arm_mat_sub_f32.c
1 /* ---------------------------------------------------------------------- 2 * Project: CMSIS DSP Library 3 * Title: arm_mat_sub_f32.c 4 * Description: Floating-point matrix subtraction 5 * 6 * $Date: 23 April 2021 7 * $Revision: V1.9.0 8 * 9 * Target Processor: Cortex-M and Cortex-A cores 10 * -------------------------------------------------------------------- */ 11 /* 12 * Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved. 13 * 14 * SPDX-License-Identifier: Apache-2.0 15 * 16 * Licensed under the Apache License, Version 2.0 (the License); you may 17 * not use this file except in compliance with the License. 18 * You may obtain a copy of the License at 19 * 20 * www.apache.org/licenses/LICENSE-2.0 21 * 22 * Unless required by applicable law or agreed to in writing, software 23 * distributed under the License is distributed on an AS IS BASIS, WITHOUT 24 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 25 * See the License for the specific language governing permissions and 26 * limitations under the License. 27 */ 28 29 #include "dsp/matrix_functions.h" 30 31 /** 32 @ingroup groupMatrix 33 */ 34 35 /** 36 @defgroup MatrixSub Matrix Subtraction 37 38 Subtract two matrices. 39 \image html MatrixSubtraction.gif "Subraction of two 3 x 3 matrices" 40 41 The functions check to make sure that 42 <code>pSrcA</code>, <code>pSrcB</code>, and <code>pDst</code> have the same 43 number of rows and columns. 44 */ 45 46 /** 47 @addtogroup MatrixSub 48 @{ 49 */ 50 51 /** 52 @brief Floating-point matrix subtraction. 53 @param[in] pSrcA points to the first input matrix structure 54 @param[in] pSrcB points to the second input matrix structure 55 @param[out] pDst points to output matrix structure 56 @return execution status 57 - \ref ARM_MATH_SUCCESS : Operation successful 58 - \ref ARM_MATH_SIZE_MISMATCH : Matrix size check failed 59 */ 60 #if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) 61 arm_status arm_mat_sub_f32( 62 const arm_matrix_instance_f32 * pSrcA, 63 const arm_matrix_instance_f32 * pSrcB, 64 arm_matrix_instance_f32 * pDst) 65 { 66 arm_status status; /* status of matrix subtraction */ 67 uint32_t numSamples; /* total number of elements in the matrix */ 68 float32_t *pDataA, *pDataB, *pDataDst; 69 f32x4_t vecA, vecB, vecDst; 70 float32_t const *pSrcAVec; 71 float32_t const *pSrcBVec; 72 uint32_t blkCnt; /* loop counters */ 73 74 pDataA = pSrcA->pData; 75 pDataB = pSrcB->pData; 76 pDataDst = pDst->pData; 77 pSrcAVec = (float32_t const *) pDataA; 78 pSrcBVec = (float32_t const *) pDataB; 79 80 #ifdef ARM_MATH_MATRIX_CHECK 81 /* Check for matrix mismatch condition */ 82 if ((pSrcA->numRows != pSrcB->numRows) || 83 (pSrcA->numCols != pSrcB->numCols) || 84 (pSrcA->numRows != pDst->numRows) || (pSrcA->numCols != pDst->numCols)) 85 { 86 /* Set status as ARM_MATH_SIZE_MISMATCH */ 87 status = ARM_MATH_SIZE_MISMATCH; 88 } 89 else 90 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */ 91 { 92 /* 93 * Total number of samples in the input matrix 94 */ 95 numSamples = (uint32_t) pSrcA->numRows * pSrcA->numCols; 96 blkCnt = numSamples >> 2; 97 while (blkCnt > 0U) 98 { 99 /* C(m,n) = A(m,n) + B(m,n) */ 100 /* sub and then store the results in the destination buffer. */ 101 vecA = vld1q(pSrcAVec); 102 pSrcAVec += 4; 103 vecB = vld1q(pSrcBVec); 104 pSrcBVec += 4; 105 vecDst = vsubq(vecA, vecB); 106 vst1q(pDataDst, vecDst); 107 pDataDst += 4; 108 /* 109 * Decrement the blockSize loop counter 110 */ 111 blkCnt--; 112 } 113 /* 114 * tail 115 * (will be merged thru tail predication) 116 */ 117 blkCnt = numSamples & 3; 118 if (blkCnt > 0U) 119 { 120 mve_pred16_t p0 = vctp32q(blkCnt); 121 vecA = vld1q(pSrcAVec); 122 vecB = vld1q(pSrcBVec); 123 vecDst = vsubq_m(vecDst, vecA, vecB, p0); 124 vstrwq_p(pDataDst, vecDst, p0); 125 } 126 status = ARM_MATH_SUCCESS; 127 } 128 129 /* Return to application */ 130 return (status); 131 } 132 133 #else 134 #if defined(ARM_MATH_NEON) 135 arm_status arm_mat_sub_f32( 136 const arm_matrix_instance_f32 * pSrcA, 137 const arm_matrix_instance_f32 * pSrcB, 138 arm_matrix_instance_f32 * pDst) 139 { 140 float32_t *pIn1 = pSrcA->pData; /* input data matrix pointer A */ 141 float32_t *pIn2 = pSrcB->pData; /* input data matrix pointer B */ 142 float32_t *pOut = pDst->pData; /* output data matrix pointer */ 143 144 145 uint32_t numSamples; /* total number of elements in the matrix */ 146 uint32_t blkCnt; /* loop counters */ 147 arm_status status; /* status of matrix subtraction */ 148 149 #ifdef ARM_MATH_MATRIX_CHECK 150 /* Check for matrix mismatch condition */ 151 if ((pSrcA->numRows != pSrcB->numRows) || 152 (pSrcA->numCols != pSrcB->numCols) || 153 (pSrcA->numRows != pDst->numRows) || (pSrcA->numCols != pDst->numCols)) 154 { 155 /* Set status as ARM_MATH_SIZE_MISMATCH */ 156 status = ARM_MATH_SIZE_MISMATCH; 157 } 158 else 159 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */ 160 { 161 float32x4_t vec1; 162 float32x4_t vec2; 163 float32x4_t res; 164 165 /* Total number of samples in the input matrix */ 166 numSamples = (uint32_t) pSrcA->numRows * pSrcA->numCols; 167 168 blkCnt = numSamples >> 2U; 169 170 /* Compute 4 outputs at a time. 171 ** a second loop below computes the remaining 1 to 3 samples. */ 172 while (blkCnt > 0U) 173 { 174 /* C(m,n) = A(m,n) - B(m,n) */ 175 /* Subtract and then store the results in the destination buffer. */ 176 /* Read values from source A */ 177 vec1 = vld1q_f32(pIn1); 178 vec2 = vld1q_f32(pIn2); 179 res = vsubq_f32(vec1, vec2); 180 vst1q_f32(pOut, res); 181 182 /* Update pointers to process next samples */ 183 pIn1 += 4U; 184 pIn2 += 4U; 185 pOut += 4U; 186 187 /* Decrement the loop counter */ 188 blkCnt--; 189 } 190 191 /* If the numSamples is not a multiple of 4, compute any remaining output samples here. 192 ** No loop unrolling is used. */ 193 blkCnt = numSamples % 0x4U; 194 195 196 while (blkCnt > 0U) 197 { 198 /* C(m,n) = A(m,n) - B(m,n) */ 199 /* Subtract and then store the results in the destination buffer. */ 200 *pOut++ = (*pIn1++) - (*pIn2++); 201 202 /* Decrement the loop counter */ 203 blkCnt--; 204 } 205 206 /* Set status as ARM_MATH_SUCCESS */ 207 status = ARM_MATH_SUCCESS; 208 } 209 210 /* Return to application */ 211 return (status); 212 } 213 #else 214 arm_status arm_mat_sub_f32( 215 const arm_matrix_instance_f32 * pSrcA, 216 const arm_matrix_instance_f32 * pSrcB, 217 arm_matrix_instance_f32 * pDst) 218 { 219 float32_t *pInA = pSrcA->pData; /* input data matrix pointer A */ 220 float32_t *pInB = pSrcB->pData; /* input data matrix pointer B */ 221 float32_t *pOut = pDst->pData; /* output data matrix pointer */ 222 223 uint32_t numSamples; /* total number of elements in the matrix */ 224 uint32_t blkCnt; /* loop counters */ 225 arm_status status; /* status of matrix subtraction */ 226 227 #ifdef ARM_MATH_MATRIX_CHECK 228 229 /* Check for matrix mismatch condition */ 230 if ((pSrcA->numRows != pSrcB->numRows) || 231 (pSrcA->numCols != pSrcB->numCols) || 232 (pSrcA->numRows != pDst->numRows) || 233 (pSrcA->numCols != pDst->numCols) ) 234 { 235 /* Set status as ARM_MATH_SIZE_MISMATCH */ 236 status = ARM_MATH_SIZE_MISMATCH; 237 } 238 else 239 240 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */ 241 242 { 243 /* Total number of samples in input matrix */ 244 numSamples = (uint32_t) pSrcA->numRows * pSrcA->numCols; 245 246 #if defined (ARM_MATH_LOOPUNROLL) 247 248 /* Loop unrolling: Compute 4 outputs at a time */ 249 blkCnt = numSamples >> 2U; 250 251 while (blkCnt > 0U) 252 { 253 /* C(m,n) = A(m,n) - B(m,n) */ 254 255 /* Subtract and store result in destination buffer. */ 256 *pOut++ = (*pInA++) - (*pInB++); 257 *pOut++ = (*pInA++) - (*pInB++); 258 *pOut++ = (*pInA++) - (*pInB++); 259 *pOut++ = (*pInA++) - (*pInB++); 260 261 /* Decrement loop counter */ 262 blkCnt--; 263 } 264 265 /* Loop unrolling: Compute remaining outputs */ 266 blkCnt = numSamples % 0x4U; 267 268 #else 269 270 /* Initialize blkCnt with number of samples */ 271 blkCnt = numSamples; 272 273 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */ 274 275 while (blkCnt > 0U) 276 { 277 /* C(m,n) = A(m,n) - B(m,n) */ 278 279 /* Subtract and store result in destination buffer. */ 280 *pOut++ = (*pInA++) - (*pInB++); 281 282 /* Decrement loop counter */ 283 blkCnt--; 284 } 285 286 /* Set status as ARM_MATH_SUCCESS */ 287 status = ARM_MATH_SUCCESS; 288 } 289 290 /* Return to application */ 291 return (status); 292 } 293 #endif /* #if defined(ARM_MATH_NEON) */ 294 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */ 295 296 /** 297 @} end of MatrixSub group 298 */