From 86608c6770cf08c138a2bdab5855072f64be09ef Mon Sep 17 00:00:00 2001 From: joshua Date: Sat, 30 Dec 2023 23:54:31 -0500 Subject: initial commit --- .../DSP/Source/StatisticsFunctions/CMakeLists.txt | 16 ++ .../StatisticsFunctions/StatisticsFunctions.c | 53 ++++ .../DSP/Source/StatisticsFunctions/arm_max_f32.c | 271 +++++++++++++++++++++ .../DSP/Source/StatisticsFunctions/arm_max_q15.c | 148 +++++++++++ .../DSP/Source/StatisticsFunctions/arm_max_q31.c | 148 +++++++++++ .../DSP/Source/StatisticsFunctions/arm_max_q7.c | 148 +++++++++++ .../DSP/Source/StatisticsFunctions/arm_mean_f32.c | 166 +++++++++++++ .../DSP/Source/StatisticsFunctions/arm_mean_q15.c | 114 +++++++++ .../DSP/Source/StatisticsFunctions/arm_mean_q31.c | 110 +++++++++ .../DSP/Source/StatisticsFunctions/arm_mean_q7.c | 112 +++++++++ .../DSP/Source/StatisticsFunctions/arm_min_f32.c | 268 ++++++++++++++++++++ .../DSP/Source/StatisticsFunctions/arm_min_q15.c | 149 +++++++++++ .../DSP/Source/StatisticsFunctions/arm_min_q31.c | 149 +++++++++++ .../DSP/Source/StatisticsFunctions/arm_min_q7.c | 149 +++++++++++ .../DSP/Source/StatisticsFunctions/arm_power_f32.c | 175 +++++++++++++ .../DSP/Source/StatisticsFunctions/arm_power_q15.c | 132 ++++++++++ .../DSP/Source/StatisticsFunctions/arm_power_q31.c | 121 +++++++++ .../DSP/Source/StatisticsFunctions/arm_power_q7.c | 136 +++++++++++ .../DSP/Source/StatisticsFunctions/arm_rms_f32.c | 176 +++++++++++++ .../DSP/Source/StatisticsFunctions/arm_rms_q15.c | 134 ++++++++++ .../DSP/Source/StatisticsFunctions/arm_rms_q31.c | 124 ++++++++++ .../DSP/Source/StatisticsFunctions/arm_std_f32.c | 188 ++++++++++++++ .../DSP/Source/StatisticsFunctions/arm_std_q15.c | 161 ++++++++++++ .../DSP/Source/StatisticsFunctions/arm_std_q31.c | 147 +++++++++++ .../DSP/Source/StatisticsFunctions/arm_var_f32.c | 234 ++++++++++++++++++ .../DSP/Source/StatisticsFunctions/arm_var_q15.c | 164 +++++++++++++ .../DSP/Source/StatisticsFunctions/arm_var_q31.c | 147 +++++++++++ 27 files changed, 4040 insertions(+) create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/CMakeLists.txt create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/StatisticsFunctions.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_f32.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q15.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q31.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q7.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_f32.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q15.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q31.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q7.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_f32.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q15.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q31.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q7.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_f32.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q15.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q31.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q7.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_f32.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_q15.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_q31.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_f32.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_q15.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_q31.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_f32.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_q15.c create mode 100644 Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_q31.c (limited to 'Drivers/CMSIS/DSP/Source/StatisticsFunctions') diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/CMakeLists.txt b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/CMakeLists.txt new file mode 100644 index 0000000..1a56e61 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/CMakeLists.txt @@ -0,0 +1,16 @@ +cmake_minimum_required (VERSION 3.6) + +project(CMSISDSPStatistics) + + +file(GLOB SRC "./*_*.c") + +add_library(CMSISDSPStatistics STATIC ${SRC}) + +configdsp(CMSISDSPStatistics ..) + +### Includes +target_include_directories(CMSISDSPStatistics PUBLIC "${DSP}/../../Include") + + + diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/StatisticsFunctions.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/StatisticsFunctions.c new file mode 100644 index 0000000..ae74d02 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/StatisticsFunctions.c @@ -0,0 +1,53 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: StatisticsFunctions.c + * Description: Combination of all statistics function source files. + * + * $Date: 18. March 2019 + * $Revision: V1.0.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * Copyright (C) 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. + */ + +#include "arm_max_f32.c" +#include "arm_max_q15.c" +#include "arm_max_q31.c" +#include "arm_max_q7.c" +#include "arm_mean_f32.c" +#include "arm_mean_q15.c" +#include "arm_mean_q31.c" +#include "arm_mean_q7.c" +#include "arm_min_f32.c" +#include "arm_min_q15.c" +#include "arm_min_q31.c" +#include "arm_min_q7.c" +#include "arm_power_f32.c" +#include "arm_power_q15.c" +#include "arm_power_q31.c" +#include "arm_power_q7.c" +#include "arm_rms_f32.c" +#include "arm_rms_q15.c" +#include "arm_rms_q31.c" +#include "arm_std_f32.c" +#include "arm_std_q15.c" +#include "arm_std_q31.c" +#include "arm_var_f32.c" +#include "arm_var_q15.c" +#include "arm_var_q31.c" diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_f32.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_f32.c new file mode 100644 index 0000000..f6ffafe --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_f32.c @@ -0,0 +1,271 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_max_f32.c + * Description: Maximum value of a floating-point vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" +#if defined(ARM_MATH_NEON) +#include +#endif + +/** + @ingroup groupStats + */ + +/** + @defgroup Max Maximum + + Computes the maximum value of an array of data. + The function returns both the maximum value and its position within the array. + There are separate functions for floating-point, Q31, Q15, and Q7 data types. + */ + +/** + @addtogroup Max + @{ + */ + +/** + @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 + @param[out] pIndex index of maximum value returned here + @return none + */ +#if defined(ARM_MATH_NEON) +void arm_max_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult, + uint32_t * pIndex) +{ + float32_t maxVal1, maxVal2, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex, count; /* loop counter */ + + float32x4_t outV, srcV; + float32x2_t outV2; + + uint32x4_t idxV; + uint32x4_t maxIdx={ULONG_MAX,ULONG_MAX,ULONG_MAX,ULONG_MAX}; + uint32x4_t index={4,5,6,7}; + uint32x4_t delta={4,4,4,4}; + uint32x4_t countV={0,1,2,3}; + uint32x2_t countV2; + + /* Initialise the count value. */ + count = 0U; + + /* Initialise the index value to zero. */ + outIndex = 0U; + + /* Load first input value that act as reference value for comparison */ + if (blockSize <= 3) + { + out = *pSrc++; + + blkCnt = blockSize - 1; + + while (blkCnt > 0U) + { + /* Initialize maxVal to the next consecutive values one by one */ + maxVal1 = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal1) + { + /* Update the maximum value and it's index */ + out = maxVal1; + outIndex = blockSize - blkCnt; + } + + /* Decrement the loop counter */ + blkCnt--; + } + } + else + { + outV = vld1q_f32(pSrc); + pSrc += 4; + + /* Compute 4 outputs at a time */ + blkCnt = (blockSize - 4 ) >> 2U; + + while (blkCnt > 0U) + { + srcV = vld1q_f32(pSrc); + pSrc += 4; + + idxV = vcgtq_f32(srcV, outV); + outV = vbslq_f32(idxV, srcV, outV ); + countV = vbslq_u32(idxV, index,countV ); + + index = vaddq_u32(index,delta); + + /* Decrement the loop counter */ + blkCnt--; + } + + outV2 = vpmax_f32(vget_low_f32(outV),vget_high_f32(outV)); + outV2 = vpmax_f32(outV2,outV2); + out = outV2[0]; + + idxV = vceqq_f32(outV, vdupq_n_f32(out)); + countV = vbslq_u32(idxV, countV,maxIdx); + + countV2 = vpmin_u32(vget_low_u32(countV),vget_high_u32(countV)); + countV2 = vpmin_u32(countV2,countV2); + outIndex = countV2[0]; + + /* if (blockSize - 1U) is not multiple of 4 */ + blkCnt = (blockSize - 4 ) % 4U; + + while (blkCnt > 0U) + { + /* Initialize maxVal to the next consecutive values one by one */ + maxVal1 = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal1) + { + /* Update the maximum value and it's index */ + out = maxVal1; + outIndex = blockSize - blkCnt ; + } + + /* Decrement the loop counter */ + blkCnt--; + } + + + } + + /* Store the maximum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} +#else +void arm_max_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult, + uint32_t * pIndex) +{ + float32_t maxVal, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex; /* Loop counter */ + +#if defined (ARM_MATH_LOOPUNROLL) + uint32_t index; /* index of maximum value */ +#endif + + /* Initialise index value to zero. */ + outIndex = 0U; + + /* Load first input value that act as reference value for comparision */ + out = *pSrc++; + +#if defined (ARM_MATH_LOOPUNROLL) + /* Initialise index of maximum value. */ + index = 0U; + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = (blockSize - 1U) >> 2U; + + while (blkCnt > 0U) + { + /* Initialize maxVal to next consecutive values one by one */ + maxVal = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal) + { + /* Update the maximum value and it's index */ + out = maxVal; + outIndex = index + 1U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 2U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 3U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 4U; + } + + index += 4U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = (blockSize - 1U) % 4U; + +#else + + /* Initialize blkCnt with number of samples */ + blkCnt = (blockSize - 1U); + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (blkCnt > 0U) + { + /* Initialize maxVal to the next consecutive values one by one */ + maxVal = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal) + { + /* Update the maximum value and it's index */ + out = maxVal; + outIndex = blockSize - blkCnt; + } + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store the maximum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} +#endif /* #if defined(ARM_MATH_NEON) */ +/** + @} end of Max group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q15.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q15.c new file mode 100644 index 0000000..dbead54 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q15.c @@ -0,0 +1,148 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_max_q15.c + * Description: Maximum value of a Q15 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup Max + @{ + */ + +/** + @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 + @param[out] pIndex index of maximum value returned here + @return none + */ + +void arm_max_q15( + const q15_t * pSrc, + uint32_t blockSize, + q15_t * pResult, + uint32_t * pIndex) +{ + q15_t maxVal, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex; /* Loop counter */ + +#if defined (ARM_MATH_LOOPUNROLL) + uint32_t index; /* index of maximum value */ +#endif + + /* Initialise index value to zero. */ + outIndex = 0U; + /* Load first input value that act as reference value for comparision */ + out = *pSrc++; + +#if defined (ARM_MATH_LOOPUNROLL) + /* Initialise index of maximum value. */ + index = 0U; + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = (blockSize - 1U) >> 2U; + + while (blkCnt > 0U) + { + /* Initialize maxVal to next consecutive values one by one */ + maxVal = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal) + { + /* Update the maximum value and it's index */ + out = maxVal; + outIndex = index + 1U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 2U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 3U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 4U; + } + + index += 4U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = (blockSize - 1U) % 4U; + +#else + + /* Initialize blkCnt with number of samples */ + blkCnt = (blockSize - 1U); + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (blkCnt > 0U) + { + /* Initialize maxVal to the next consecutive values one by one */ + maxVal = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal) + { + /* Update the maximum value and it's index */ + out = maxVal; + outIndex = blockSize - blkCnt; + } + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store the maximum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} + +/** + @} end of Max group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q31.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q31.c new file mode 100644 index 0000000..5c6badf --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q31.c @@ -0,0 +1,148 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_max_q31.c + * Description: Maximum value of a Q31 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup Max + @{ + */ + +/** + @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 + @param[out] pIndex index of maximum value returned here + @return none + */ + +void arm_max_q31( + const q31_t * pSrc, + uint32_t blockSize, + q31_t * pResult, + uint32_t * pIndex) +{ + q31_t maxVal, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex; /* Loop counter */ + +#if defined (ARM_MATH_LOOPUNROLL) + uint32_t index; /* index of maximum value */ +#endif + + /* Initialise index value to zero. */ + outIndex = 0U; + /* Load first input value that act as reference value for comparision */ + out = *pSrc++; + +#if defined (ARM_MATH_LOOPUNROLL) + /* Initialise index of maximum value. */ + index = 0U; + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = (blockSize - 1U) >> 2U; + + while (blkCnt > 0U) + { + /* Initialize maxVal to next consecutive values one by one */ + maxVal = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal) + { + /* Update the maximum value and it's index */ + out = maxVal; + outIndex = index + 1U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 2U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 3U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 4U; + } + + index += 4U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = (blockSize - 1U) % 4U; + +#else + + /* Initialize blkCnt with number of samples */ + blkCnt = (blockSize - 1U); + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (blkCnt > 0U) + { + /* Initialize maxVal to the next consecutive values one by one */ + maxVal = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal) + { + /* Update the maximum value and it's index */ + out = maxVal; + outIndex = blockSize - blkCnt; + } + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store the maximum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} + +/** + @} end of Max group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q7.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q7.c new file mode 100644 index 0000000..d4977e5 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_max_q7.c @@ -0,0 +1,148 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_max_q7.c + * Description: Maximum value of a Q7 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup Max + @{ + */ + +/** + @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 + @param[out] pIndex index of maximum value returned here + @return none + */ + +void arm_max_q7( + const q7_t * pSrc, + uint32_t blockSize, + q7_t * pResult, + uint32_t * pIndex) +{ + q7_t maxVal, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex; /* Loop counter */ + +#if defined (ARM_MATH_LOOPUNROLL) + uint32_t index; /* index of maximum value */ +#endif + + /* Initialise index value to zero. */ + outIndex = 0U; + /* Load first input value that act as reference value for comparision */ + out = *pSrc++; + +#if defined (ARM_MATH_LOOPUNROLL) + /* Initialise index of maximum value. */ + index = 0U; + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = (blockSize - 1U) >> 2U; + + while (blkCnt > 0U) + { + /* Initialize maxVal to next consecutive values one by one */ + maxVal = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal) + { + /* Update the maximum value and it's index */ + out = maxVal; + outIndex = index + 1U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 2U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 3U; + } + + maxVal = *pSrc++; + if (out < maxVal) + { + out = maxVal; + outIndex = index + 4U; + } + + index += 4U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = (blockSize - 1U) % 4U; + +#else + + /* Initialize blkCnt with number of samples */ + blkCnt = (blockSize - 1U); + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (blkCnt > 0U) + { + /* Initialize maxVal to the next consecutive values one by one */ + maxVal = *pSrc++; + + /* compare for the maximum value */ + if (out < maxVal) + { + /* Update the maximum value and it's index */ + out = maxVal; + outIndex = blockSize - blkCnt; + } + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store the maximum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} + +/** + @} end of Max group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_f32.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_f32.c new file mode 100644 index 0000000..3373f82 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_f32.c @@ -0,0 +1,166 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_mean_f32.c + * Description: Mean value of a floating-point vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @defgroup mean Mean + + Calculates the mean of the input vector. Mean is defined as the average of the elements in the vector. + The underlying algorithm is used: + + 
+      Result = (pSrc[0] + pSrc[1] + pSrc[2] + ... + pSrc[blockSize-1]) / blockSize;
+
+ + There are separate functions for floating-point, Q31, Q15, and Q7 data types. + */ + +/** + @addtogroup mean + @{ + */ + +/** + @brief Mean 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 mean value returned here. + @return none + */ +#if defined(ARM_MATH_NEON_EXPERIMENTAL) +void arm_mean_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + float32_t sum = 0.0f; /* Temporary result storage */ + float32x4_t sumV = vdupq_n_f32(0.0f); /* Temporary result storage */ + float32x2_t sumV2; + + uint32_t blkCnt; /* Loop counter */ + + float32_t in1, in2, in3, in4; + float32x4_t inV; + + blkCnt = blockSize >> 2U; + + /* Compute 4 outputs at a time. + ** a second loop below computes the remaining 1 to 3 samples. */ + while (blkCnt > 0U) + { + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ + inV = vld1q_f32(pSrc); + sumV = vaddq_f32(sumV, inV); + + pSrc += 4; + /* Decrement the loop counter */ + blkCnt--; + } + + sumV2 = vpadd_f32(vget_low_f32(sumV),vget_high_f32(sumV)); + sum = sumV2[0] + sumV2[1]; + + /* If the blockSize is not a multiple of 4, compute any remaining output samples here. + ** No loop unrolling is used. */ + blkCnt = blockSize & 3; + + while (blkCnt > 0U) + { + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ + sum += *pSrc++; + + /* Decrement the loop counter */ + blkCnt--; + } + + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) / blockSize */ + /* Store the result to the destination */ + *pResult = sum / (float32_t) blockSize; +} +#else +void arm_mean_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + float32_t sum = 0.0f; /* Temporary result storage */ + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ + sum += *pSrc++; + + sum += *pSrc++; + + sum += *pSrc++; + + sum += *pSrc++; + + /* Decrement the 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] + A[1] + A[2] + ... + A[blockSize-1]) */ + sum += *pSrc++; + + /* Decrement loop counter */ + blkCnt--; + } + + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) / blockSize */ + /* Store result to destination */ + *pResult = (sum / blockSize); +} +#endif /* #if defined(ARM_MATH_NEON) */ + +/** + @} end of mean group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q15.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q15.c new file mode 100644 index 0000000..09e6da9 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q15.c @@ -0,0 +1,114 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_mean_q15.c + * Description: Mean value of a Q15 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup mean + @{ + */ + +/** + @brief Mean 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 mean value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 32-bit internal accumulator. + The input is represented in 1.15 format and is accumulated in a 32-bit + accumulator in 17.15 format. + There is no risk of internal overflow with this approach, and the + full precision of intermediate result is preserved. + Finally, the accumulator is truncated to yield a result of 1.15 format. + */ + +void arm_mean_q15( + const q15_t * pSrc, + uint32_t blockSize, + q15_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q31_t sum = 0; /* Temporary result storage */ + +#if defined (ARM_MATH_LOOPUNROLL) + q31_t in; +#endif + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ + in = read_q15x2_ia ((q15_t **) &pSrc); + sum += ((in << 16U) >> 16U); + sum += (in >> 16U); + + in = read_q15x2_ia ((q15_t **) &pSrc); + sum += ((in << 16U) >> 16U); + sum += (in >> 16U); + + /* Decrement the 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] + A[1] + A[2] + ... + A[blockSize-1]) */ + sum += *pSrc++; + + /* Decrement loop counter */ + blkCnt--; + } + + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) / blockSize */ + /* Store result to destination */ + *pResult = (q15_t) (sum / (int32_t) blockSize); +} + +/** + @} end of mean group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q31.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q31.c new file mode 100644 index 0000000..7667ad8 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q31.c @@ -0,0 +1,110 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_mean_q31.c + * Description: Mean value of a Q31 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup mean + @{ + */ + +/** + @brief Mean 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 mean value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 64-bit internal accumulator. + The input is represented in 1.31 format and is accumulated in a 64-bit + accumulator in 33.31 format. + There is no risk of internal overflow with this approach, and the + full precision of intermediate result is preserved. + Finally, the accumulator is truncated to yield a result of 1.31 format. + */ + +void arm_mean_q31( + const q31_t * pSrc, + uint32_t blockSize, + q31_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q63_t sum = 0; /* Temporary result storage */ + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ + sum += *pSrc++; + + sum += *pSrc++; + + sum += *pSrc++; + + sum += *pSrc++; + + /* Decrement the 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] + A[1] + A[2] + ... + A[blockSize-1]) */ + sum += *pSrc++; + + /* Decrement loop counter */ + blkCnt--; + } + + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) / blockSize */ + /* Store result to destination */ + *pResult = (q31_t) (sum / blockSize); +} + +/** + @} end of mean group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q7.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q7.c new file mode 100644 index 0000000..3f90ea8 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_mean_q7.c @@ -0,0 +1,112 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_mean_q7.c + * Description: Mean value of a Q7 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup mean + @{ + */ + +/** + @brief Mean 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 mean value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 32-bit internal accumulator. + The input is represented in 1.7 format and is accumulated in a 32-bit + accumulator in 25.7 format. + There is no risk of internal overflow with this approach, and the + full precision of intermediate result is preserved. + Finally, the accumulator is truncated to yield a result of 1.7 format. + */ + +void arm_mean_q7( + const q7_t * pSrc, + uint32_t blockSize, + q7_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q31_t sum = 0; /* Temporary result storage */ + +#if defined (ARM_MATH_LOOPUNROLL) + q31_t in; +#endif + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ + in = read_q7x4_ia ((q7_t **) &pSrc); + sum += ((in << 24U) >> 24U); + sum += ((in << 16U) >> 24U); + sum += ((in << 8U) >> 24U); + sum += (in >> 24U); + + /* Decrement the 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] + A[1] + A[2] + ... + A[blockSize-1]) */ + sum += *pSrc++; + + /* Decrement loop counter */ + blkCnt--; + } + + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) / blockSize */ + /* Store result to destination */ + *pResult = (q7_t) (sum / (int32_t) blockSize); +} + +/** + @} end of mean group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_f32.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_f32.c new file mode 100644 index 0000000..f6504aa --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_f32.c @@ -0,0 +1,268 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_min_f32.c + * Description: Minimum value of a floating-point vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" +#include + +/** + @ingroup groupStats + */ + +/** + @defgroup Min Minimum + + Computes the minimum value of an array of data. + The function returns both the minimum value and its position within the array. + There are separate functions for floating-point, Q31, Q15, and Q7 data types. + */ + +/** + @addtogroup Min + @{ + */ + +/** + @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 + @param[out] pIndex index of minimum value returned here + @return none + */ +#if defined(ARM_MATH_NEON) +void arm_min_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult, + uint32_t * pIndex) +{ + float32_t maxVal1, maxVal2, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex, count; /* loop counter */ + + float32x4_t outV, srcV; + float32x2_t outV2; + + uint32x4_t idxV; + uint32x4_t maxIdx={ULONG_MAX,ULONG_MAX,ULONG_MAX,ULONG_MAX}; + uint32x4_t index={4,5,6,7}; + uint32x4_t delta={4,4,4,4}; + uint32x4_t countV={0,1,2,3}; + uint32x2_t countV2; + + /* Initialise the count value. */ + count = 0U; + + /* Initialise the index value to zero. */ + outIndex = 0U; + + /* Load first input value that act as reference value for comparison */ + if (blockSize <= 3) + { + out = *pSrc++; + + blkCnt = blockSize - 1; + + while (blkCnt > 0U) + { + /* Initialize maxVal to the next consecutive values one by one */ + maxVal1 = *pSrc++; + + /* compare for the maximum value */ + if (out > maxVal1) + { + /* Update the maximum value and it's index */ + out = maxVal1; + outIndex = blockSize - blkCnt; + } + + /* Decrement the loop counter */ + blkCnt--; + } + } + else + { + outV = vld1q_f32(pSrc); + pSrc += 4; + + /* Compute 4 outputs at a time */ + blkCnt = (blockSize - 4 ) >> 2U; + + while (blkCnt > 0U) + { + srcV = vld1q_f32(pSrc); + pSrc += 4; + + idxV = vcltq_f32(srcV, outV); + outV = vbslq_f32(idxV, srcV, outV ); + countV = vbslq_u32(idxV, index,countV ); + + index = vaddq_u32(index,delta); + + /* Decrement the loop counter */ + blkCnt--; + } + + outV2 = vpmin_f32(vget_low_f32(outV),vget_high_f32(outV)); + outV2 = vpmin_f32(outV2,outV2); + out = outV2[0]; + + idxV = vceqq_f32(outV, vdupq_n_f32(out)); + countV = vbslq_u32(idxV, countV,maxIdx); + + countV2 = vpmin_u32(vget_low_u32(countV),vget_high_u32(countV)); + countV2 = vpmin_u32(countV2,countV2); + outIndex = countV2[0]; + + /* if (blockSize - 1U) is not multiple of 4 */ + blkCnt = (blockSize - 4 ) % 4U; + + while (blkCnt > 0U) + { + /* Initialize maxVal to the next consecutive values one by one */ + maxVal1 = *pSrc++; + + /* compare for the maximum value */ + if (out > maxVal1) + { + /* Update the maximum value and it's index */ + out = maxVal1; + outIndex = blockSize - blkCnt ; + } + + /* Decrement the loop counter */ + blkCnt--; + } + } + + /* Store the maximum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} +#else +void arm_min_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult, + uint32_t * pIndex) +{ + float32_t minVal, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex; /* Loop counter */ + +#if defined (ARM_MATH_LOOPUNROLL) + uint32_t index; /* index of maximum value */ +#endif + + /* Initialise index value to zero. */ + outIndex = 0U; + + /* Load first input value that act as reference value for comparision */ + out = *pSrc++; + +#if defined (ARM_MATH_LOOPUNROLL) + /* Initialise index of maximum value. */ + index = 0U; + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = (blockSize - 1U) >> 2U; + + while (blkCnt > 0U) + { + /* Initialize minVal to next consecutive values one by one */ + minVal = *pSrc++; + + /* compare for the minimum value */ + if (out > minVal) + { + /* Update the minimum value and it's index */ + out = minVal; + outIndex = index + 1U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 2U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 3U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 4U; + } + + index += 4U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = (blockSize - 1U) % 4U; + +#else + + /* Initialize blkCnt with number of samples */ + blkCnt = (blockSize - 1U); + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (blkCnt > 0U) + { + /* Initialize minVal to the next consecutive values one by one */ + minVal = *pSrc++; + + /* compare for the minimum value */ + if (out > minVal) + { + /* Update the minimum value and it's index */ + out = minVal; + outIndex = blockSize - blkCnt; + } + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store the minimum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} +#endif /* #if defined(ARM_MATH_NEON) */ + +/** + @} end of Min group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q15.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q15.c new file mode 100644 index 0000000..67fd49e --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q15.c @@ -0,0 +1,149 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_min_q15.c + * Description: Minimum value of a Q15 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + + +/** + @addtogroup Min + @{ + */ + +/** + @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 + @param[out] pIndex index of minimum value returned here + @return none + */ + +void arm_min_q15( + const q15_t * pSrc, + uint32_t blockSize, + q15_t * pResult, + uint32_t * pIndex) +{ + q15_t minVal, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex; /* Loop counter */ + +#if defined (ARM_MATH_LOOPUNROLL) + uint32_t index; /* index of maximum value */ +#endif + + /* Initialise index value to zero. */ + outIndex = 0U; + /* Load first input value that act as reference value for comparision */ + out = *pSrc++; + +#if defined (ARM_MATH_LOOPUNROLL) + /* Initialise index of maximum value. */ + index = 0U; + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = (blockSize - 1U) >> 2U; + + while (blkCnt > 0U) + { + /* Initialize minVal to next consecutive values one by one */ + minVal = *pSrc++; + + /* compare for the minimum value */ + if (out > minVal) + { + /* Update the minimum value and it's index */ + out = minVal; + outIndex = index + 1U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 2U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 3U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 4U; + } + + index += 4U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = (blockSize - 1U) % 4U; + +#else + + /* Initialize blkCnt with number of samples */ + blkCnt = (blockSize - 1U); + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (blkCnt > 0U) + { + /* Initialize minVal to the next consecutive values one by one */ + minVal = *pSrc++; + + /* compare for the minimum value */ + if (out > minVal) + { + /* Update the minimum value and it's index */ + out = minVal; + outIndex = blockSize - blkCnt; + } + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store the minimum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} + +/** + @} end of Min group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q31.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q31.c new file mode 100644 index 0000000..c362173 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q31.c @@ -0,0 +1,149 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_min_q31.c + * Description: Minimum value of a Q31 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + + +/** + @addtogroup Min + @{ + */ + +/** + @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 + @param[out] pIndex index of minimum value returned here + @return none + */ + +void arm_min_q31( + const q31_t * pSrc, + uint32_t blockSize, + q31_t * pResult, + uint32_t * pIndex) +{ + q31_t minVal, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex; /* Loop counter */ + +#if defined (ARM_MATH_LOOPUNROLL) + uint32_t index; /* index of maximum value */ +#endif + + /* Initialise index value to zero. */ + outIndex = 0U; + /* Load first input value that act as reference value for comparision */ + out = *pSrc++; + +#if defined (ARM_MATH_LOOPUNROLL) + /* Initialise index of maximum value. */ + index = 0U; + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = (blockSize - 1U) >> 2U; + + while (blkCnt > 0U) + { + /* Initialize minVal to next consecutive values one by one */ + minVal = *pSrc++; + + /* compare for the minimum value */ + if (out > minVal) + { + /* Update the minimum value and it's index */ + out = minVal; + outIndex = index + 1U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 2U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 3U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 4U; + } + + index += 4U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = (blockSize - 1U) % 4U; + +#else + + /* Initialize blkCnt with number of samples */ + blkCnt = (blockSize - 1U); + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (blkCnt > 0U) + { + /* Initialize minVal to the next consecutive values one by one */ + minVal = *pSrc++; + + /* compare for the minimum value */ + if (out > minVal) + { + /* Update the minimum value and it's index */ + out = minVal; + outIndex = blockSize - blkCnt; + } + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store the minimum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} + +/** + @} end of Min group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q7.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q7.c new file mode 100644 index 0000000..4329630 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_min_q7.c @@ -0,0 +1,149 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_min_q7.c + * Description: Minimum value of a Q7 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + + +/** + @addtogroup Min + @{ + */ + +/** + @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 + @param[out] pIndex index of minimum value returned here + @return none + */ + +void arm_min_q7( + const q7_t * pSrc, + uint32_t blockSize, + q7_t * pResult, + uint32_t * pIndex) +{ + q7_t minVal, out; /* Temporary variables to store the output value. */ + uint32_t blkCnt, outIndex; /* Loop counter */ + +#if defined (ARM_MATH_LOOPUNROLL) + uint32_t index; /* index of maximum value */ +#endif + + /* Initialise index value to zero. */ + outIndex = 0U; + /* Load first input value that act as reference value for comparision */ + out = *pSrc++; + +#if defined (ARM_MATH_LOOPUNROLL) + /* Initialise index of maximum value. */ + index = 0U; + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = (blockSize - 1U) >> 2U; + + while (blkCnt > 0U) + { + /* Initialize minVal to next consecutive values one by one */ + minVal = *pSrc++; + + /* compare for the minimum value */ + if (out > minVal) + { + /* Update the minimum value and it's index */ + out = minVal; + outIndex = index + 1U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 2U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 3U; + } + + minVal = *pSrc++; + if (out > minVal) + { + out = minVal; + outIndex = index + 4U; + } + + index += 4U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = (blockSize - 1U) % 4U; + +#else + + /* Initialize blkCnt with number of samples */ + blkCnt = (blockSize - 1U); + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (blkCnt > 0U) + { + /* Initialize minVal to the next consecutive values one by one */ + minVal = *pSrc++; + + /* compare for the minimum value */ + if (out > minVal) + { + /* Update the minimum value and it's index */ + out = minVal; + outIndex = blockSize - blkCnt; + } + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store the minimum value and it's index into destination pointers */ + *pResult = out; + *pIndex = outIndex; +} + +/** + @} end of Min group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_f32.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_f32.c new file mode 100644 index 0000000..abf444b --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_f32.c @@ -0,0 +1,175 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_power_f32.c + * Description: Sum of the squares of the elements of a floating-point vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @defgroup power Power + + Calculates the sum of the squares of the elements in the input vector. + The underlying algorithm is used: + + 
+      Result = pSrc[0] * pSrc[0] + pSrc[1] * pSrc[1] + pSrc[2] * pSrc[2] + ... + pSrc[blockSize-1] * pSrc[blockSize-1];
+
+ + There are separate functions for floating point, Q31, Q15, and Q7 data types. + */ + +/** + @addtogroup power + @{ + */ + +/** + @brief Sum of the squares of the elements 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 sum of the squares value returned here + @return none + */ +#if defined(ARM_MATH_NEON) +void arm_power_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + float32_t sum = 0.0f; /* accumulator */ + float32_t in; /* Temporary variable to store input value */ + uint32_t blkCnt; /* loop counter */ + + float32x4_t sumV = vdupq_n_f32(0.0f); /* Temporary result storage */ + float32x2_t sumV2; + float32x4_t inV; + + blkCnt = blockSize >> 2U; + + /* Compute 4 outputs at a time. + ** a second loop below computes the remaining 1 to 3 samples. */ + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + ... + A[blockSize-1] * A[blockSize-1] */ + /* Compute Power and then store the result in a temporary variable, sum. */ + inV = vld1q_f32(pSrc); + sumV = vmlaq_f32(sumV, inV, inV); + pSrc += 4; + + /* Decrement the loop counter */ + blkCnt--; + } + sumV2 = vpadd_f32(vget_low_f32(sumV),vget_high_f32(sumV)); + sum = sumV2[0] + sumV2[1]; + + /* If the blockSize is not a multiple of 4, compute any remaining output samples here. + ** No loop unrolling is used. */ + blkCnt = blockSize % 0x4U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + ... + A[blockSize-1] * A[blockSize-1] */ + /* compute power and then store the result in a temporary variable, sum. */ + in = *pSrc++; + sum += in * in; + + /* Decrement the loop counter */ + blkCnt--; + } + + /* Store the result to the destination */ + *pResult = sum; +} +#else +void arm_power_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + float32_t sum = 0.0f; /* Temporary result storage */ + float32_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute Power and store result in a temporary variable, sum. */ + in = *pSrc++; + sum += in * in; + + in = *pSrc++; + sum += in * in; + + in = *pSrc++; + sum += in * in; + + in = *pSrc++; + sum += in * in; + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute Power and store result in a temporary variable, sum. */ + in = *pSrc++; + sum += in * in; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store result to destination */ + *pResult = sum; +} +#endif /* #if defined(ARM_MATH_NEON) */ + +/** + @} end of power group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q15.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q15.c new file mode 100644 index 0000000..62e567f --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q15.c @@ -0,0 +1,132 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_power_q15.c + * Description: Sum of the squares of the elements of a Q15 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup power + @{ + */ + +/** + @brief Sum of the squares of the elements of a Q15 vector. + @param[in] pSrc points to the input vector + @param[in] blockSize number of samples in input vector + @param[out] pResult sum of the squares value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 64-bit internal accumulator. + The input is represented in 1.15 format. + Intermediate multiplication yields a 2.30 format, and this + result is added without saturation to a 64-bit accumulator in 34.30 format. + With 33 guard bits in the accumulator, there is no risk of overflow, and the + full precision of the intermediate multiplication is preserved. + Finally, the return result is in 34.30 format. + */ + +void arm_power_q15( + const q15_t * pSrc, + uint32_t blockSize, + q63_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q63_t sum = 0; /* Temporary result storage */ + q15_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) && defined (ARM_MATH_DSP) + q31_t in32; /* Temporary variable to store packed input value */ +#endif + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute Power and store result in a temporary variable, sum. */ +#if defined (ARM_MATH_DSP) + in32 = read_q15x2_ia ((q15_t **) &pSrc); + sum = __SMLALD(in32, in32, sum); + + in32 = read_q15x2_ia ((q15_t **) &pSrc); + sum = __SMLALD(in32, in32, sum); +#else + in = *pSrc++; + sum += ((q31_t) in * in); + + in = *pSrc++; + sum += ((q31_t) in * in); + + in = *pSrc++; + sum += ((q31_t) in * in); + + in = *pSrc++; + sum += ((q31_t) in * in); +#endif /* #if defined (ARM_MATH_DSP) */ + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute Power and store result in a temporary variable, sum. */ + in = *pSrc++; + sum += ((q31_t) in * in); + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store result in 34.30 format */ + *pResult = sum; +} + +/** + @} end of power group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q31.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q31.c new file mode 100644 index 0000000..aa51d0f --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q31.c @@ -0,0 +1,121 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_power_q31.c + * Description: Sum of the squares of the elements of a Q31 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup power + @{ + */ + +/** + @brief Sum of the squares of the elements of a Q31 vector. + @param[in] pSrc points to the input vector + @param[in] blockSize number of samples in input vector + @param[out] pResult sum of the squares value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 64-bit internal accumulator. + The input is represented in 1.31 format. + Intermediate multiplication yields a 2.62 format, and this + result is 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. + With 15 guard bits in the accumulator, there is no risk of overflow, and the + full precision of the intermediate multiplication is preserved. + Finally, the return result is in 16.48 format. + */ + +void arm_power_q31( + const q31_t * pSrc, + uint32_t blockSize, + q63_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q63_t sum = 0; /* Temporary result storage */ + q31_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute Power then shift intermediate results by 14 bits to maintain 16.48 format and store result in a temporary variable sum, providing 15 guard bits. */ + in = *pSrc++; + sum += ((q63_t) in * in) >> 14U; + + in = *pSrc++; + sum += ((q63_t) in * in) >> 14U; + + in = *pSrc++; + sum += ((q63_t) in * in) >> 14U; + + in = *pSrc++; + sum += ((q63_t) in * in) >> 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute Power and store result in a temporary variable, sum. */ + in = *pSrc++; + sum += ((q63_t) in * in) >> 14U; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store results in 16.48 format */ + *pResult = sum; +} + +/** + @} end of power group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q7.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q7.c new file mode 100644 index 0000000..eefa8ba --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_power_q7.c @@ -0,0 +1,136 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_power_q7.c + * Description: Sum of the squares of the elements of a Q7 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup power + @{ + */ + +/** + @brief Sum of the squares of the elements of a Q7 vector. + @param[in] pSrc points to the input vector + @param[in] blockSize number of samples in input vector + @param[out] pResult sum of the squares value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 32-bit internal accumulator. + The input is represented in 1.7 format. + Intermediate multiplication yields a 2.14 format, and this + result is added without saturation to an accumulator in 18.14 format. + With 17 guard bits in the accumulator, there is no risk of overflow, and the + full precision of the intermediate multiplication is preserved. + Finally, the return result is in 18.14 format. + */ + +void arm_power_q7( + const q7_t * pSrc, + uint32_t blockSize, + q31_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q31_t sum = 0; /* Temporary result storage */ + q7_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) && defined (ARM_MATH_DSP) + q31_t in32; /* Temporary variable to store packed input value */ + q31_t in1, in2; /* Temporary variables to store input value */ +#endif + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute Power and store result in a temporary variable, sum. */ +#if defined (ARM_MATH_DSP) + in32 = read_q7x4_ia ((q7_t **) &pSrc); + + in1 = __SXTB16(__ROR(in32, 8)); + in2 = __SXTB16(in32); + + /* calculate power and accumulate to accumulator */ + sum = __SMLAD(in1, in1, sum); + sum = __SMLAD(in2, in2, sum); +#else + in = *pSrc++; + sum += ((q15_t) in * in); + + in = *pSrc++; + sum += ((q15_t) in * in); + + in = *pSrc++; + sum += ((q15_t) in * in); + + in = *pSrc++; + sum += ((q15_t) in * in); +#endif /* #if defined (ARM_MATH_DSP) */ + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute Power and store result in a temporary variable, sum. */ + in = *pSrc++; + sum += ((q15_t) in * in); + + /* Decrement loop counter */ + blkCnt--; + } + + /* Store result in 18.14 format */ + *pResult = sum; +} + +/** + @} end of power group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_f32.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_f32.c new file mode 100644 index 0000000..4f316fb --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_f32.c @@ -0,0 +1,176 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_rms_f32.c + * Description: Root mean square value of the elements of a floating-point vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @defgroup RMS Root mean square (RMS) + + Calculates the Root Mean Square of the elements in the input vector. + The underlying algorithm is used: + + 
+      Result = sqrt(((pSrc[0] * pSrc[0] + pSrc[1] * pSrc[1] + ... + pSrc[blockSize-1] * pSrc[blockSize-1]) / blockSize));
+
+ + There are separate functions for floating point, Q31, and Q15 data types. + */ + +/** + @addtogroup RMS + @{ + */ + +/** + @brief Root Mean Square of the elements 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 root mean square value returned here + @return none + */ +#if defined(ARM_MATH_NEON) +void arm_rms_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + float32_t sum = 0.0f; /* accumulator */ + float32_t in; /* Temporary variable to store input value */ + uint32_t blkCnt; /* loop counter */ + + float32x4_t sumV = vdupq_n_f32(0.0f); /* Temporary result storage */ + float32x2_t sumV2; + float32x4_t inV; + + blkCnt = blockSize >> 2U; + + /* Compute 4 outputs at a time. + ** a second loop below computes the remaining 1 to 3 samples. */ + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + ... + A[blockSize-1] * A[blockSize-1] */ + /* Compute Power and then store the result in a temporary variable, sum. */ + inV = vld1q_f32(pSrc); + sumV = vmlaq_f32(sumV, inV, inV); + pSrc += 4; + + /* Decrement the loop counter */ + blkCnt--; + } + + sumV2 = vpadd_f32(vget_low_f32(sumV),vget_high_f32(sumV)); + sum = sumV2[0] + sumV2[1]; + + /* If the blockSize is not a multiple of 4, compute any remaining output samples here. + ** No loop unrolling is used. */ + blkCnt = blockSize % 0x4U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + ... + A[blockSize-1] * A[blockSize-1] */ + /* compute power and then store the result in a temporary variable, sum. */ + in = *pSrc++; + sum += in * in; + + /* Decrement the loop counter */ + blkCnt--; + } + + /* Compute Rms and store the result in the destination */ + arm_sqrt_f32(sum / (float32_t) blockSize, pResult); +} +#else +void arm_rms_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + float32_t sum = 0.0f; /* Temporary result storage */ + float32_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable, sum. */ + sum += in * in; + + in = *pSrc++; + sum += in * in; + + in = *pSrc++; + sum += in * in; + + in = *pSrc++; + sum += in * in; + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable. */ + sum += ( in * in); + + /* Decrement loop counter */ + blkCnt--; + } + + /* Compute Rms and store result in destination */ + arm_sqrt_f32(sum / (float32_t) blockSize, pResult); +} +#endif /* #if defined(ARM_MATH_NEON) */ + +/** + @} end of RMS group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_q15.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_q15.c new file mode 100644 index 0000000..c8a0a13 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_q15.c @@ -0,0 +1,134 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_rms_q15.c + * Description: Root Mean Square of the elements of a Q15 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup RMS + @{ + */ + +/** + @brief Root Mean Square of the elements of a Q15 vector. + @param[in] pSrc points to the input vector + @param[in] blockSize number of samples in input vector + @param[out] pResult root mean square value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 64-bit internal accumulator. + The input is represented in 1.15 format. + Intermediate multiplication yields a 2.30 format, and this + result is added without saturation to a 64-bit accumulator in 34.30 format. + With 33 guard bits in the accumulator, there is no risk of overflow, and the + full precision of the intermediate multiplication is preserved. + Finally, the 34.30 result is truncated to 34.15 format by discarding the lower + 15 bits, and then saturated to yield a result in 1.15 format. + */ + +void arm_rms_q15( + const q15_t * pSrc, + uint32_t blockSize, + q15_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q63_t sum = 0; /* Temporary result storage */ + q15_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) && defined (ARM_MATH_DSP) + q31_t in32; /* Temporary variable to store input value */ +#endif + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + /* Compute sum of squares and store result in a temporary variable. */ +#if defined (ARM_MATH_DSP) + in32 = read_q15x2_ia ((q15_t **) &pSrc); + sum = __SMLALD(in32, in32, sum); + + in32 = read_q15x2_ia ((q15_t **) &pSrc); + sum = __SMLALD(in32, in32, sum); +#else + in = *pSrc++; + sum += ((q31_t) in * in); + + in = *pSrc++; + sum += ((q31_t) in * in); + + in = *pSrc++; + sum += ((q31_t) in * in); + + in = *pSrc++; + sum += ((q31_t) in * in); +#endif /* #if defined (ARM_MATH_DSP) */ + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable. */ + sum += ((q31_t) in * in); + + /* Decrement loop counter */ + blkCnt--; + } + + /* Truncating and saturating the accumulator to 1.15 format */ + /* Store result in destination */ + arm_sqrt_q15(__SSAT((sum / (q63_t)blockSize) >> 15, 16), pResult); +} + +/** + @} end of RMS group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_q31.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_q31.c new file mode 100644 index 0000000..4cf086f --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_rms_q31.c @@ -0,0 +1,124 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_rms_q31.c + * Description: Root Mean Square of the elements of a Q31 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup RMS + @{ + */ + +/** + @brief Root Mean Square of the elements of a Q31 vector. + @param[in] pSrc points to the input vector + @param[in] blockSize number of samples in input vector + @param[out] pResult root mean square value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using an internal 64-bit accumulator. + The input is represented in 1.31 format, and intermediate multiplication + yields a 2.62 format. + The accumulator maintains full precision of the intermediate multiplication results, + but provides only a single guard bit. + There is no saturation on intermediate additions. + If the accumulator overflows, it wraps around and distorts the result. + In order to avoid overflows completely, the input signal must be scaled down by + log2(blockSize) bits, as a total of blockSize additions are performed internally. + Finally, the 2.62 accumulator is right shifted by 31 bits to yield a 1.31 format value. + */ + +void arm_rms_q31( + const q31_t * pSrc, + uint32_t blockSize, + q31_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + uint64_t sum = 0; /* Temporary result storage (can get never negative. changed type from q63 to uint64 */ + q31_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable, sum. */ + sum += ((q63_t) in * in); + + in = *pSrc++; + sum += ((q63_t) in * in); + + in = *pSrc++; + sum += ((q63_t) in * in); + + in = *pSrc++; + sum += ((q63_t) in * in); + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable. */ + sum += ((q63_t) in * in); + + /* Decrement loop counter */ + blkCnt--; + } + + /* Convert data in 2.62 to 1.31 by 31 right shifts and saturate */ + /* Compute Rms and store result in destination vector */ + arm_sqrt_q31(clip_q63_to_q31((sum / (q63_t) blockSize) >> 31), pResult); +} + +/** + @} end of RMS group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_f32.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_f32.c new file mode 100644 index 0000000..214a794 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_f32.c @@ -0,0 +1,188 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_std_f32.c + * Description: Standard deviation of the elements of a floating-point vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @defgroup STD Standard deviation + + Calculates the standard deviation of the elements in the input vector. + The underlying algorithm is used: + + 
+      Result = sqrt((sumOfSquares - sum2 / blockSize) / (blockSize - 1))
+
+      sumOfSquares = pSrc[0] * pSrc[0] + pSrc[1] * pSrc[1] + ... + pSrc[blockSize-1] * pSrc[blockSize-1]
+      sum = pSrc[0] + pSrc[1] + pSrc[2] + ... + pSrc[blockSize-1]
+
+ + There are separate functions for floating point, Q31, and Q15 data types. + */ + +/** + @addtogroup STD + @{ + */ + +/** + @brief Standard deviation of the elements 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 standard deviation value returned here + @return none + */ +#if defined(ARM_MATH_NEON_EXPERIMENTAL) +void arm_std_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + float32_t var; + arm_var_f32(pSrc,blockSize,&var); + arm_sqrt_f32(var, pResult); +} +#else +void arm_std_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + float32_t sum = 0.0f; /* Temporary result storage */ + float32_t sumOfSquares = 0.0f; /* Sum of squares */ + float32_t in; /* Temporary variable to store input value */ + +#ifndef ARM_MATH_CM0_FAMILY + float32_t meanOfSquares, mean, squareOfMean; /* Temporary variables */ +#else + float32_t squareOfSum; /* Square of Sum */ + float32_t var; /* Temporary varaince storage */ +#endif + + if (blockSize <= 1U) + { + *pResult = 0; + return; + } + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + sumOfSquares += in * in; + /* Compute sum and store result in a temporary variable, sum. */ + sum += in; + + in = *pSrc++; + sumOfSquares += in * in; + sum += in; + + in = *pSrc++; + sumOfSquares += in * in; + sum += in; + + in = *pSrc++; + sumOfSquares += in * in; + sum += in; + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + sumOfSquares += ( in * in); + /* Compute sum and store result in a temporary variable, sum. */ + sum += in; + + /* Decrement loop counter */ + blkCnt--; + } + +#ifndef ARM_MATH_CM0_FAMILY + + /* Compute Mean of squares and store result in a temporary variable, meanOfSquares. */ + meanOfSquares = sumOfSquares / ((float32_t) blockSize - 1.0f); + + /* Compute mean of all input values */ + mean = sum / (float32_t) blockSize; + + /* Compute square of mean */ + squareOfMean = (mean * mean) * (((float32_t) blockSize) / + ((float32_t) blockSize - 1.0f)); + + /* Compute standard deviation and store result to destination */ + arm_sqrt_f32((meanOfSquares - squareOfMean), pResult); + +#else + /* Run the below code for Cortex-M0 */ + + /* Compute square of sum */ + squareOfSum = ((sum * sum) / (float32_t) blockSize); + + /* Compute variance */ + var = ((sumOfSquares - squareOfSum) / (float32_t) (blockSize - 1.0f)); + + /* Compute standard deviation and store result in destination */ + arm_sqrt_f32(var, pResult); + +#endif /* #ifndef ARM_MATH_CM0_FAMILY */ + +} +#endif /* #if defined(ARM_MATH_NEON) */ + +/** + @} end of STD group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_q15.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_q15.c new file mode 100644 index 0000000..70d142d --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_q15.c @@ -0,0 +1,161 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_std_q15.c + * Description: Standard deviation of an array of Q15 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup STD + @{ + */ + +/** + @brief Standard deviation of the elements of a Q15 vector. + @param[in] pSrc points to the input vector + @param[in] blockSize number of samples in input vector + @param[out] pResult standard deviation value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 64-bit internal accumulator. + The input is represented in 1.15 format. + Intermediate multiplication yields a 2.30 format, and this + result is added without saturation to a 64-bit accumulator in 34.30 format. + With 33 guard bits in the accumulator, there is no risk of overflow, and the + full precision of the intermediate multiplication is preserved. + Finally, the 34.30 result is truncated to 34.15 format by discarding the lower + 15 bits, and then saturated to yield a result in 1.15 format. + */ + +void arm_std_q15( + const q15_t * pSrc, + uint32_t blockSize, + q15_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q31_t sum = 0; /* Accumulator */ + q31_t meanOfSquares, squareOfMean; /* Square of mean and mean of square */ + q63_t sumOfSquares = 0; /* Sum of squares */ + q15_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) && defined (ARM_MATH_DSP) + q31_t in32; /* Temporary variable to store input value */ +#endif + + if (blockSize <= 1U) + { + *pResult = 0; + return; + } + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + /* Compute sum and store result in a temporary variable, sum. */ +#if defined (ARM_MATH_DSP) + in32 = read_q15x2_ia ((q15_t **) &pSrc); + sumOfSquares = __SMLALD(in32, in32, sumOfSquares); + sum += ((in32 << 16U) >> 16U); + sum += (in32 >> 16U); + + in32 = read_q15x2_ia ((q15_t **) &pSrc); + sumOfSquares = __SMLALD(in32, in32, sumOfSquares); + sum += ((in32 << 16U) >> 16U); + sum += (in32 >> 16U); +#else + in = *pSrc++; + sumOfSquares += (in * in); + sum += in; + + in = *pSrc++; + sumOfSquares += (in * in); + sum += in; + + in = *pSrc++; + sumOfSquares += (in * in); + sum += in; + + in = *pSrc++; + sumOfSquares += (in * in); + sum += in; +#endif /* #if defined (ARM_MATH_DSP) */ + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + sumOfSquares += (in * in); + /* Compute sum and store result in a temporary variable, sum. */ + sum += in; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Compute Mean of squares and store result in a temporary variable, meanOfSquares. */ + meanOfSquares = (q31_t) (sumOfSquares / (q63_t)(blockSize - 1U)); + + /* Compute square of mean */ + squareOfMean = (q31_t) ((q63_t) sum * sum / (q63_t)(blockSize * (blockSize - 1U))); + + /* mean of squares minus the square of mean. */ + /* Compute standard deviation and store result in destination */ + arm_sqrt_q15(__SSAT((meanOfSquares - squareOfMean) >> 15U, 16U), pResult); +} + +/** + @} end of STD group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_q31.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_q31.c new file mode 100644 index 0000000..1452a69 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_std_q31.c @@ -0,0 +1,147 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_std_q31.c + * Description: Standard deviation of the elements of a Q31 vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup STD + @{ + */ + +/** + @brief Standard deviation of the elements of a Q31 vector. + @param[in] pSrc points to the input vector. + @param[in] blockSize number of samples in input vector. + @param[out] pResult standard deviation value returned here. + @return none + + @par Scaling and Overflow Behavior + The function is implemented using an internal 64-bit accumulator. + The input is represented in 1.31 format, which is then downshifted by 8 bits + which yields 1.23, and intermediate multiplication yields a 2.46 format. + The accumulator maintains full precision of the intermediate multiplication results, + but provides only a 16 guard bits. + There is no saturation on intermediate additions. + If the accumulator overflows it wraps around and distorts the result. + In order to avoid overflows completely the input signal must be scaled down by + log2(blockSize)-8 bits, as a total of blockSize additions are performed internally. + After division, internal variables should be Q18.46 + Finally, the 18.46 accumulator is right shifted by 15 bits to yield a 1.31 format value. + */ + +void arm_std_q31( + const q31_t * pSrc, + uint32_t blockSize, + q31_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q63_t sum = 0; /* Accumulator */ + q63_t meanOfSquares, squareOfMean; /* Square of mean and mean of square */ + q63_t sumOfSquares = 0; /* Sum of squares */ + q31_t in; /* Temporary variable to store input value */ + + if (blockSize <= 1U) + { + *pResult = 0; + return; + } + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + in = *pSrc++ >> 8U; + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + sumOfSquares += ((q63_t) (in) * (in)); + /* Compute sum and store result in a temporary variable, sum. */ + sum += in; + + in = *pSrc++ >> 8U; + sumOfSquares += ((q63_t) (in) * (in)); + sum += in; + + in = *pSrc++ >> 8U; + sumOfSquares += ((q63_t) (in) * (in)); + sum += in; + + in = *pSrc++ >> 8U; + sumOfSquares += ((q63_t) (in) * (in)); + sum += in; + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + in = *pSrc++ >> 8U; + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + sumOfSquares += ((q63_t) (in) * (in)); + /* Compute sum and store result in a temporary variable, sum. */ + sum += in; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Compute Mean of squares and store result in a temporary variable, meanOfSquares. */ + meanOfSquares = (sumOfSquares / (q63_t)(blockSize - 1U)); + + /* Compute square of mean */ + squareOfMean = ( sum * sum / (q63_t)(blockSize * (blockSize - 1U))); + + /* Compute standard deviation and store result in destination */ + arm_sqrt_q31((meanOfSquares - squareOfMean) >> 15U, pResult); +} + +/** + @} end of STD group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_f32.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_f32.c new file mode 100644 index 0000000..18f5806 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_f32.c @@ -0,0 +1,234 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_var_f32.c + * Description: Variance of the elements of a floating-point vector + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @defgroup variance Variance + + Calculates the variance of the elements in the input vector. + The underlying algorithm used is the direct method sometimes referred to as the two-pass method: + + 
+      Result = sum(element - meanOfElements)^2) / numElement - 1
+
+      meanOfElements = ( pSrc[0] * pSrc[0] + pSrc[1] * pSrc[1] + ... + pSrc[blockSize-1] ) / blockSize
+
+ + There are separate functions for floating point, Q31, and Q15 data types. + */ + +/** + @addtogroup variance + @{ + */ + +/** + @brief Variance of the elements 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 variance value returned here + @return none + */ +#if defined(ARM_MATH_NEON_EXPERIMENTAL) +void arm_var_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + float32_t mean; + + float32_t sum = 0.0f; /* accumulator */ + float32_t in; /* Temporary variable to store input value */ + uint32_t blkCnt; /* loop counter */ + + float32x4_t sumV = vdupq_n_f32(0.0f); /* Temporary result storage */ + float32x2_t sumV2; + float32x4_t inV; + float32x4_t avg; + + arm_mean_f32(pSrc,blockSize,&mean); + avg = vdupq_n_f32(mean); + + blkCnt = blockSize >> 2U; + + /* Compute 4 outputs at a time. + ** a second loop below computes the remaining 1 to 3 samples. */ + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + ... + A[blockSize-1] * A[blockSize-1] */ + /* Compute Power and then store the result in a temporary variable, sum. */ + inV = vld1q_f32(pSrc); + inV = vsubq_f32(inV, avg); + sumV = vmlaq_f32(sumV, inV, inV); + pSrc += 4; + + /* Decrement the loop counter */ + blkCnt--; + } + + sumV2 = vpadd_f32(vget_low_f32(sumV),vget_high_f32(sumV)); + sum = sumV2[0] + sumV2[1]; + + /* If the blockSize is not a multiple of 4, compute any remaining output samples here. + ** No loop unrolling is used. */ + blkCnt = blockSize % 0x4U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + A[2] * A[2] + ... + A[blockSize-1] * A[blockSize-1] */ + /* compute power and then store the result in a temporary variable, sum. */ + in = *pSrc++; + in = in - mean; + sum += in * in; + + /* Decrement the loop counter */ + blkCnt--; + } + + /* Variance */ + *pResult = sum / (float32_t)(blockSize - 1.0f); + +} + +#else +void arm_var_f32( + const float32_t * pSrc, + uint32_t blockSize, + float32_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + float32_t sum = 0.0f; /* Temporary result storage */ + float32_t fSum = 0.0f; + float32_t fMean, fValue; + const float32_t * pInput = pSrc; + + if (blockSize <= 1U) + { + *pResult = 0; + return; + } + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */ + + sum += *pInput++; + sum += *pInput++; + sum += *pInput++; + sum += *pInput++; + + + /* 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] + A[1] + A[2] + ... + A[blockSize-1]) */ + + sum += *pInput++; + + /* Decrement loop counter */ + blkCnt--; + } + + /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) / blockSize */ + fMean = sum / (float32_t) blockSize; + + pInput = pSrc; + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + fValue = *pInput++ - fMean; + fSum += fValue * fValue; + + fValue = *pInput++ - fMean; + fSum += fValue * fValue; + + fValue = *pInput++ - fMean; + fSum += fValue * fValue; + + fValue = *pInput++ - fMean; + fSum += fValue * fValue; + + /* 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) + { + fValue = *pInput++ - fMean; + fSum += fValue * fValue; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Variance */ + *pResult = fSum / (float32_t)(blockSize - 1.0f); +} +#endif /* #if defined(ARM_MATH_NEON) */ + +/** + @} end of variance group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_q15.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_q15.c new file mode 100644 index 0000000..4218522 --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_q15.c @@ -0,0 +1,164 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_var_q15.c + * Description: Variance of an array of Q15 type + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup variance + @{ + */ + +/** + @brief Variance of the elements of a Q15 vector. + @param[in] pSrc points to the input vector + @param[in] blockSize number of samples in input vector + @param[out] pResult variance value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using a 64-bit internal accumulator. + The input is represented in 1.15 format. + Intermediate multiplication yields a 2.30 format, and this + result is added without saturation to a 64-bit accumulator in 34.30 format. + With 33 guard bits in the accumulator, there is no risk of overflow, and the + full precision of the intermediate multiplication is preserved. + Finally, the 34.30 result is truncated to 34.15 format by discarding the lower + 15 bits, and then saturated to yield a result in 1.15 format. + */ + +void arm_var_q15( + const q15_t * pSrc, + uint32_t blockSize, + q15_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q31_t sum = 0; /* Accumulator */ + q31_t meanOfSquares, squareOfMean; /* Square of mean and mean of square */ + q63_t sumOfSquares = 0; /* Sum of squares */ + q15_t in; /* Temporary variable to store input value */ + +#if defined (ARM_MATH_LOOPUNROLL) && defined (ARM_MATH_DSP) + q31_t in32; /* Temporary variable to store input value */ +#endif + + if (blockSize <= 1U) + { + *pResult = 0; + return; + } + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + /* Compute sum and store result in a temporary variable, sum. */ +#if defined (ARM_MATH_DSP) + in32 = read_q15x2_ia ((q15_t **) &pSrc); + sumOfSquares = __SMLALD(in32, in32, sumOfSquares); + sum += ((in32 << 16U) >> 16U); + sum += (in32 >> 16U); + + in32 = read_q15x2_ia ((q15_t **) &pSrc); + sumOfSquares = __SMLALD(in32, in32, sumOfSquares); + sum += ((in32 << 16U) >> 16U); + sum += (in32 >> 16U); +#else + in = *pSrc++; + sumOfSquares += (in * in); + sum += in; + + in = *pSrc++; + sumOfSquares += (in * in); + sum += in; + + in = *pSrc++; + sumOfSquares += (in * in); + sum += in; + + in = *pSrc++; + sumOfSquares += (in * in); + sum += in; +#endif /* #if defined (ARM_MATH_DSP) */ + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + in = *pSrc++; + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ +#if defined (ARM_MATH_DSP) + sumOfSquares = __SMLALD(in, in, sumOfSquares); +#else + sumOfSquares += (in * in); +#endif /* #if defined (ARM_MATH_DSP) */ + /* Compute sum and store result in a temporary variable, sum. */ + sum += in; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Compute Mean of squares and store result in a temporary variable, meanOfSquares. */ + meanOfSquares = (q31_t) (sumOfSquares / (q63_t)(blockSize - 1U)); + + /* Compute square of mean */ + squareOfMean = (q31_t) ((q63_t) sum * sum / (q63_t)(blockSize * (blockSize - 1U))); + + /* mean of squares minus the square of mean. */ + *pResult = (meanOfSquares - squareOfMean) >> 15U; +} + +/** + @} end of variance group + */ diff --git a/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_q31.c b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_q31.c new file mode 100644 index 0000000..6b552db --- /dev/null +++ b/Drivers/CMSIS/DSP/Source/StatisticsFunctions/arm_var_q31.c @@ -0,0 +1,147 @@ +/* ---------------------------------------------------------------------- + * Project: CMSIS DSP Library + * Title: arm_var_q31.c + * Description: Variance of an array of Q31 type + * + * $Date: 18. March 2019 + * $Revision: V1.6.0 + * + * Target Processor: Cortex-M cores + * -------------------------------------------------------------------- */ +/* + * 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. + */ + +#include "arm_math.h" + +/** + @ingroup groupStats + */ + +/** + @addtogroup variance + @{ + */ + +/** + @brief Variance of the elements of a Q31 vector. + @param[in] pSrc points to the input vector + @param[in] blockSize number of samples in input vector + @param[out] pResult variance value returned here + @return none + + @par Scaling and Overflow Behavior + The function is implemented using an internal 64-bit accumulator. + The input is represented in 1.31 format, which is then downshifted by 8 bits + which yields 1.23, and intermediate multiplication yields a 2.46 format. + The accumulator maintains full precision of the intermediate multiplication results, + but provides only a 16 guard bits. + There is no saturation on intermediate additions. + If the accumulator overflows it wraps around and distorts the result. + In order to avoid overflows completely the input signal must be scaled down by + log2(blockSize)-8 bits, as a total of blockSize additions are performed internally. + After division, internal variables should be Q18.46 + Finally, the 18.46 accumulator is right shifted by 15 bits to yield a 1.31 format value. + */ + +void arm_var_q31( + const q31_t * pSrc, + uint32_t blockSize, + q31_t * pResult) +{ + uint32_t blkCnt; /* Loop counter */ + q63_t sum = 0; /* Temporary result storage */ + q63_t meanOfSquares, squareOfMean; /* Square of mean and mean of square */ + q63_t sumOfSquares = 0; /* Sum of squares */ + q31_t in; /* Temporary variable to store input value */ + + if (blockSize <= 1U) + { + *pResult = 0; + return; + } + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = blockSize >> 2U; + + while (blkCnt > 0U) + { + /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + in = *pSrc++ >> 8U; + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + sumOfSquares += ((q63_t) (in) * (in)); + /* Compute sum and store result in a temporary variable, sum. */ + sum += in; + + in = *pSrc++ >> 8U; + sumOfSquares += ((q63_t) (in) * (in)); + sum += in; + + in = *pSrc++ >> 8U; + sumOfSquares += ((q63_t) (in) * (in)); + sum += in; + + in = *pSrc++ >> 8U; + sumOfSquares += ((q63_t) (in) * (in)); + sum += in; + + /* 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] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */ + /* C = A[0] + A[1] + ... + A[blockSize-1] */ + + in = *pSrc++ >> 8U; + /* Compute sum of squares and store result in a temporary variable, sumOfSquares. */ + sumOfSquares += ((q63_t) (in) * (in)); + /* Compute sum and store result in a temporary variable, sum. */ + sum += in; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Compute Mean of squares and store result in a temporary variable, meanOfSquares. */ + meanOfSquares = (sumOfSquares / (q63_t)(blockSize - 1U)); + + /* Compute square of mean */ + squareOfMean = ( sum * sum / (q63_t)(blockSize * (blockSize - 1U))); + + /* Compute variance and store result in destination */ + *pResult = (meanOfSquares - squareOfMean) >> 15U; +} + +/** + @} end of variance group + */ -- cgit v1.2.3