val[0]; \
(b) = vtrn32_tmp.val[1]; \
} while (0)
#define vtrn256_32q(a, b) \
do { \
float32x4x2_t vtrn32_tmp_1 = vtrnq_f32((a.val[0]), (b.val[0])); \
float32x4x2_t vtrn32_tmp_2 = vtrnq_f32((a.val[1]), (b.val[1])); \
a.val[0] = vtrn32_tmp_1.val[0]; \
a.val[1] = vtrn32_tmp_2.val[0]; \
b.val[0] = vtrn32_tmp_1.val[1]; \
b.val[1] = vtrn32_tmp_2.val[1]; \
} while (0)
#define vminmaxq(a, b) \
do { \
float32x4_t minmax_tmp = (a); \
(a) = vminq_f32((a), (b)); \
(b) = vmaxq_f32(minmax_tmp, (b)); \
} while (0)
#define vminmax256q(a, b) \
do { \
float32x4x2_t minmax256_tmp = (a); \
a.val[0] = vminq_f32(a.val[0], b.val[0]); \
a.val[1] = vminq_f32(a.val[1], b.val[1]); \
b.val[0] = vmaxq_f32(minmax256_tmp.val[0], b.val[0]); \
b.val[1] = vmaxq_f32(minmax256_tmp.val[1], b.val[1]); \
} while (0)
#define vrev128q_f32(a) \
vcombine_f32(vrev64_f32(vget_high_f32(a)), vrev64_f32(vget_low_f32(a)))
#define vrev256q_f32(a) \
do { \
float32x4_t rev_tmp = vcombine_f32(vrev64_f32(vget_high_f32(a.val[0])), vrev64_f32(vget_low_f32(a.val[0]))); \
a.val[0] = vcombine_f32(vrev64_f32(vget_high_f32(a.val[1])), vrev64_f32(vget_low_f32(a.val[1]))); \
a.val[1] = rev_tmp; \
} while (0)
#define vldrev128q_f32(a, p) \
do { \
a = vld1q_f32(p); \
a = vrev128q_f32(a); \
} while (0)
#endif /* ARM_MATH_NEON */
#ifdef __cplusplus
}
#endif
#endif /* _ARM_SORTING_H */

325
sdk/CMSIS/DSP/PrivateInclude/arm_vec_fft.h Normal file
View file

@ -0,0 +1,325 @@
/******************************************************************************
* @file arm_vec_fft.h
* @brief Private header file for CMSIS DSP Library
* @version V1.7.0
* @date 07. January 2020
******************************************************************************/
/*
* Copyright (c) 2010-2020 Arm Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef ARM_VEC_FFT_H_
#define ARM_VEC_FFT_H_
#include "arm_math.h"
#include "arm_helium_utils.h"
#ifdef __cplusplus
extern "C"
{
#endif
#if (defined(ARM_MATH_MVEF) || defined(ARM_MATH_MVEI) || defined(ARM_MATH_HELIUM)) && !defined(ARM_MATH_AUTOVECTORIZE)
#define MVE_CMPLX_ADD_A_ixB(A, B) vcaddq_rot90(A,B)
#define MVE_CMPLX_SUB_A_ixB(A,B) vcaddq_rot270(A,B)
#define MVE_CMPLX_MULT_FLT_AxB(A,B) vcmlaq_rot90(vcmulq(A, B), A, B)
#define MVE_CMPLX_MULT_FLT_Conj_AxB(A,B) vcmlaq_rot270(vcmulq(A, B), A, B)
#define MVE_CMPLX_MULT_FX_AxB(A,B,TyA) vqdmladhxq(vqdmlsdhq((TyA)vuninitializedq_s32(), A, B), A, B)
#define MVE_CMPLX_MULT_FX_AxConjB(A,B,TyA) vqdmladhq(vqdmlsdhxq((TyA)vuninitializedq_s32(), A, B), A, B)
#define MVE_CMPLX_ADD_FX_A_ixB(A, B) vhcaddq_rot90(A,B)
#define MVE_CMPLX_SUB_FX_A_ixB(A,B) vhcaddq_rot270(A,B)
/**
@brief In-place 32 bit reversal function for helium
@param[in,out] pSrc points to in-place buffer of unknown 32-bit data type
@param[in] bitRevLen bit reversal table length
@param[in] pBitRevTab points to bit reversal table
@return none
*/
__STATIC_INLINE void arm_bitreversal_32_inpl_mve(
uint32_t *pSrc,
const uint16_t bitRevLen,
const uint16_t *pBitRevTab)
{
uint64_t *src = (uint64_t *) pSrc;
int32_t blkCnt; /* loop counters */
uint32x4_t bitRevTabOff;
uint32x4_t one = vdupq_n_u32(1);
uint64x2_t inLow, inHigh;
uint64x2_t bitRevOff1Low, bitRevOff0Low;
uint64x2_t bitRevOff1High, bitRevOff0High;
/* load scheduling to increase gather load idx update / gather load distance */
bitRevTabOff = vldrhq_u32(pBitRevTab);
pBitRevTab += 4;
bitRevOff0Low = vmullbq_int_u32(bitRevTabOff, one);
bitRevOff0High = vmulltq_int_u32(bitRevTabOff, one);
blkCnt = bitRevLen / 8;
while (blkCnt > 0) {
bitRevTabOff = vldrhq_u32(pBitRevTab);
pBitRevTab += 4;
/* 64-bit index expansion */
bitRevOff1Low = vmullbq_int_u32(bitRevTabOff, one);
bitRevOff1High = vmulltq_int_u32(bitRevTabOff, one);
inLow = vldrdq_gather_offset_u64(src, bitRevOff0Low);
inHigh = vldrdq_gather_offset_u64(src, bitRevOff0High);
vstrdq_scatter_offset_u64(src, bitRevOff0Low, inHigh);
vstrdq_scatter_offset_u64(src, bitRevOff0High, inLow);
/* unrolled */
bitRevTabOff = vldrhq_u32(pBitRevTab);
pBitRevTab += 4;
bitRevOff0Low = vmullbq_int_u32(bitRevTabOff, one);
bitRevOff0High = vmulltq_int_u32(bitRevTabOff, one);
inLow = vldrdq_gather_offset_u64(src, bitRevOff1Low);
inHigh = vldrdq_gather_offset_u64(src, bitRevOff1High);
vstrdq_scatter_offset_u64(src, bitRevOff1Low, inHigh);
vstrdq_scatter_offset_u64(src, bitRevOff1High, inLow);
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
}
if (bitRevLen & 7) {
/* FFT size = 16 */
inLow = vldrdq_gather_offset_u64(src, bitRevOff0Low);
inHigh = vldrdq_gather_offset_u64(src, bitRevOff0High);
vstrdq_scatter_offset_u64(src, bitRevOff0Low, inHigh);
vstrdq_scatter_offset_u64(src, bitRevOff0High, inLow);
}
}
/**
@brief In-place 16 bit reversal function for helium
@param[in,out] pSrc points to in-place buffer of unknown 16-bit data type
@param[in] bitRevLen bit reversal table length
@param[in] pBitRevTab points to bit reversal table
@return none
*/
__STATIC_INLINE void arm_bitreversal_16_inpl_mve(
uint16_t *pSrc,
const uint16_t bitRevLen,
const uint16_t *pBitRevTab)
{
uint32_t *src = (uint32_t *) pSrc;
int32_t blkCnt; /* loop counters */
uint32x4_t bitRevTabOff;
uint16x8_t one = vdupq_n_u16(1);
uint32x4_t bitRevOff1Low, bitRevOff0Low;
uint32x4_t bitRevOff1High, bitRevOff0High;
uint32x4_t inLow, inHigh;
/* load scheduling to increase gather load idx update / gather load distance */
bitRevTabOff = vldrhq_u16(pBitRevTab);
pBitRevTab += 8;
bitRevOff0Low = vmullbq_int_u16((uint16x8_t)bitRevTabOff, one);
bitRevOff0High = vmulltq_int_u16((uint16x8_t)bitRevTabOff, one);
bitRevOff0Low = vshrq_n_u16((uint16x8_t)bitRevOff0Low, 3);
bitRevOff0High = vshrq_n_u16((uint16x8_t)bitRevOff0High, 3);
blkCnt = (bitRevLen / 16);
while (blkCnt > 0) {
bitRevTabOff = vldrhq_u16(pBitRevTab);
pBitRevTab += 8;
bitRevOff1Low = vmullbq_int_u16((uint16x8_t)bitRevTabOff, one);
bitRevOff1High = vmulltq_int_u16((uint16x8_t)bitRevTabOff, one);
bitRevOff1Low = vshrq_n_u16((uint16x8_t)bitRevOff1Low, 3);
bitRevOff1High = vshrq_n_u16((uint16x8_t)bitRevOff1High, 3);
inLow = vldrwq_gather_shifted_offset_u32(src, bitRevOff0Low);
inHigh = vldrwq_gather_shifted_offset_u32(src, bitRevOff0High);
vstrwq_scatter_shifted_offset_u32(src, bitRevOff0Low, inHigh);
vstrwq_scatter_shifted_offset_u32(src, bitRevOff0High, inLow);
/* loop unrolling */
bitRevTabOff = vldrhq_u16(pBitRevTab);
pBitRevTab += 8;
bitRevOff0Low = vmullbq_int_u16((uint16x8_t)bitRevTabOff, one);
bitRevOff0High = vmulltq_int_u16((uint16x8_t)bitRevTabOff, one);
bitRevOff0Low = vshrq_n_u16((uint16x8_t)bitRevOff0Low, 3);
bitRevOff0High = vshrq_n_u16((uint16x8_t)bitRevOff0High, 3);
inLow = vldrwq_gather_shifted_offset_u32(src, bitRevOff1Low);
inHigh = vldrwq_gather_shifted_offset_u32(src, bitRevOff1High);
vstrwq_scatter_shifted_offset_u32(src, bitRevOff1Low, inHigh);
vstrwq_scatter_shifted_offset_u32(src, bitRevOff1High, inLow);
blkCnt--;
}
/* tail handling */
blkCnt = bitRevLen & 0xf;
if (blkCnt == 8) {
inLow = vldrwq_gather_shifted_offset_u32(src, bitRevOff0Low);
inHigh = vldrwq_gather_shifted_offset_u32(src, bitRevOff0High);
vstrwq_scatter_shifted_offset_u32(src, bitRevOff0Low, inHigh);
vstrwq_scatter_shifted_offset_u32(src, bitRevOff0High, inLow);
} else if (blkCnt == 12) {
/* FFT 16 special case */
mve_pred16_t p = vctp16q(4);
bitRevTabOff = vldrhq_z_u16(pBitRevTab, p);
inLow = vldrwq_gather_shifted_offset_u32(src, bitRevOff0Low);
inHigh = vldrwq_gather_shifted_offset_u32(src, bitRevOff0High);
vstrwq_scatter_shifted_offset_u32(src, bitRevOff0Low, inHigh);
vstrwq_scatter_shifted_offset_u32(src, bitRevOff0High, inLow);
bitRevOff0Low = vmullbq_int_u16((uint16x8_t)bitRevTabOff, one);
bitRevOff0High = vmulltq_int_u16((uint16x8_t)bitRevTabOff, one);
bitRevOff0Low = vshrq_n_u16((uint16x8_t)bitRevOff0Low, 3);
bitRevOff0High = vshrq_n_u16((uint16x8_t)bitRevOff0High, 3);
inLow = vldrwq_gather_shifted_offset_z_u32(src, bitRevOff0Low, p);
inHigh = vldrwq_gather_shifted_offset_z_u32(src, bitRevOff0High, p);
vstrwq_scatter_shifted_offset_p_u32(src, bitRevOff0Low, inHigh, p);
vstrwq_scatter_shifted_offset_p_u32(src, bitRevOff0High, inLow, p);
}
}
/**
@brief Out-of-place 32 bit reversal function for helium
@param[out] pDst points to destination buffer of unknown 32-bit data type
@param[in] pSrc points to input buffer of unknown 32-bit data type
@param[in] fftLen FFT length
@return none
*/
__STATIC_INLINE void arm_bitreversal_32_outpl_mve(void *pDst, void *pSrc, uint32_t fftLen)
{
uint32x4_t idxOffs0, idxOffs1, bitRevOffs0, bitRevOffs1;
uint32_t bitRevPos, blkCnt;
uint32_t *pDst32 = (uint32_t *) pDst;
/* fwd indexes */
idxOffs0 = vdupq_n_u32(0);
idxOffs1 = vdupq_n_u32(0);
idxOffs0[0] = 0; idxOffs0[2] = 4;
idxOffs1[0] = 8; idxOffs1[2] = 12;
bitRevPos = (31 - __CLZ(fftLen)) + 5;
blkCnt = fftLen >> 2;
/* issued earlier to increase gather load idx update / gather load distance */
/* bit-reverse fwd indexes */
bitRevOffs0 = vbrsrq(idxOffs0, bitRevPos);
bitRevOffs1 = vbrsrq(idxOffs1, bitRevPos);
while (blkCnt > 0) {
uint64x2_t vecIn;
vecIn = vldrdq_gather_offset_u64(pSrc, (uint64x2_t) bitRevOffs0);
idxOffs0 = idxOffs0 + 16;
vst1q(pDst32, (uint32x4_t) vecIn);
pDst32 += 4;
bitRevOffs0 = vbrsrq(idxOffs0, bitRevPos);
vecIn = vldrdq_gather_offset_u64(pSrc, (uint64x2_t) bitRevOffs1);
idxOffs1 = idxOffs1 + 16;
vst1q(pDst32, (uint32x4_t) vecIn);
pDst32 += 4;
bitRevOffs1 = vbrsrq(idxOffs1, bitRevPos);
blkCnt--;
}
}
/**
@brief Out-of-place 16 bit reversal function for helium
@param[out] pDst points to destination buffer of unknown 16-bit data type
@param[in] pSrc points to input buffer of unknown 16-bit data type
@param[in] fftLen FFT length
@return none
*/
__STATIC_INLINE void arm_bitreversal_16_outpl_mve(void *pDst, void *pSrc, uint32_t fftLen)
{
uint32x4_t idxOffs0, idxOffs1, bitRevOffs0, bitRevOffs1;
uint32_t bitRevPos, blkCnt;
uint16_t *pDst16 = (uint16_t *) pDst;
uint32_t incrIdx = 0;
/* fwd indexes */
idxOffs0 = vidupq_wb_u32(&incrIdx, 4); // {0, 4, 8, 12}
idxOffs1 = vidupq_wb_u32(&incrIdx, 4); // {16, 20, 24, 28}
bitRevPos = (31 - __CLZ(fftLen)) + 4;
blkCnt = fftLen >> 3;
/* issued earlier to increase gather load idx update / gather load distance */
/* bit-reverse fwd indexes */
bitRevOffs0 = vbrsrq(idxOffs0, bitRevPos);
bitRevOffs1 = vbrsrq(idxOffs1, bitRevPos);
while (blkCnt > 0) {
uint32x4_t vecIn;
vecIn = vldrwq_gather_offset_s32(pSrc, bitRevOffs0);
idxOffs0 = idxOffs0 + 32;
vst1q(pDst16, (uint16x8_t) vecIn);
pDst16 += 8;
bitRevOffs0 = vbrsrq(idxOffs0, bitRevPos);
vecIn = vldrwq_gather_offset_s32(pSrc, bitRevOffs1);
idxOffs1 = idxOffs1 + 32;
vst1q(pDst16, (uint16x8_t) vecIn);
pDst16 += 8;
bitRevOffs1 = vbrsrq(idxOffs1, bitRevPos);
blkCnt--;
}
}
#endif /* (defined(ARM_MATH_MVEF) || defined(ARM_MATH_HELIUM)) && !defined(ARM_MATH_AUTOVECTORIZE)*/
#ifdef __cplusplus
}
#endif
#endif /* _ARM_VEC_FFT_H_ */

1586
sdk/CMSIS/DSP/PrivateInclude/arm_vec_filtering.h Normal file
View file

File diff suppressed because it is too large Load diff

87
sdk/CMSIS/DSP/Source/BasicMathFunctions/BasicMathFunctions.c Normal file
View file

@ -0,0 +1,87 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: BasicMathFunctions.c
* Description: Combination of all basic math function source files.
*
* $Date: 16. March 2020
* $Revision: V1.1.0
*
* Target Processor: Cortex-M cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2019-2020 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "arm_abs_f32.c"
#include "arm_abs_f64.c"
#include "arm_abs_q15.c"
#include "arm_abs_q31.c"
#include "arm_abs_q7.c"
#include "arm_add_f32.c"
#include "arm_add_f64.c"
#include "arm_add_q15.c"
#include "arm_add_q31.c"
#include "arm_add_q7.c"
#include "arm_and_u16.c"
#include "arm_and_u32.c"
#include "arm_and_u8.c"
#include "arm_dot_prod_f32.c"
#include "arm_dot_prod_f64.c"
#include "arm_dot_prod_q15.c"
#include "arm_dot_prod_q31.c"
#include "arm_dot_prod_q7.c"
#include "arm_mult_f32.c"
#include "arm_mult_f64.c"
#include "arm_mult_q15.c"
#include "arm_mult_q31.c"
#include "arm_mult_q7.c"
#include "arm_negate_f32.c"
#include "arm_negate_f64.c"
#include "arm_negate_q15.c"
#include "arm_negate_q31.c"
#include "arm_negate_q7.c"
#include "arm_not_u16.c"
#include "arm_not_u32.c"
#include "arm_not_u8.c"
#include "arm_offset_f32.c"
#include "arm_offset_f64.c"
#include "arm_offset_q15.c"
#include "arm_offset_q31.c"
#include "arm_offset_q7.c"
#include "arm_or_u16.c"
#include "arm_or_u32.c"
#include "arm_or_u8.c"
#include "arm_scale_f32.c"
#include "arm_scale_f64.c"
#include "arm_scale_q15.c"
#include "arm_scale_q31.c"
#include "arm_scale_q7.c"
#include "arm_shift_q15.c"
#include "arm_shift_q31.c"
#include "arm_shift_q7.c"
#include "arm_sub_f32.c"
#include "arm_sub_f64.c"
#include "arm_sub_q15.c"
#include "arm_sub_q31.c"
#include "arm_sub_q7.c"
#include "arm_xor_u16.c"
#include "arm_xor_u32.c"
#include "arm_xor_u8.c"
#include "arm_clip_f32.c"
#include "arm_clip_q31.c"
#include "arm_clip_q15.c"
#include "arm_clip_q7.c"

37
sdk/CMSIS/DSP/Source/BasicMathFunctions/BasicMathFunctionsF16.c Normal file
View file

@ -0,0 +1,37 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: BasicMathFunctionsF16.c
* Description: Combination of all basic math function f16 source files.
*
* $Date: 20. April 2020
* $Revision: V1.1.0
*
* Target Processor: Cortex-M cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2019-2020 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "arm_abs_f16.c"
#include "arm_add_f16.c"
#include "arm_dot_prod_f16.c"
#include "arm_mult_f16.c"
#include "arm_negate_f16.c"
#include "arm_offset_f16.c"
#include "arm_scale_f16.c"
#include "arm_sub_f16.c"
#include "arm_clip_f16.c"

184
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_abs_f16.c Normal file
View file

@ -0,0 +1,184 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_abs_f16.c
* Description: Floating-point vector absolute value
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions_f16.h"
#include <math.h>
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAbs
@{
*/
/**
@brief Floating-point vector absolute value.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_abs_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
f16x8_t vec1;
f16x8_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 3U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute values and then store the results in the destination buffer. */
vec1 = vld1q(pSrc);
res = vabsq(vec1);
vst1q(pDst, res);
/* Increment pointers */
pSrc += 8;
pDst += 8;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x7;
if (blkCnt > 0U)
{
/* C = |A| */
mve_pred16_t p0 = vctp16q(blkCnt);
vec1 = vld1q(pSrc);
vstrhq_p(pDst, vabsq(vec1), p0);
}
}
#else
#if defined(ARM_FLOAT16_SUPPORTED)
ARM_DSP_ATTRIBUTE void arm_abs_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined(ARM_MATH_NEON_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
f16x8_t vec1;
f16x8_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute values and then store the results in the destination buffer. */
vec1 = vld1q_f16(pSrc);
res = vabsq_f16(vec1);
vst1q_f16(pDst, res);
/* Increment pointers */
pSrc += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x3;
#else
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute and store result in destination buffer. */
*pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
*pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
*pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
*pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
#endif /* #if defined(ARM_MATH_NEON) */
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute and store result in destination buffer. */
*pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_FLOAT16_SUPPORTED */
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicAbs group
*/

195
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_abs_f32.c Normal file
View file

@ -0,0 +1,195 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_abs_f32.c
* Description: Floating-point vector absolute value
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
#include <math.h>
/**
@ingroup groupMath
*/
/**
@defgroup BasicAbs Vector Absolute Value
Computes the absolute value of a vector on an element-by-element basis.
<pre>
pDst[n] = abs(pSrc[n]), 0 <= n < blockSize.
</pre>
The functions support in-place computation allowing the source and
destination pointers to reference the same memory buffer.
There are separate functions for floating-point, Q7, Q15, and Q31 data types.
*/
/**
@addtogroup BasicAbs
@{
*/
/**
@brief Floating-point vector absolute value.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_abs_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
f32x4_t vec1;
f32x4_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute values and then store the results in the destination buffer. */
vec1 = vld1q(pSrc);
res = vabsq(vec1);
vst1q(pDst, res);
/* Increment pointers */
pSrc += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x3;
if (blkCnt > 0U)
{
/* C = |A| */
mve_pred16_t p0 = vctp32q(blkCnt);
vec1 = vld1q(pSrc);
vstrwq_p(pDst, vabsq(vec1), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_abs_f32(
const float32_t * pSrc,
float32_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
f32x4_t vec1;
f32x4_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute values and then store the results in the destination buffer. */
vec1 = vld1q_f32(pSrc);
res = vabsq_f32(vec1);
vst1q_f32(pDst, res);
/* Increment pointers */
pSrc += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x3;
#else
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute and store result in destination buffer. */
*pDst++ = fabsf(*pSrc++);
*pDst++ = fabsf(*pSrc++);
*pDst++ = fabsf(*pSrc++);
*pDst++ = fabsf(*pSrc++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
#endif /* #if defined(ARM_MATH_NEON) */
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute and store result in destination buffer. */
*pDst++ = fabsf(*pSrc++);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicAbs group
*/

73
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_abs_f64.c Normal file
View file

@ -0,0 +1,73 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_abs_f64.c
* Description: Floating-point vector absolute value
*
* $Date: 13 September 2021
* $Revision: V1.10.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
#include <math.h>
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAbs
@{
*/
/**
@brief Floating-point vector absolute value.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
*/
ARM_DSP_ATTRIBUTE void arm_abs_f64(
const float64_t * pSrc,
float64_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute and store result in destination buffer. */
*pDst++ = fabs(*pSrc++);
/* Decrement loop counter */
blkCnt--;
}
}
/**
@} end of BasicAbs group
*/

177
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_abs_q15.c Normal file
View file

@ -0,0 +1,177 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_abs_q15.c
* Description: Q15 vector absolute value
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAbs
@{
*/
/**
@brief Q15 vector absolute value.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
The Q15 value -1 (0x8000) will be saturated to the maximum allowable positive value 0x7FFF.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_abs_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* loop counters */
q15x8_t vecSrc;
/* Compute 8 outputs at a time */
blkCnt = blockSize >> 3;
while (blkCnt > 0U)
{
/*
* C = |A|
* Calculate absolute and then store the results in the destination buffer.
*/
vecSrc = vld1q(pSrc);
vst1q(pDst, vqabsq(vecSrc));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrc += 8;
pDst += 8;
}
/*
* tail
*/
blkCnt = blockSize & 7;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp16q(blkCnt);
vecSrc = vld1q(pSrc);
vstrhq_p(pDst, vqabsq(vecSrc), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_abs_q15(
const q15_t * pSrc,
q15_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
q15_t in; /* Temporary input variable */
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute of input (if -1 then saturated to 0x7fff) and store result in destination buffer. */
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q15_t)__QSUB16(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q15_t) 0x8000) ? 0x7fff : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q15_t)__QSUB16(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q15_t) 0x8000) ? 0x7fff : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q15_t)__QSUB16(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q15_t) 0x8000) ? 0x7fff : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q15_t)__QSUB16(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q15_t) 0x8000) ? 0x7fff : -in);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute of input (if -1 then saturated to 0x7fff) and store result in destination buffer. */
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q15_t)__QSUB16(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q15_t) 0x8000) ? 0x7fff : -in);
#endif
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicAbs group
*/

207
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_abs_q31.c Normal file
View file

@ -0,0 +1,207 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_abs_q31.c
* Description: Q31 vector absolute value
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAbs
@{
*/
/**
@brief Q31 vector absolute value.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
The Q31 value -1 (0x80000000) will be saturated to the maximum allowable positive value 0x7FFFFFFF.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_abs_q31(
const q31_t * pSrc,
q31_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counters */
q31x4_t vecSrc;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2;
while (blkCnt > 0U)
{
/*
* C = |A|
* Calculate absolute and then store the results in the destination buffer.
*/
vecSrc = vld1q(pSrc);
vst1q(pDst, vqabsq(vecSrc));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* Advance vector source and destination pointers
*/
pSrc += 4;
pDst += 4;
}
/*
* Tail
*/
blkCnt = blockSize & 3;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp32q(blkCnt);
vecSrc = vld1q(pSrc);
vstrwq_p(pDst, vqabsq(vecSrc), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_abs_q31(
const q31_t * pSrc,
q31_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
q31_t in; /* Temporary variable */
#if defined(ARM_MATH_NEON)
int32x4_t vec1;
int32x4_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute and then store the results in the destination buffer. */
vec1 = vld1q_s32(pSrc);
res = vqabsq_s32(vec1);
vst1q_s32(pDst, res);
/* Increment pointers */
pSrc += 4;
pDst += 4;
/* Decrement the blockSize loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x3;
#else
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute of input (if -1 then saturated to 0x7fffffff) and store result in destination buffer. */
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
#endif /* #if defined (ARM_MATH_NEON) */
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute of input (if -1 then saturated to 0x7fffffff) and store result in destination buffer. */
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
#endif
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* #if defined (ARM_MATH_MVEI) */
/**
@} end of BasicAbs group
*/

179
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_abs_q7.c Normal file
View file

@ -0,0 +1,179 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_abs_q7.c
* Description: Q7 vector absolute value
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAbs
@{
*/
/**
@brief Q7 vector absolute value.
@param[in] pSrc points to the input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
@par Conditions for optimum performance
Input and output buffers should be aligned by 32-bit
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
The Q7 value -1 (0x80) will be saturated to the maximum allowable positive value 0x7F.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_abs_q7(
const q7_t * pSrc,
q7_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* loop counters */
q7x16_t vecSrc;
/* Compute 16 outputs at a time */
blkCnt = blockSize >> 4;
while (blkCnt > 0U)
{
/*
* C = |A|
* Calculate absolute and then store the results in the destination buffer.
*/
vecSrc = vld1q(pSrc);
vst1q(pDst, vqabsq(vecSrc));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrc += 16;
pDst += 16;
}
/*
* tail
*/
blkCnt = blockSize & 0xF;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp8q(blkCnt);
vecSrc = vld1q(pSrc);
vstrbq_p(pDst, vqabsq(vecSrc), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_abs_q7(
const q7_t * pSrc,
q7_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
q7_t in; /* Temporary input variable */
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute of input (if -1 then saturated to 0x7f) and store result in destination buffer. */
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q7_t)__QSUB8(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q7_t)__QSUB8(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q7_t)__QSUB8(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
#endif
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q7_t)__QSUB8(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute of input (if -1 then saturated to 0x7f) and store result in destination buffer. */
in = *pSrc++;
#if defined (ARM_MATH_DSP)
*pDst++ = (in > 0) ? in : (q7_t) __QSUB8(0, in);
#else
*pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
#endif
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicAbs group
*/

157
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_add_f16.c Normal file
View file

@ -0,0 +1,157 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_add_f16.c
* Description: Floating-point vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions_f16.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAdd
@{
*/
/**
@brief Floating-point vector addition.
@param[in] pSrcA points to first input vector
@param[in] pSrcB points to second input vector
@param[out] pDst points to output vector
@param[in] blockSize number of samples in each vector
*/
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_add_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
float16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
f16x8_t vec1;
f16x8_t vec2;
f16x8_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 3U;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and then store the results in the destination buffer. */
vec1 = vld1q(pSrcA);
vec2 = vld1q(pSrcB);
res = vaddq(vec1, vec2);
vst1q(pDst, res);
/* Increment pointers */
pSrcA += 8;
pSrcB += 8;
pDst += 8;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x7;
if (blkCnt > 0U)
{
/* C = A + B */
mve_pred16_t p0 = vctp16q(blkCnt);
vec1 = vld1q(pSrcA);
vec2 = vld1q(pSrcB);
vstrhq_p(pDst, vaddq(vec1,vec2), p0);
}
}
#else
#if defined(ARM_FLOAT16_SUPPORTED)
ARM_DSP_ATTRIBUTE void arm_add_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
float16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
*pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
*pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
*pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
*pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
*pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_FLOAT16_SUPPORTED) */
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicAdd group
*/

198
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_add_f32.c Normal file
View file

@ -0,0 +1,198 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_add_f32.c
* Description: Floating-point vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@defgroup BasicAdd Vector Addition
Element-by-element addition of two vectors.
<pre>
pDst[n] = pSrcA[n] + pSrcB[n], 0 <= n < blockSize.
</pre>
There are separate functions for floating-point, Q7, Q15, and Q31 data types.
*/
/**
@addtogroup BasicAdd
@{
*/
/**
@brief Floating-point vector addition.
@param[in] pSrcA points to first input vector
@param[in] pSrcB points to second input vector
@param[out] pDst points to output vector
@param[in] blockSize number of samples in each vector
*/
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_add_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
float32_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
f32x4_t vec1;
f32x4_t vec2;
f32x4_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and then store the results in the destination buffer. */
vec1 = vld1q(pSrcA);
vec2 = vld1q(pSrcB);
res = vaddq(vec1, vec2);
vst1q(pDst, res);
/* Increment pointers */
pSrcA += 4;
pSrcB += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x3;
if (blkCnt > 0U)
{
/* C = A + B */
mve_pred16_t p0 = vctp32q(blkCnt);
vec1 = vld1q(pSrcA);
vec2 = vld1q(pSrcB);
vstrwq_p(pDst, vaddq(vec1,vec2), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_add_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
float32_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
f32x4_t vec1;
f32x4_t vec2;
f32x4_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and then store the results in the destination buffer. */
vec1 = vld1q_f32(pSrcA);
vec2 = vld1q_f32(pSrcB);
res = vaddq_f32(vec1, vec2);
vst1q_f32(pDst, res);
/* Increment pointers */
pSrcA += 4;
pSrcB += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x3;
#else
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
*pDst++ = (*pSrcA++) + (*pSrcB++);
*pDst++ = (*pSrcA++) + (*pSrcB++);
*pDst++ = (*pSrcA++) + (*pSrcB++);
*pDst++ = (*pSrcA++) + (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
#endif /* #if defined(ARM_MATH_NEON) */
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
*pDst++ = (*pSrcA++) + (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicAdd group
*/

74
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_add_f64.c Normal file
View file

@ -0,0 +1,74 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_add_f64.c
* Description: Floating-point vector addition
*
* $Date: 13 September 2021
* $Revision: V1.10.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAdd
@{
*/
/**
@brief Floating-point vector addition.
@param[in] pSrcA points to first input vector
@param[in] pSrcB points to second input vector
@param[out] pDst points to output vector
@param[in] blockSize number of samples in each vector
*/
ARM_DSP_ATTRIBUTE void arm_add_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
float64_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
*pDst++ = (*pSrcA++) + (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
}
/**
@} end of BasicAdd group
*/

175
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_add_q15.c Normal file
View file

@ -0,0 +1,175 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_add_q15.c
* Description: Q15 vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAdd
@{
*/
/**
@brief Q15 vector addition.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
Results outside of the allowable Q15 range [0x8000 0x7FFF] are saturated.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_add_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
q15_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* loop counters */
q15x8_t vecA;
q15x8_t vecB;
/* Compute 8 outputs at a time */
blkCnt = blockSize >> 3;
while (blkCnt > 0U)
{
/*
* C = A + B
* Add and then store the results in the destination buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vst1q(pDst, vqaddq(vecA, vecB));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 8;
pSrcB += 8;
pDst += 8;
}
/*
* tail
*/
blkCnt = blockSize & 7;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp16q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vstrhq_p(pDst, vqaddq(vecA, vecB), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_add_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
q15_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined (ARM_MATH_LOOPUNROLL)
#if defined (ARM_MATH_DSP)
q31_t inA1, inA2;
q31_t inB1, inB2;
#endif
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A + B */
#if defined (ARM_MATH_DSP)
/* read 2 times 2 samples at a time from sourceA */
inA1 = read_q15x2_ia (&pSrcA);
inA2 = read_q15x2_ia (&pSrcA);
/* read 2 times 2 samples at a time from sourceB */
inB1 = read_q15x2_ia (&pSrcB);
inB2 = read_q15x2_ia (&pSrcB);
/* Add and store 2 times 2 samples at a time */
write_q15x2_ia (&pDst, __QADD16(inA1, inB1));
write_q15x2_ia (&pDst, __QADD16(inA2, inB2));
#else
*pDst++ = (q15_t) __SSAT(((q31_t) *pSrcA++ + *pSrcB++), 16);
*pDst++ = (q15_t) __SSAT(((q31_t) *pSrcA++ + *pSrcB++), 16);
*pDst++ = (q15_t) __SSAT(((q31_t) *pSrcA++ + *pSrcB++), 16);
*pDst++ = (q15_t) __SSAT(((q31_t) *pSrcA++ + *pSrcB++), 16);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
#if defined (ARM_MATH_DSP)
*pDst++ = (q15_t) __QADD16(*pSrcA++, *pSrcB++);
#else
*pDst++ = (q15_t) __SSAT(((q31_t) *pSrcA++ + *pSrcB++), 16);
#endif
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicAdd group
*/

158
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_add_q31.c Normal file
View file

@ -0,0 +1,158 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_add_q31.c
* Description: Q31 vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAdd
@{
*/
/**
@brief Q31 vector addition.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
Results outside of the allowable Q31 range [0x80000000 0x7FFFFFFF] are saturated.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_add_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt;
q31x4_t vecA;
q31x4_t vecB;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2;
while (blkCnt > 0U)
{
/*
* C = A + B
* Add and then store the results in the destination buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vst1q(pDst, vqaddq(vecA, vecB));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 4;
pSrcB += 4;
pDst += 4;
}
/*
* tail
*/
blkCnt = blockSize & 3;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp32q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vstrwq_p(pDst, vqaddq(vecA, vecB), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_add_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
*pDst++ = __QADD(*pSrcA++, *pSrcB++);
*pDst++ = __QADD(*pSrcA++, *pSrcB++);
*pDst++ = __QADD(*pSrcA++, *pSrcB++);
*pDst++ = __QADD(*pSrcA++, *pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
*pDst++ = __QADD(*pSrcA++, *pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicAdd group
*/

158
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_add_q7.c Normal file
View file

@ -0,0 +1,158 @@
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_add_q7.c
* Description: Q7 vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicAdd
@{
*/
/**
@brief Q7 vector addition.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
Results outside of the allowable Q7 range [0x80 0x7F] are saturated.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_add_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
q7_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* loop counters */
q7x16_t vecA;
q7x16_t vecB;
/* Compute 16 outputs at a time */
blkCnt = blockSize >> 4;
while (blkCnt > 0U)
{
/*
* C = A + B
* Add and then store the results in the destination buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vst1q(pDst, vqaddq(vecA, vecB));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 16;
pSrcB += 16;
pDst += 16;
}
/*
* tail
*/
blkCnt = blockSize & 0xF;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp8q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vstrbq_p(pDst, vqaddq(vecA, vecB), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_add_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
q7_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A + B */
#if defined (ARM_MATH_DSP)
/* Add and store result in destination buffer (4 samples at a time). */
write_q7x4_ia (&pDst, __QADD8 (read_q7x4_ia (&pSrcA), read_q7x4_ia (&pSrcB)));
#else
*pDst++ = (q7_t) __SSAT ((q15_t) *pSrcA++ + *pSrcB++, 8);
*pDst++ = (q7_t) __SSAT ((q15_t) *pSrcA++ + *pSrcB++, 8);
*pDst++ = (q7_t) __SSAT ((q15_t) *pSrcA++ + *pSrcB++, 8);
*pDst++ = (q7_t) __SSAT ((q15_t) *pSrcA++ + *pSrcB++, 8);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and store result in destination buffer. */
*pDst++ = (q7_t) __SSAT((q15_t) *pSrcA++ + *pSrcB++, 8);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicAdd group
*/

136
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_and_u16.c Normal file
View file

@ -0,0 +1,136 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_and_u16.c
* Description: uint16_t bitwise AND
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@defgroup And Vector bitwise AND
Compute the logical bitwise AND.
There are separate functions for uint32_t, uint16_t, and uint7_t data types.
*/
/**
@addtogroup And
@{
*/
/**
@brief Compute the logical bitwise AND of two fixed-point vectors.
@param[in] pSrcA points to input vector A
@param[in] pSrcB points to input vector B
@param[out] pDst points to output vector
@param[in] blockSize number of samples in each vector
*/
ARM_DSP_ATTRIBUTE void arm_and_u16(
const uint16_t * pSrcA,
const uint16_t * pSrcB,
uint16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
uint16x8_t vecSrcA, vecSrcB;
/* Compute 8 outputs at a time */
blkCnt = blockSize >> 3;
while (blkCnt > 0U)
{
vecSrcA = vld1q(pSrcA);
vecSrcB = vld1q(pSrcB);
vst1q(pDst, vandq_u16(vecSrcA, vecSrcB) );
pSrcA += 8;
pSrcB += 8;
pDst += 8;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 7;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp16q(blkCnt);
vecSrcA = vld1q(pSrcA);
vecSrcB = vld1q(pSrcB);
vstrhq_p(pDst, vandq_u16(vecSrcA, vecSrcB), p0);
}
#else
#if defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
uint16x8_t vecA, vecB;
/* Compute 8 outputs at a time */
blkCnt = blockSize >> 3U;
while (blkCnt > 0U)
{
vecA = vld1q_u16(pSrcA);
vecB = vld1q_u16(pSrcB);
vst1q_u16(pDst, vandq_u16(vecA, vecB) );
pSrcA += 8;
pSrcB += 8;
pDst += 8;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 7;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif
while (blkCnt > 0U)
{
*pDst++ = (*pSrcA++)&(*pSrcB++);
/* Decrement the loop counter */
blkCnt--;
}
#endif /* if defined(ARM_MATH_MVEI) */
}
/**
@} end of And group
*/

128
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_and_u32.c Normal file
View file

@ -0,0 +1,128 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_and_u32.c
* Description: uint32_t bitwise AND
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup And
@{
*/
/**
@brief Compute the logical bitwise AND of two fixed-point vectors.
@param[in] pSrcA points to input vector A
@param[in] pSrcB points to input vector B
@param[out] pDst points to output vector
@param[in] blockSize number of samples in each vector
*/
ARM_DSP_ATTRIBUTE void arm_and_u32(
const uint32_t * pSrcA,
const uint32_t * pSrcB,
uint32_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
uint32x4_t vecSrcA, vecSrcB;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2;
while (blkCnt > 0U)
{
vecSrcA = vld1q(pSrcA);
vecSrcB = vld1q(pSrcB);
vst1q(pDst, vandq_u32(vecSrcA, vecSrcB) );
pSrcA += 4;
pSrcB += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 3;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp32q(blkCnt);
vecSrcA = vld1q(pSrcA);
vecSrcB = vld1q(pSrcB);
vstrwq_p(pDst, vandq_u32(vecSrcA, vecSrcB), p0);
}
#else
#if defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
uint32x4_t vecA, vecB;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
vecA = vld1q_u32(pSrcA);
vecB = vld1q_u32(pSrcB);
vst1q_u32(pDst, vandq_u32(vecA, vecB) );
pSrcA += 4;
pSrcB += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 3;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif
while (blkCnt > 0U)
{
*pDst++ = (*pSrcA++)&(*pSrcB++);
/* Decrement the loop counter */
blkCnt--;
}
#endif /* if defined(ARM_MATH_MVEI) */
}
/**
@} end of And group
*/

129
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_and_u8.c Normal file
View file

@ -0,0 +1,129 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_and_u8.c
* Description: uint8_t bitwise AND
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup And
@{
*/
/**
@brief Compute the logical bitwise AND of two fixed-point vectors.
@param[in] pSrcA points to input vector A
@param[in] pSrcB points to input vector B
@param[out] pDst points to output vector
@param[in] blockSize number of samples in each vector
*/
ARM_DSP_ATTRIBUTE void arm_and_u8(
const uint8_t * pSrcA,
const uint8_t * pSrcB,
uint8_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
uint8x16_t vecSrcA, vecSrcB;
/* Compute 16 outputs at a time */
blkCnt = blockSize >> 4;
while (blkCnt > 0U)
{
vecSrcA = vld1q(pSrcA);
vecSrcB = vld1q(pSrcB);
vst1q(pDst, vandq_u8(vecSrcA, vecSrcB) );
pSrcA += 16;
pSrcB += 16;
pDst += 16;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0xF;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp8q(blkCnt);
vecSrcA = vld1q(pSrcA);
vecSrcB = vld1q(pSrcB);
vstrbq_p(pDst, vandq_u8(vecSrcA, vecSrcB), p0);
}
#else
#if defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
uint8x16_t vecA, vecB;
/* Compute 16 outputs at a time */
blkCnt = blockSize >> 4U;
while (blkCnt > 0U)
{
vecA = vld1q_u8(pSrcA);
vecB = vld1q_u8(pSrcB);
vst1q_u8(pDst, vandq_u8(vecA, vecB) );
pSrcA += 16;
pSrcB += 16;
pDst += 16;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0xF;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif
while (blkCnt > 0U)
{
*pDst++ = (*pSrcA++)&(*pSrcB++);
/* Decrement the loop counter */
blkCnt--;
}
#endif /* if defined(ARM_MATH_MVEI) */
}
/**
@} end of And group
*/

140
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_clip_f16.c Normal file
View file

@ -0,0 +1,140 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_clip_f16.c
* Description: Floating-point vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions_f16.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicClip
@{
*/
/**
@brief Elementwise floating-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_clip_f16(const float16_t * pSrc,
float16_t * pDst,
float16_t low,
float16_t high,
uint32_t numSamples)
{
uint32_t blkCnt;
f16x8_t curVec0, curVec1;
f16x8_t vecLow, vecHigh;
vecLow = vdupq_n_f16(low);
vecHigh = vdupq_n_f16(high);
curVec0 = vld1q(pSrc);
pSrc += 8;
/*
* unrolled x 2 to allow
* vldr/vstr/vmin/vmax
* stall free interleaving
*/
blkCnt = numSamples >> 4;
while (blkCnt--)
{
curVec0 = vmaxnmq(curVec0, vecLow);
curVec1 = vld1q(pSrc);
pSrc += 8;
curVec0 = vminnmq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 8;
curVec1 = vmaxnmq(curVec1, vecLow);
curVec0 = vld1q(pSrc);
pSrc += 8;
curVec1 = vminnmq(curVec1, vecHigh);
vst1q(pDst, curVec1);
pDst += 8;
}
/*
* Tail handling
*/
blkCnt = numSamples - ((numSamples >> 4) << 4);
if (blkCnt >= 8)
{
curVec0 = vmaxnmq(curVec0, vecLow);
curVec0 = vminnmq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 8;
curVec0 = vld1q(pSrc);
pSrc += 8;
}
if (blkCnt > 0)
{
mve_pred16_t p0 = vctp16q(blkCnt & 7);
curVec0 = vmaxnmq(curVec0, vecLow);
curVec0 = vminnmq(curVec0, vecHigh);
vstrhq_p(pDst, curVec0, p0);
}
}
#else
#if defined(ARM_FLOAT16_SUPPORTED)
ARM_DSP_ATTRIBUTE void arm_clip_f16(const float16_t * pSrc,
float16_t * pDst,
float16_t low,
float16_t high,
uint32_t numSamples)
{
for (uint32_t i = 0; i < numSamples; i++)
{
if ((_Float16)pSrc[i] > (_Float16)high)
pDst[i] = high;
else if ((_Float16)pSrc[i] < (_Float16)low)
pDst[i] = low;
else
pDst[i] = pSrc[i];
}
}
#endif /* defined(ARM_FLOAT16_SUPPORTED */
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicClip group
*/

143
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_clip_f32.c Normal file
View file

@ -0,0 +1,143 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_clip_f32.c
* Description: Floating-point vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@defgroup BasicClip Elementwise clipping
Element-by-element clipping of a value.
The value is constrained between 2 bounds.
There are separate functions for floating-point, Q7, Q15, and Q31 data types.
*/
/**
@addtogroup BasicClip
@{
*/
/**
@brief Elementwise floating-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_clip_f32(const float32_t * pSrc,
float32_t * pDst,
float32_t low,
float32_t high,
uint32_t numSamples)
{
uint32_t blkCnt;
f32x4_t curVec0, curVec1;
f32x4_t vecLow, vecHigh;
vecLow = vdupq_n_f32(low);
vecHigh = vdupq_n_f32(high);
curVec0 = vld1q(pSrc);
pSrc += 4;
/*
* unrolled x 2 to allow
* vldr/vstr/vmin/vmax
* stall free interleaving
*/
blkCnt = numSamples >> 3;
while (blkCnt--)
{
curVec0 = vmaxnmq(curVec0, vecLow);
curVec1 = vld1q(pSrc);
pSrc += 4;
curVec0 = vminnmq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 4;
curVec1 = vmaxnmq(curVec1, vecLow);
curVec0 = vld1q(pSrc);
pSrc += 4;
curVec1 = vminnmq(curVec1, vecHigh);
vst1q(pDst, curVec1);
pDst += 4;
}
/*
* Tail handling
*/
blkCnt = numSamples - ((numSamples >> 3) << 3);
if (blkCnt >= 4)
{
curVec0 = vmaxnmq(curVec0, vecLow);
curVec0 = vminnmq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 4;
curVec0 = vld1q(pSrc);
pSrc += 4;
}
if (blkCnt > 0)
{
mve_pred16_t p0 = vctp32q(blkCnt & 3);
curVec0 = vmaxnmq(curVec0, vecLow);
curVec0 = vminnmq(curVec0, vecHigh);
vstrwq_p(pDst, curVec0, p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_clip_f32(const float32_t * pSrc,
float32_t * pDst,
float32_t low,
float32_t high,
uint32_t numSamples)
{
uint32_t i;
for (i = 0; i < numSamples; i++)
{
if (pSrc[i] > high)
pDst[i] = high;
else if (pSrc[i] < low)
pDst[i] = low;
else
pDst[i] = pSrc[i];
}
}
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicClip group
*/

133
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_clip_q15.c Normal file
View file

@ -0,0 +1,133 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_clip_q15.c
* Description: Floating-point vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicClip
@{
*/
/**
@brief Elementwise fixed-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_clip_q15(const q15_t * pSrc,
q15_t * pDst,
q15_t low,
q15_t high,
uint32_t numSamples)
{
uint32_t blkCnt;
q15x8_t curVec0, curVec1;
q15x8_t vecLow, vecHigh;
vecLow = vdupq_n_s16(low);
vecHigh = vdupq_n_s16(high);
curVec0 = vld1q(pSrc);
pSrc += 8;
/*
* unrolled x 2 to allow
* vldr/vstr/vmin/vmax
* stall free interleaving
*/
blkCnt = numSamples >> 4;
while (blkCnt--)
{
curVec0 = vmaxq(curVec0, vecLow);
curVec1 = vld1q(pSrc);
pSrc += 8;
curVec0 = vminq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 8;
curVec1 = vmaxq(curVec1, vecLow);
curVec0 = vld1q(pSrc);
pSrc += 8;
curVec1 = vminq(curVec1, vecHigh);
vst1q(pDst, curVec1);
pDst += 8;
}
/*
* Tail handling
*/
blkCnt = numSamples - ((numSamples >> 4) << 4);
if (blkCnt >= 8)
{
curVec0 = vmaxq(curVec0, vecLow);
curVec0 = vminq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 8;
curVec0 = vld1q(pSrc);
pSrc += 8;
}
if (blkCnt > 0)
{
mve_pred16_t p0 = vctp16q(blkCnt & 7);
curVec0 = vmaxq(curVec0, vecLow);
curVec0 = vminq(curVec0, vecHigh);
vstrhq_p(pDst, curVec0, p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_clip_q15(const q15_t * pSrc,
q15_t * pDst,
q15_t low,
q15_t high,
uint32_t numSamples)
{
uint32_t i;
for (i = 0; i < numSamples; i++)
{
if (pSrc[i] > high)
pDst[i] = high;
else if (pSrc[i] < low)
pDst[i] = low;
else
pDst[i] = pSrc[i];
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicClip group
*/

133
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_clip_q31.c Normal file
View file

@ -0,0 +1,133 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_clip_q31.c
* Description: Floating-point vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicClip
@{
*/
/**
@brief Elementwise fixed-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_clip_q31(const q31_t * pSrc,
q31_t * pDst,
q31_t low,
q31_t high,
uint32_t numSamples)
{
uint32_t blkCnt;
q31x4_t curVec0, curVec1;
q31x4_t vecLow, vecHigh;
vecLow = vdupq_n_s32(low);
vecHigh = vdupq_n_s32(high);
curVec0 = vld1q(pSrc);
pSrc += 4;
/*
* unrolled x 2 to allow
* vldr/vstr/vmin/vmax
* stall free interleaving
*/
blkCnt = numSamples >> 3;
while (blkCnt--)
{
curVec0 = vmaxq(curVec0, vecLow);
curVec1 = vld1q(pSrc);
pSrc += 4;
curVec0 = vminq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 4;
curVec1 = vmaxq(curVec1, vecLow);
curVec0 = vld1q(pSrc);
pSrc += 4;
curVec1 = vminq(curVec1, vecHigh);
vst1q(pDst, curVec1);
pDst += 4;
}
/*
* Tail handling
*/
blkCnt = numSamples - ((numSamples >> 3) << 3);
if (blkCnt >= 4)
{
curVec0 = vmaxq(curVec0, vecLow);
curVec0 = vminq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 4;
curVec0 = vld1q(pSrc);
pSrc += 4;
}
if (blkCnt > 0)
{
mve_pred16_t p0 = vctp32q(blkCnt & 3);
curVec0 = vmaxq(curVec0, vecLow);
curVec0 = vminq(curVec0, vecHigh);
vstrwq_p(pDst, curVec0, p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_clip_q31(const q31_t * pSrc,
q31_t * pDst,
q31_t low,
q31_t high,
uint32_t numSamples)
{
uint32_t i;
for (i = 0; i < numSamples; i++)
{
if (pSrc[i] > high)
pDst[i] = high;
else if (pSrc[i] < low)
pDst[i] = low;
else
pDst[i] = pSrc[i];
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicClip group
*/

133
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_clip_q7.c Normal file
View file

@ -0,0 +1,133 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_clip_q7.c
* Description: Floating-point vector addition
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicClip
@{
*/
/**
@brief Elementwise fixed-point clipping
@param[in] pSrc points to input values
@param[out] pDst points to output clipped values
@param[in] low lower bound
@param[in] high higher bound
@param[in] numSamples number of samples to clip
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_clip_q7(const q7_t * pSrc,
q7_t * pDst,
q7_t low,
q7_t high,
uint32_t numSamples)
{
uint32_t blkCnt;
q7x16_t curVec0, curVec1;
q7x16_t vecLow, vecHigh;
vecLow = vdupq_n_s8(low);
vecHigh = vdupq_n_s8(high);
curVec0 = vld1q(pSrc);
pSrc += 16;
/*
* unrolled x 2 to allow
* vldr/vstr/vmin/vmax
* stall free interleaving
*/
blkCnt = numSamples >> 5;
while (blkCnt--)
{
curVec0 = vmaxq(curVec0, vecLow);
curVec1 = vld1q(pSrc);
pSrc += 16;
curVec0 = vminq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 16;
curVec1 = vmaxq(curVec1, vecLow);
curVec0 = vld1q(pSrc);
pSrc += 16;
curVec1 = vminq(curVec1, vecHigh);
vst1q(pDst, curVec1);
pDst += 16;
}
/*
* Tail handling
*/
blkCnt = numSamples - ((numSamples >> 5) << 5);
if (blkCnt >= 16)
{
curVec0 = vmaxq(curVec0, vecLow);
curVec0 = vminq(curVec0, vecHigh);
vst1q(pDst, curVec0);
pDst += 16;
curVec0 = vld1q(pSrc);
pSrc += 16;
}
if (blkCnt > 0)
{
mve_pred16_t p0 = vctp8q(blkCnt & 0xf);
curVec0 = vmaxq(curVec0, vecLow);
curVec0 = vminq(curVec0, vecHigh);
vstrbq_p(pDst, curVec0, p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_clip_q7(const q7_t * pSrc,
q7_t * pDst,
q7_t low,
q7_t high,
uint32_t numSamples)
{
uint32_t i;
for (i = 0; i < numSamples; i++)
{
if (pSrc[i] > high)
pDst[i] = high;
else if (pSrc[i] < low)
pDst[i] = low;
else
pDst[i] = pSrc[i];
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicClip group
*/

171
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_dot_prod_f16.c Normal file
View file

@ -0,0 +1,171 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_dot_prod_f16.c
* Description: Floating-point dot product
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions_f16.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicDotProd
@{
*/
/**
@brief Dot product of floating-point vectors.
@param[in] pSrcA points to the first input vector.
@param[in] pSrcB points to the second input vector.
@param[in] blockSize number of samples in each vector.
@param[out] result output result returned here.
*/
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_dot_prod_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
uint32_t blockSize,
float16_t * result)
{
f16x8_t vecA, vecB;
f16x8_t vecSum;
uint32_t blkCnt;
float16_t sum = 0.0f;
vecSum = vdupq_n_f16(0.0f);
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 3U;
while (blkCnt > 0U)
{
/*
* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1]
* Calculate dot product and then store the result in a temporary buffer.
* and advance vector source and destination pointers
*/
vecA = vld1q(pSrcA);
pSrcA += 8;
vecB = vld1q(pSrcB);
pSrcB += 8;
vecSum = vfmaq(vecSum, vecA, vecB);
/*
* Decrement the blockSize loop counter
*/
blkCnt --;
}
blkCnt = blockSize & 7;
if (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
mve_pred16_t p0 = vctp16q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vecSum = vfmaq_m(vecSum, vecA, vecB, p0);
}
sum = vecAddAcrossF16Mve(vecSum);
/* Store result in destination buffer */
*result = sum;
}
#else
#if defined(ARM_FLOAT16_SUPPORTED)
ARM_DSP_ATTRIBUTE void arm_dot_prod_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
uint32_t blockSize,
float16_t * result)
{
uint32_t blkCnt; /* Loop counter */
_Float16 sum = 0.0f; /* Temporary return variable */
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
/* First part of the processing with loop unrolling. Compute 4 outputs at a time.
** a second loop below computes the remaining 1 to 3 samples. */
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
sum += (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
sum += (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
sum += (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
sum += (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
sum += (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Store result in destination buffer */
*result = sum;
}
#endif
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicDotProd group
*/

227
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_dot_prod_f32.c Normal file
View file

@ -0,0 +1,227 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_dot_prod_f32.c
* Description: Floating-point dot product
*
* $Date: 05 October 2021
* $Revision: V1.9.1
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@defgroup BasicDotProd Vector Dot Product
Computes the dot product of two vectors.
The vectors are multiplied element-by-element and then summed.
<pre>
sum = pSrcA[0]*pSrcB[0] + pSrcA[1]*pSrcB[1] + ... + pSrcA[blockSize-1]*pSrcB[blockSize-1]
</pre>
There are separate functions for floating-point, Q7, Q15, and Q31 data types.
*/
/**
@addtogroup BasicDotProd
@{
*/
/**
@brief Dot product of floating-point vectors.
@param[in] pSrcA points to the first input vector.
@param[in] pSrcB points to the second input vector.
@param[in] blockSize number of samples in each vector.
@param[out] result output result returned here.
*/
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_dot_prod_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
uint32_t blockSize,
float32_t * result)
{
f32x4_t vecA, vecB;
f32x4_t vecSum;
uint32_t blkCnt;
float32_t sum = 0.0f;
vecSum = vdupq_n_f32(0.0f);
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/*
* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1]
* Calculate dot product and then store the result in a temporary buffer.
* and advance vector source and destination pointers
*/
vecA = vld1q(pSrcA);
pSrcA += 4;
vecB = vld1q(pSrcB);
pSrcB += 4;
vecSum = vfmaq(vecSum, vecA, vecB);
/*
* Decrement the blockSize loop counter
*/
blkCnt --;
}
blkCnt = blockSize & 3;
if (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
mve_pred16_t p0 = vctp32q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vecSum = vfmaq_m(vecSum, vecA, vecB, p0);
}
sum = vecAddAcrossF32Mve(vecSum);
/* Store result in destination buffer */
*result = sum;
}
#else
ARM_DSP_ATTRIBUTE void arm_dot_prod_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
uint32_t blockSize,
float32_t * result)
{
uint32_t blkCnt; /* Loop counter */
float32_t sum = 0.0f; /* Temporary return variable */
#if defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
f32x4_t vec1;
f32x4_t vec2;
f32x4_t accum = vdupq_n_f32(0);
#if !defined(__aarch64__)
f32x2_t tmp = vdup_n_f32(0);
#endif
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
vec1 = vld1q_f32(pSrcA);
vec2 = vld1q_f32(pSrcB);
while (blkCnt > 0U)
{
/* C = A[0]*B[0] + A[1]*B[1] + A[2]*B[2] + ... + A[blockSize-1]*B[blockSize-1] */
/* Calculate dot product and then store the result in a temporary buffer. */
accum = vmlaq_f32(accum, vec1, vec2);
/* Increment pointers */
pSrcA += 4;
pSrcB += 4;
vec1 = vld1q_f32(pSrcA);
vec2 = vld1q_f32(pSrcB);
/* Decrement the loop counter */
blkCnt--;
}
#if defined(__aarch64__)
sum = vpadds_f32(vpadd_f32(vget_low_f32(accum), vget_high_f32(accum)));
#else
tmp = vpadd_f32(vget_low_f32(accum), vget_high_f32(accum));
sum = vget_lane_f32(tmp, 0) + vget_lane_f32(tmp, 1);
#endif
/* Tail */
blkCnt = blockSize & 0x3;
#else
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
/* First part of the processing with loop unrolling. Compute 4 outputs at a time.
** a second loop below computes the remaining 1 to 3 samples. */
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
sum += (*pSrcA++) * (*pSrcB++);
sum += (*pSrcA++) * (*pSrcB++);
sum += (*pSrcA++) * (*pSrcB++);
sum += (*pSrcA++) * (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
#endif /* #if defined(ARM_MATH_NEON) */
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
sum += (*pSrcA++) * (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Store result in destination buffer */
*result = sum;
}
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicDotProd group
*/

131
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_dot_prod_f64.c Normal file
View file

@ -0,0 +1,131 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_dot_prod_f64.c
* Description: Floating-point dot product
*
* $Date: 03 June 2022
* $Revision: V1.10.1
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicDotProd
@{
*/
/**
@brief Dot product of floating-point vectors.
@param[in] pSrcA points to the first input vector.
@param[in] pSrcB points to the second input vector.
@param[in] blockSize number of samples in each vector.
@param[out] result output result returned here.
*/
#if defined(ARM_MATH_NEON) && defined(__aarch64__)
ARM_DSP_ATTRIBUTE void arm_dot_prod_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
uint32_t blockSize,
float64_t * result)
{
uint32_t blkCnt; /* Loop counter */
float64_t sum = 0.; /* Temporary return variable */
/* Neon Buffer Initialisation */
float64x2_t sumV = vdupq_n_f64(0.0); /* Neon buffer for sum variable */
/* Neon Buffer for sources */
float64x2_t pSrcAV;
float64x2_t pSrcBV;
/* Initialize blkCnt with number of samples */
blkCnt = blockSize >> 1U;
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Load source value in Neon Buffer */
pSrcAV = vld1q_f64(pSrcA);
pSrcBV = vld1q_f64(pSrcB);
/* Calculate dot product and store result in a temporary buffer. */
sumV = vmlaq_f64(sumV, pSrcAV, pSrcBV);
pSrcA+=2;
pSrcB+=2;
/* Decrement loop counter */
blkCnt--;
}
/* Sum both 64 bits part in the float64x2 */
sum = vaddvq_f64(sumV);
/* Tail */
blkCnt = blockSize & 1 ;
while(blkCnt > 0U)
{
sum += (*pSrcA++) * (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Store result in destination buffer */
*result = sum;
}
#else
ARM_DSP_ATTRIBUTE void arm_dot_prod_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
uint32_t blockSize,
float64_t * result)
{
uint32_t blkCnt; /* Loop counter */
float64_t sum = 0.; /* Temporary return variable */
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
sum += (*pSrcA++) * (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Store result in destination buffer */
*result = sum;
}
#endif
/**
@} end of BasicDotProd group
*/

167
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_dot_prod_q15.c Normal file
View file

@ -0,0 +1,167 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_dot_prod_q15.c
* Description: Q15 dot product
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicDotProd
@{
*/
/**
@brief Dot product of Q15 vectors.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[in] blockSize number of samples in each vector
@param[out] result output result returned here
@par Scaling and Overflow Behavior
The intermediate multiplications are in 1.15 x 1.15 = 2.30 format and these
results are added to a 64-bit accumulator in 34.30 format.
Nonsaturating additions are used and given that there are 33 guard bits in the accumulator
there is no risk of overflow.
The return result is in 34.30 format.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_dot_prod_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
uint32_t blockSize,
q63_t * result)
{
uint32_t blkCnt; /* loop counters */
q15x8_t vecA;
q15x8_t vecB;
q63_t sum = 0LL;
/* Compute 8 outputs at a time */
blkCnt = blockSize >> 3;
while (blkCnt > 0U)
{
/*
* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1]
* Calculate dot product and then store the result in a temporary buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
sum = vmlaldavaq(sum, vecA, vecB);
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 8;
pSrcB += 8;
}
/*
* tail
*/
blkCnt = blockSize & 7;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp16q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
sum = vmlaldavaq_p(sum, vecA, vecB, p0);
}
*result = sum;
}
#else
ARM_DSP_ATTRIBUTE void arm_dot_prod_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
uint32_t blockSize,
q63_t * result)
{
uint32_t blkCnt; /* Loop counter */
q63_t sum = 0; /* Temporary return variable */
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
#if defined (ARM_MATH_DSP)
/* Calculate dot product and store result in a temporary buffer. */
sum = __SMLALD(read_q15x2_ia (&pSrcA), read_q15x2_ia (&pSrcB), sum);
sum = __SMLALD(read_q15x2_ia (&pSrcA), read_q15x2_ia (&pSrcB), sum);
#else
sum += (q63_t)((q31_t) *pSrcA++ * *pSrcB++);
sum += (q63_t)((q31_t) *pSrcA++ * *pSrcB++);
sum += (q63_t)((q31_t) *pSrcA++ * *pSrcB++);
sum += (q63_t)((q31_t) *pSrcA++ * *pSrcB++);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
sum += (q63_t)((q31_t) *pSrcA++ * *pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Store result in destination buffer in 34.30 format */
*result = sum;
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicDotProd group
*/

173
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_dot_prod_q31.c Normal file
View file

@ -0,0 +1,173 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_dot_prod_q31.c
* Description: Q31 dot product
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicDotProd
@{
*/
/**
@brief Dot product of Q31 vectors.
@param[in] pSrcA points to the first input vector.
@param[in] pSrcB points to the second input vector.
@param[in] blockSize number of samples in each vector.
@param[out] result output result returned here.
@par Scaling and Overflow Behavior
The intermediate multiplications are in 1.31 x 1.31 = 2.62 format and these
are truncated to 2.48 format by discarding the lower 14 bits.
The 2.48 result is then added without saturation to a 64-bit accumulator in 16.48 format.
There are 15 guard bits in the accumulator and there is no risk of overflow as long as
the length of the vectors is less than 2^16 elements.
The return result is in 16.48 format.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_dot_prod_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
uint32_t blockSize,
q63_t * result)
{
uint32_t blkCnt; /* loop counters */
q31x4_t vecA;
q31x4_t vecB;
q63_t sum = 0LL;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2;
while (blkCnt > 0U)
{
/*
* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1]
* Calculate dot product and then store the result in a temporary buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
sum = vrmlaldavhaq(sum, vecA, vecB);
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 4;
pSrcB += 4;
}
/*
* tail
*/
blkCnt = blockSize & 3;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp32q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
sum = vrmlaldavhaq_p(sum, vecA, vecB, p0);
}
/*
* vrmlaldavhaq provides extra intermediate accumulator headroom.
* limiting the need of intermediate scaling
* Scalar variant uses 2.48 accu format by right shifting accumulators by 14.
* 16.48 output conversion is performed outside the loop by scaling accu. by 6
*/
*result = asrl(sum, (14 - 8));
}
#else
ARM_DSP_ATTRIBUTE void arm_dot_prod_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
uint32_t blockSize,
q63_t * result)
{
uint32_t blkCnt; /* Loop counter */
q63_t sum = 0; /* Temporary return variable */
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
sum += ((q63_t) *pSrcA++ * *pSrcB++) >> 14U;
sum += ((q63_t) *pSrcA++ * *pSrcB++) >> 14U;
sum += ((q63_t) *pSrcA++ * *pSrcB++) >> 14U;
sum += ((q63_t) *pSrcA++ * *pSrcB++) >> 14U;
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
sum += ((q63_t) *pSrcA++ * *pSrcB++) >> 14U;
/* Decrement loop counter */
blkCnt--;
}
/* Store result in destination buffer in 16.48 format */
*result = sum;
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicDotProd group
*/

190
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_dot_prod_q7.c Normal file
View file

@ -0,0 +1,190 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_dot_prod_q7.c
* Description: Q7 dot product
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicDotProd
@{
*/
/**
@brief Dot product of Q7 vectors.
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[in] blockSize number of samples in each vector
@param[out] result output result returned here
@par Scaling and Overflow Behavior
The intermediate multiplications are in 1.7 x 1.7 = 2.14 format and these
results are added to an accumulator in 18.14 format.
Nonsaturating additions are used and there is no danger of wrap around as long as
the vectors are less than 2^18 elements long.
The return result is in 18.14 format.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_dot_prod_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
uint32_t blockSize,
q31_t * result)
{
uint32_t blkCnt; /* loop counters */
q7x16_t vecA;
q7x16_t vecB;
q31_t sum = 0;
/* Compute 16 outputs at a time */
blkCnt = blockSize >> 4;
while (blkCnt > 0U)
{
/*
* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1]
* Calculate dot product and then store the result in a temporary buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
sum = vmladavaq(sum, vecA, vecB);
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 16;
pSrcB += 16;
}
/*
* tail
*/
blkCnt = blockSize & 0xF;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp8q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
sum = vmladavaq_p(sum, vecA, vecB, p0);
}
*result = sum;
}
#else
ARM_DSP_ATTRIBUTE void arm_dot_prod_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
uint32_t blockSize,
q31_t * result)
{
uint32_t blkCnt; /* Loop counter */
q31_t sum = 0; /* Temporary return variable */
#if defined (ARM_MATH_LOOPUNROLL)
#if defined (ARM_MATH_DSP)
q31_t input1, input2; /* Temporary variables */
q31_t inA1, inA2, inB1, inB2; /* Temporary variables */
#endif
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
#if defined (ARM_MATH_DSP)
/* read 4 samples at a time from sourceA */
input1 = read_q7x4_ia (&pSrcA);
/* read 4 samples at a time from sourceB */
input2 = read_q7x4_ia (&pSrcB);
/* extract two q7_t samples to q15_t samples */
inA1 = __SXTB16(__ROR(input1, 8));
/* extract reminaing two samples */
inA2 = __SXTB16(input1);
/* extract two q7_t samples to q15_t samples */
inB1 = __SXTB16(__ROR(input2, 8));
/* extract reminaing two samples */
inB2 = __SXTB16(input2);
/* multiply and accumulate two samples at a time */
sum = __SMLAD(inA1, inB1, sum);
sum = __SMLAD(inA2, inB2, sum);
#else
sum += (q31_t) ((q15_t) *pSrcA++ * *pSrcB++);
sum += (q31_t) ((q15_t) *pSrcA++ * *pSrcB++);
sum += (q31_t) ((q15_t) *pSrcA++ * *pSrcB++);
sum += (q31_t) ((q15_t) *pSrcA++ * *pSrcB++);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
/* Calculate dot product and store result in a temporary buffer. */
//#if defined (ARM_MATH_DSP)
// sum = __SMLAD(*pSrcA++, *pSrcB++, sum);
//#else
sum += (q31_t) ((q15_t) *pSrcA++ * *pSrcB++);
//#endif
/* Decrement loop counter */
blkCnt--;
}
/* Store result in destination buffer in 18.14 format */
*result = sum;
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicDotProd group
*/

159
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_mult_f16.c Normal file
View file

@ -0,0 +1,159 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_mult_f16.c
* Description: Floating-point vector multiplication
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions_f16.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicMult
@{
*/
/**
@brief Floating-point vector multiplication.
@param[in] pSrcA points to the first input vector.
@param[in] pSrcB points to the second input vector.
@param[out] pDst points to the output vector.
@param[in] blockSize number of samples in each vector.
*/
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_mult_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
float16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
f16x8_t vec1;
f16x8_t vec2;
f16x8_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 3U;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and then store the results in the destination buffer. */
vec1 = vld1q(pSrcA);
vec2 = vld1q(pSrcB);
res = vmulq(vec1, vec2);
vst1q(pDst, res);
/* Increment pointers */
pSrcA += 8;
pSrcB += 8;
pDst += 8;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x7;
if (blkCnt > 0U)
{
/* C = A + B */
mve_pred16_t p0 = vctp16q(blkCnt);
vec1 = vld1q(pSrcA);
vec2 = vld1q(pSrcB);
vstrhq_p(pDst, vmulq(vec1,vec2), p0);
}
}
#else
#if defined(ARM_FLOAT16_SUPPORTED)
ARM_DSP_ATTRIBUTE void arm_mult_f16(
const float16_t * pSrcA,
const float16_t * pSrcB,
float16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply inputs and store result in destination buffer. */
*pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
*pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
*pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
*pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply input and store result in destination buffer. */
*pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
}
#endif
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicMult group
*/

199
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_mult_f32.c Normal file
View file

@ -0,0 +1,199 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_mult_f32.c
* Description: Floating-point vector multiplication
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@defgroup BasicMult Vector Multiplication
Element-by-element multiplication of two vectors.
<pre>
pDst[n] = pSrcA[n] * pSrcB[n], 0 <= n < blockSize.
</pre>
There are separate functions for floating-point, Q7, Q15, and Q31 data types.
*/
/**
@addtogroup BasicMult
@{
*/
/**
@brief Floating-point vector multiplication.
@param[in] pSrcA points to the first input vector.
@param[in] pSrcB points to the second input vector.
@param[out] pDst points to the output vector.
@param[in] blockSize number of samples in each vector.
*/
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_mult_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
float32_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
f32x4_t vec1;
f32x4_t vec2;
f32x4_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A + B */
/* Add and then store the results in the destination buffer. */
vec1 = vld1q(pSrcA);
vec2 = vld1q(pSrcB);
res = vmulq(vec1, vec2);
vst1q(pDst, res);
/* Increment pointers */
pSrcA += 4;
pSrcB += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x3;
if (blkCnt > 0U)
{
/* C = A + B */
mve_pred16_t p0 = vctp32q(blkCnt);
vec1 = vld1q(pSrcA);
vec2 = vld1q(pSrcB);
vstrwq_p(pDst, vmulq(vec1,vec2), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_mult_f32(
const float32_t * pSrcA,
const float32_t * pSrcB,
float32_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
f32x4_t vec1;
f32x4_t vec2;
f32x4_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply the inputs and then store the results in the destination buffer. */
vec1 = vld1q_f32(pSrcA);
vec2 = vld1q_f32(pSrcB);
res = vmulq_f32(vec1, vec2);
vst1q_f32(pDst, res);
/* Increment pointers */
pSrcA += 4;
pSrcB += 4;
pDst += 4;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x3;
#else
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply inputs and store result in destination buffer. */
*pDst++ = (*pSrcA++) * (*pSrcB++);
*pDst++ = (*pSrcA++) * (*pSrcB++);
*pDst++ = (*pSrcA++) * (*pSrcB++);
*pDst++ = (*pSrcA++) * (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
#endif /* #if defined(ARM_MATH_NEON) */
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply input and store result in destination buffer. */
*pDst++ = (*pSrcA++) * (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicMult group
*/

74
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_mult_f64.c Normal file
View file

@ -0,0 +1,74 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_mult_f64.c
* Description: Floating-point vector multiplication
*
* $Date: 13 September 2021
* $Revision: V1.10.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicMult
@{
*/
/**
@brief Floating-point vector multiplication.
@param[in] pSrcA points to the first input vector.
@param[in] pSrcB points to the second input vector.
@param[out] pDst points to the output vector.
@param[in] blockSize number of samples in each vector.
*/
ARM_DSP_ATTRIBUTE void arm_mult_f64(
const float64_t * pSrcA,
const float64_t * pSrcB,
float64_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply input and store result in destination buffer. */
*pDst++ = (*pSrcA++) * (*pSrcB++);
/* Decrement loop counter */
blkCnt--;
}
}
/**
@} end of BasicMult group
*/

191
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_mult_q15.c Normal file
View file

@ -0,0 +1,191 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_mult_q15.c
* Description: Q15 vector multiplication
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicMult
@{
*/
/**
@brief Q15 vector multiplication
@param[in] pSrcA points to first input vector
@param[in] pSrcB points to second input vector
@param[out] pDst points to output vector
@param[in] blockSize number of samples in each vector
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
Results outside of the allowable Q15 range [0x8000 0x7FFF] are saturated.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_mult_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
q15_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* loop counters */
q15x8_t vecA, vecB;
/* Compute 8 outputs at a time */
blkCnt = blockSize >> 3;
while (blkCnt > 0U)
{
/*
* C = A * B
* Multiply the inputs and then store the results in the destination buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vst1q(pDst, vqdmulhq(vecA, vecB));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 8;
pSrcB += 8;
pDst += 8;
}
/*
* tail
*/
blkCnt = blockSize & 7;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp16q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vstrhq_p(pDst, vqdmulhq(vecA, vecB), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_mult_q15(
const q15_t * pSrcA,
const q15_t * pSrcB,
q15_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined (ARM_MATH_LOOPUNROLL)
#if defined (ARM_MATH_DSP)
q31_t inA1, inA2, inB1, inB2; /* Temporary input variables */
q15_t out1, out2, out3, out4; /* Temporary output variables */
q31_t mul1, mul2, mul3, mul4; /* Temporary variables */
#endif
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A * B */
#if defined (ARM_MATH_DSP)
/* read 2 samples at a time from sourceA */
inA1 = read_q15x2_ia (&pSrcA);
/* read 2 samples at a time from sourceB */
inB1 = read_q15x2_ia (&pSrcB);
/* read 2 samples at a time from sourceA */
inA2 = read_q15x2_ia (&pSrcA);
/* read 2 samples at a time from sourceB */
inB2 = read_q15x2_ia (&pSrcB);
/* multiply mul = sourceA * sourceB */
mul1 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1 >> 16));
mul2 = (q31_t) ((q15_t) (inA1 ) * (q15_t) (inB1 ));
mul3 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) (inB2 >> 16));
mul4 = (q31_t) ((q15_t) (inA2 ) * (q15_t) (inB2 ));
/* saturate result to 16 bit */
out1 = (q15_t) __SSAT(mul1 >> 15, 16);
out2 = (q15_t) __SSAT(mul2 >> 15, 16);
out3 = (q15_t) __SSAT(mul3 >> 15, 16);
out4 = (q15_t) __SSAT(mul4 >> 15, 16);
/* store result to destination */
#ifndef ARM_MATH_BIG_ENDIAN
write_q15x2_ia (&pDst, __PKHBT(out2, out1, 16));
write_q15x2_ia (&pDst, __PKHBT(out4, out3, 16));
#else
write_q15x2_ia (&pDst, __PKHBT(out1, out2, 16));
write_q15x2_ia (&pDst, __PKHBT(out3, out4, 16));
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
#else
*pDst++ = (q15_t) __SSAT((((q31_t) (*pSrcA++) * (*pSrcB++)) >> 15), 16);
*pDst++ = (q15_t) __SSAT((((q31_t) (*pSrcA++) * (*pSrcB++)) >> 15), 16);
*pDst++ = (q15_t) __SSAT((((q31_t) (*pSrcA++) * (*pSrcB++)) >> 15), 16);
*pDst++ = (q15_t) __SSAT((((q31_t) (*pSrcA++) * (*pSrcB++)) >> 15), 16);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply inputs and store result in destination buffer. */
*pDst++ = (q15_t) __SSAT((((q31_t) (*pSrcA++) * (*pSrcB++)) >> 15), 16);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicMult group
*/

167
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_mult_q31.c Normal file
View file

@ -0,0 +1,167 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_mult_q31.c
* Description: Q31 vector multiplication
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicMult
@{
*/
/**
@brief Q31 vector multiplication.
@param[in] pSrcA points to the first input vector.
@param[in] pSrcB points to the second input vector.
@param[out] pDst points to the output vector.
@param[in] blockSize number of samples in each vector.
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
Results outside of the allowable Q31 range[0x80000000 0x7FFFFFFF] are saturated.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_mult_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* loop counters */
q31x4_t vecA, vecB;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 2;
while (blkCnt > 0U)
{
/*
* C = A * B
* Multiply the inputs and then store the results in the destination buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vst1q(pDst, vqdmulhq(vecA, vecB));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 4;
pSrcB += 4;
pDst += 4;
}
/*
* tail
*/
blkCnt = blockSize & 3;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp32q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vstrwq_p(pDst, vqdmulhq(vecA, vecB), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_mult_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
q31_t out; /* Temporary output variable */
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply inputs and store result in destination buffer. */
out = ((q63_t) *pSrcA++ * *pSrcB++) >> 32;
out = __SSAT(out, 31);
*pDst++ = out << 1U;
out = ((q63_t) *pSrcA++ * *pSrcB++) >> 32;
out = __SSAT(out, 31);
*pDst++ = out << 1U;
out = ((q63_t) *pSrcA++ * *pSrcB++) >> 32;
out = __SSAT(out, 31);
*pDst++ = out << 1U;
out = ((q63_t) *pSrcA++ * *pSrcB++) >> 32;
out = __SSAT(out, 31);
*pDst++ = out << 1U;
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply inputs and store result in destination buffer. */
out = ((q63_t) *pSrcA++ * *pSrcB++) >> 32;
out = __SSAT(out, 31);
*pDst++ = out << 1U;
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicMult group
*/

167
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_mult_q7.c Normal file
View file

@ -0,0 +1,167 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_mult_q7.c
* Description: Q7 vector multiplication
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicMult
@{
*/
/**
@brief Q7 vector multiplication
@param[in] pSrcA points to the first input vector
@param[in] pSrcB points to the second input vector
@param[out] pDst points to the output vector
@param[in] blockSize number of samples in each vector
@par Scaling and Overflow Behavior
The function uses saturating arithmetic.
Results outside of the allowable Q7 range [0x80 0x7F] are saturated.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_mult_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
q7_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* loop counters */
q7x16_t vecA, vecB;
/* Compute 16 outputs at a time */
blkCnt = blockSize >> 4;
while (blkCnt > 0U)
{
/*
* C = A * B
* Multiply the inputs and then store the results in the destination buffer.
*/
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vst1q(pDst, vqdmulhq(vecA, vecB));
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
/*
* advance vector source and destination pointers
*/
pSrcA += 16;
pSrcB += 16;
pDst += 16;
}
/*
* tail
*/
blkCnt = blockSize & 0xF;
if (blkCnt > 0U)
{
mve_pred16_t p0 = vctp8q(blkCnt);
vecA = vld1q(pSrcA);
vecB = vld1q(pSrcB);
vstrbq_p(pDst, vqdmulhq(vecA, vecB), p0);
}
}
#else
ARM_DSP_ATTRIBUTE void arm_mult_q7(
const q7_t * pSrcA,
const q7_t * pSrcB,
q7_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined (ARM_MATH_LOOPUNROLL)
#if defined (ARM_MATH_DSP)
q7_t out1, out2, out3, out4; /* Temporary output variables */
#endif
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = A * B */
#if defined (ARM_MATH_DSP)
/* Multiply inputs and store results in temporary variables */
out1 = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
out2 = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
out3 = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
out4 = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
/* Pack and store result in destination buffer (in single write) */
write_q7x4_ia (&pDst, __PACKq7(out1, out2, out3, out4));
#else
*pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
*pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
*pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
*pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
#endif
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = A * B */
/* Multiply input and store result in destination buffer. */
*pDst++ = (q7_t) __SSAT((((q15_t) (*pSrcA++) * (*pSrcB++)) >> 7), 8);
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of BasicMult group
*/

152
sdk/CMSIS/DSP/Source/BasicMathFunctions/arm_negate_f16.c Normal file
View file

@ -0,0 +1,152 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_negate_f16.c
* Description: Negates floating-point vectors
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/basic_math_functions_f16.h"
/**
@ingroup groupMath
*/
/**
@addtogroup BasicNegate
@{
*/
/**
@brief Negates the elements of a floating-point vector.
@param[in] pSrc points to input vector.
@param[out] pDst points to output vector.
@param[in] blockSize number of samples in each vector.
*/
#if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
ARM_DSP_ATTRIBUTE void arm_negate_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
f16x8_t vec1;
f16x8_t res;
/* Compute 4 outputs at a time */
blkCnt = blockSize >> 3U;
while (blkCnt > 0U)
{
/* C = |A| */
/* Calculate absolute values and then store the results in the destination buffer. */
vec1 = vld1q(pSrc);
res = vnegq(vec1);
vst1q(pDst, res);
/* Increment pointers */
pSrc += 8;
pDst += 8;
/* Decrement the loop counter */
blkCnt--;
}
/* Tail */
blkCnt = blockSize & 0x7;
if (blkCnt > 0U)
{
/* C = |A| */
mve_pred16_t p0 = vctp16q(blkCnt);
vec1 = vld1q((float16_t const *) pSrc);
vstrhq_p(pDst, vnegq(vec1), p0);
}
}
#else
#if defined(ARM_FLOAT16_SUPPORTED)
ARM_DSP_ATTRIBUTE void arm_negate_f16(
const float16_t * pSrc,
float16_t * pDst,
uint32_t blockSize)
{
uint32_t blkCnt; /* Loop counter */
#if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* C = -A */
/* Negate and store result in destination buffer. */
*pDst++ = -(_Float16)*pSrc++;
*pDst++ = -(_Float16)*pSrc++;
*pDst++ = -(_Float16)*pSrc++;
*pDst++ = -(_Float16)*pSrc++;
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = blockSize;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C = -A */
/* Negate and store result in destination buffer. */
*pDst++ = -(_Float16)*pSrc++;
/* Decrement loop counter */
blkCnt--;
}
}
#endif
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of BasicNegate group
*/

Some files were not shown because too many files have changed in this diff Show more