Laxmi Kant Tiwari / mbed-dsp

CMSIS DSP Lib

Fork of mbed-dsp by mbed official

cmsis_dsp/MatrixFunctions/arm_mat_mult_f32.c@2:da51fb522205, 2013-05-30 (annotated)

Committer:
emilmont
Date:
Thu May 30 17:10:11 2013 +0100
Revision:
2:da51fb522205
Parent:
1:fdd22bb7aa52
Child:
3:7a284390b0ce
Keep "cmsis-dsp" module in synch with its source

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emilmont 1:fdd22bb7aa52 1 /* ----------------------------------------------------------------------
emilmont 1:fdd22bb7aa52 2 * Copyright (C) 2010 ARM Limited. All rights reserved.
emilmont 1:fdd22bb7aa52 3 *
emilmont 1:fdd22bb7aa52 4 * $Date: 15. February 2012
emilmont 2:da51fb522205 5 * $Revision: V1.1.0
emilmont 1:fdd22bb7aa52 6 *
emilmont 2:da51fb522205 7 * Project: CMSIS DSP Library
emilmont 2:da51fb522205 8 * Title: arm_mat_mult_f32.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 1:fdd22bb7aa52 10 * Description: Floating-point matrix multiplication.
emilmont 1:fdd22bb7aa52 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
emilmont 1:fdd22bb7aa52 13 *
emilmont 1:fdd22bb7aa52 14 * Version 1.1.0 2012/02/15
emilmont 1:fdd22bb7aa52 15 * Updated with more optimizations, bug fixes and minor API changes.
emilmont 1:fdd22bb7aa52 16 *
emilmont 1:fdd22bb7aa52 17 * Version 1.0.10 2011/7/15
emilmont 1:fdd22bb7aa52 18 * Big Endian support added and Merged M0 and M3/M4 Source code.
emilmont 1:fdd22bb7aa52 19 *
emilmont 1:fdd22bb7aa52 20 * Version 1.0.3 2010/11/29
emilmont 1:fdd22bb7aa52 21 * Re-organized the CMSIS folders and updated documentation.
emilmont 1:fdd22bb7aa52 22 *
emilmont 1:fdd22bb7aa52 23 * Version 1.0.2 2010/11/11
emilmont 1:fdd22bb7aa52 24 * Documentation updated.
emilmont 1:fdd22bb7aa52 25 *
emilmont 1:fdd22bb7aa52 26 * Version 1.0.1 2010/10/05
emilmont 1:fdd22bb7aa52 27 * Production release and review comments incorporated.
emilmont 1:fdd22bb7aa52 28 *
emilmont 1:fdd22bb7aa52 29 * Version 1.0.0 2010/09/20
emilmont 1:fdd22bb7aa52 30 * Production release and review comments incorporated.
emilmont 1:fdd22bb7aa52 31 *
emilmont 1:fdd22bb7aa52 32 * Version 0.0.5 2010/04/26
emilmont 1:fdd22bb7aa52 33 * incorporated review comments and updated with latest CMSIS layer
emilmont 1:fdd22bb7aa52 34 *
emilmont 1:fdd22bb7aa52 35 * Version 0.0.3 2010/03/10
emilmont 1:fdd22bb7aa52 36 * Initial version
emilmont 1:fdd22bb7aa52 37 * -------------------------------------------------------------------- */
emilmont 1:fdd22bb7aa52 38
emilmont 1:fdd22bb7aa52 39 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 40
emilmont 1:fdd22bb7aa52 41 /**
emilmont 1:fdd22bb7aa52 42 * @ingroup groupMatrix
emilmont 1:fdd22bb7aa52 43 */
emilmont 1:fdd22bb7aa52 44
emilmont 1:fdd22bb7aa52 45 /**
emilmont 1:fdd22bb7aa52 46 * @defgroup MatrixMult Matrix Multiplication
emilmont 1:fdd22bb7aa52 47 *
emilmont 1:fdd22bb7aa52 48 * Multiplies two matrices.
emilmont 1:fdd22bb7aa52 49 *
emilmont 1:fdd22bb7aa52 50 * \image html MatrixMultiplication.gif "Multiplication of two 3 x 3 matrices"
emilmont 1:fdd22bb7aa52 51
emilmont 1:fdd22bb7aa52 52 * Matrix multiplication is only defined if the number of columns of the
emilmont 1:fdd22bb7aa52 53 * first matrix equals the number of rows of the second matrix.
emilmont 1:fdd22bb7aa52 54 * Multiplying an <code>M x N</code> matrix with an <code>N x P</code> matrix results
emilmont 1:fdd22bb7aa52 55 * in an <code>M x P</code> matrix.
emilmont 1:fdd22bb7aa52 56 * When matrix size checking is enabled, the functions check: (1) that the inner dimensions of
emilmont 1:fdd22bb7aa52 57 * <code>pSrcA</code> and <code>pSrcB</code> are equal; and (2) that the size of the output
emilmont 1:fdd22bb7aa52 58 * matrix equals the outer dimensions of <code>pSrcA</code> and <code>pSrcB</code>.
emilmont 1:fdd22bb7aa52 59 */
emilmont 1:fdd22bb7aa52 60
emilmont 1:fdd22bb7aa52 61
emilmont 1:fdd22bb7aa52 62 /**
emilmont 1:fdd22bb7aa52 63 * @addtogroup MatrixMult
emilmont 1:fdd22bb7aa52 64 * @{
emilmont 1:fdd22bb7aa52 65 */
emilmont 1:fdd22bb7aa52 66
emilmont 1:fdd22bb7aa52 67 /**
emilmont 1:fdd22bb7aa52 68 * @brief Floating-point matrix multiplication.
emilmont 1:fdd22bb7aa52 69 * @param[in] *pSrcA points to the first input matrix structure
emilmont 1:fdd22bb7aa52 70 * @param[in] *pSrcB points to the second input matrix structure
emilmont 1:fdd22bb7aa52 71 * @param[out] *pDst points to output matrix structure
emilmont 2:da51fb522205 72 * @return The function returns either
emilmont 1:fdd22bb7aa52 73 * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
emilmont 1:fdd22bb7aa52 74 */
emilmont 1:fdd22bb7aa52 75
emilmont 1:fdd22bb7aa52 76 arm_status arm_mat_mult_f32(
emilmont 1:fdd22bb7aa52 77 const arm_matrix_instance_f32 * pSrcA,
emilmont 1:fdd22bb7aa52 78 const arm_matrix_instance_f32 * pSrcB,
emilmont 1:fdd22bb7aa52 79 arm_matrix_instance_f32 * pDst)
emilmont 1:fdd22bb7aa52 80 {
emilmont 1:fdd22bb7aa52 81 float32_t *pIn1 = pSrcA->pData; /* input data matrix pointer A */
emilmont 1:fdd22bb7aa52 82 float32_t *pIn2 = pSrcB->pData; /* input data matrix pointer B */
emilmont 1:fdd22bb7aa52 83 float32_t *pInA = pSrcA->pData; /* input data matrix pointer A */
emilmont 1:fdd22bb7aa52 84 float32_t *pOut = pDst->pData; /* output data matrix pointer */
emilmont 1:fdd22bb7aa52 85 float32_t *px; /* Temporary output data matrix pointer */
emilmont 1:fdd22bb7aa52 86 float32_t sum; /* Accumulator */
emilmont 1:fdd22bb7aa52 87 uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */
emilmont 1:fdd22bb7aa52 88 uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */
emilmont 1:fdd22bb7aa52 89 uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */
emilmont 1:fdd22bb7aa52 90
emilmont 1:fdd22bb7aa52 91 #ifndef ARM_MATH_CM0
emilmont 1:fdd22bb7aa52 92
emilmont 1:fdd22bb7aa52 93 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 94
emilmont 1:fdd22bb7aa52 95 float32_t in1, in2, in3, in4;
emilmont 1:fdd22bb7aa52 96 uint16_t col, i = 0u, j, row = numRowsA, colCnt; /* loop counters */
emilmont 1:fdd22bb7aa52 97 arm_status status; /* status of matrix multiplication */
emilmont 1:fdd22bb7aa52 98
emilmont 1:fdd22bb7aa52 99 #ifdef ARM_MATH_MATRIX_CHECK
emilmont 1:fdd22bb7aa52 100
emilmont 1:fdd22bb7aa52 101
emilmont 1:fdd22bb7aa52 102 /* Check for matrix mismatch condition */
emilmont 1:fdd22bb7aa52 103 if((pSrcA->numCols != pSrcB->numRows) ||
emilmont 1:fdd22bb7aa52 104 (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
emilmont 1:fdd22bb7aa52 105 {
emilmont 1:fdd22bb7aa52 106
emilmont 1:fdd22bb7aa52 107 /* Set status as ARM_MATH_SIZE_MISMATCH */
emilmont 1:fdd22bb7aa52 108 status = ARM_MATH_SIZE_MISMATCH;
emilmont 1:fdd22bb7aa52 109 }
emilmont 1:fdd22bb7aa52 110 else
emilmont 1:fdd22bb7aa52 111 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
emilmont 1:fdd22bb7aa52 112
emilmont 1:fdd22bb7aa52 113 {
emilmont 1:fdd22bb7aa52 114 /* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */
emilmont 1:fdd22bb7aa52 115 /* row loop */
emilmont 1:fdd22bb7aa52 116 do
emilmont 1:fdd22bb7aa52 117 {
emilmont 1:fdd22bb7aa52 118 /* Output pointer is set to starting address of the row being processed */
emilmont 1:fdd22bb7aa52 119 px = pOut + i;
emilmont 1:fdd22bb7aa52 120
emilmont 1:fdd22bb7aa52 121 /* For every row wise process, the column loop counter is to be initiated */
emilmont 1:fdd22bb7aa52 122 col = numColsB;
emilmont 1:fdd22bb7aa52 123
emilmont 1:fdd22bb7aa52 124 /* For every row wise process, the pIn2 pointer is set
emilmont 1:fdd22bb7aa52 125 ** to the starting address of the pSrcB data */
emilmont 1:fdd22bb7aa52 126 pIn2 = pSrcB->pData;
emilmont 1:fdd22bb7aa52 127
emilmont 1:fdd22bb7aa52 128 j = 0u;
emilmont 1:fdd22bb7aa52 129
emilmont 1:fdd22bb7aa52 130 /* column loop */
emilmont 1:fdd22bb7aa52 131 do
emilmont 1:fdd22bb7aa52 132 {
emilmont 1:fdd22bb7aa52 133 /* Set the variable sum, that acts as accumulator, to zero */
emilmont 1:fdd22bb7aa52 134 sum = 0.0f;
emilmont 1:fdd22bb7aa52 135
emilmont 1:fdd22bb7aa52 136 /* Initiate the pointer pIn1 to point to the starting address of the column being processed */
emilmont 1:fdd22bb7aa52 137 pIn1 = pInA;
emilmont 1:fdd22bb7aa52 138
emilmont 1:fdd22bb7aa52 139 /* Apply loop unrolling and compute 4 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 140 colCnt = numColsA >> 2u;
emilmont 1:fdd22bb7aa52 141
emilmont 1:fdd22bb7aa52 142 /* matrix multiplication */
emilmont 1:fdd22bb7aa52 143 while(colCnt > 0u)
emilmont 1:fdd22bb7aa52 144 {
emilmont 1:fdd22bb7aa52 145 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
emilmont 1:fdd22bb7aa52 146 in3 = *pIn2;
emilmont 1:fdd22bb7aa52 147 pIn2 += numColsB;
emilmont 1:fdd22bb7aa52 148 in1 = pIn1[0];
emilmont 1:fdd22bb7aa52 149 in2 = pIn1[1];
emilmont 1:fdd22bb7aa52 150 sum += in1 * in3;
emilmont 1:fdd22bb7aa52 151 in4 = *pIn2;
emilmont 1:fdd22bb7aa52 152 pIn2 += numColsB;
emilmont 1:fdd22bb7aa52 153 sum += in2 * in4;
emilmont 1:fdd22bb7aa52 154
emilmont 1:fdd22bb7aa52 155 in3 = *pIn2;
emilmont 1:fdd22bb7aa52 156 pIn2 += numColsB;
emilmont 1:fdd22bb7aa52 157 in1 = pIn1[2];
emilmont 1:fdd22bb7aa52 158 in2 = pIn1[3];
emilmont 1:fdd22bb7aa52 159 sum += in1 * in3;
emilmont 1:fdd22bb7aa52 160 in4 = *pIn2;
emilmont 1:fdd22bb7aa52 161 pIn2 += numColsB;
emilmont 1:fdd22bb7aa52 162 sum += in2 * in4;
emilmont 1:fdd22bb7aa52 163 pIn1 += 4u;
emilmont 1:fdd22bb7aa52 164
emilmont 1:fdd22bb7aa52 165 /* Decrement the loop count */
emilmont 1:fdd22bb7aa52 166 colCnt--;
emilmont 1:fdd22bb7aa52 167 }
emilmont 1:fdd22bb7aa52 168
emilmont 1:fdd22bb7aa52 169 /* If the columns of pSrcA is not a multiple of 4, compute any remaining MACs here.
emilmont 1:fdd22bb7aa52 170 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 171 colCnt = numColsA % 0x4u;
emilmont 1:fdd22bb7aa52 172
emilmont 1:fdd22bb7aa52 173 while(colCnt > 0u)
emilmont 1:fdd22bb7aa52 174 {
emilmont 1:fdd22bb7aa52 175 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
emilmont 1:fdd22bb7aa52 176 sum += *pIn1++ * (*pIn2);
emilmont 1:fdd22bb7aa52 177 pIn2 += numColsB;
emilmont 1:fdd22bb7aa52 178
emilmont 1:fdd22bb7aa52 179 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 180 colCnt--;
emilmont 1:fdd22bb7aa52 181 }
emilmont 1:fdd22bb7aa52 182
emilmont 1:fdd22bb7aa52 183 /* Store the result in the destination buffer */
emilmont 1:fdd22bb7aa52 184 *px++ = sum;
emilmont 1:fdd22bb7aa52 185
emilmont 1:fdd22bb7aa52 186 /* Update the pointer pIn2 to point to the starting address of the next column */
emilmont 1:fdd22bb7aa52 187 j++;
emilmont 1:fdd22bb7aa52 188 pIn2 = pSrcB->pData + j;
emilmont 1:fdd22bb7aa52 189
emilmont 1:fdd22bb7aa52 190 /* Decrement the column loop counter */
emilmont 1:fdd22bb7aa52 191 col--;
emilmont 1:fdd22bb7aa52 192
emilmont 1:fdd22bb7aa52 193 } while(col > 0u);
emilmont 1:fdd22bb7aa52 194
emilmont 1:fdd22bb7aa52 195 #else
emilmont 1:fdd22bb7aa52 196
emilmont 1:fdd22bb7aa52 197 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 198
emilmont 1:fdd22bb7aa52 199 float32_t *pInB = pSrcB->pData; /* input data matrix pointer B */
emilmont 1:fdd22bb7aa52 200 uint16_t col, i = 0u, row = numRowsA, colCnt; /* loop counters */
emilmont 1:fdd22bb7aa52 201 arm_status status; /* status of matrix multiplication */
emilmont 1:fdd22bb7aa52 202
emilmont 1:fdd22bb7aa52 203 #ifdef ARM_MATH_MATRIX_CHECK
emilmont 1:fdd22bb7aa52 204
emilmont 1:fdd22bb7aa52 205 /* Check for matrix mismatch condition */
emilmont 1:fdd22bb7aa52 206 if((pSrcA->numCols != pSrcB->numRows) ||
emilmont 1:fdd22bb7aa52 207 (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
emilmont 1:fdd22bb7aa52 208 {
emilmont 1:fdd22bb7aa52 209
emilmont 1:fdd22bb7aa52 210 /* Set status as ARM_MATH_SIZE_MISMATCH */
emilmont 1:fdd22bb7aa52 211 status = ARM_MATH_SIZE_MISMATCH;
emilmont 1:fdd22bb7aa52 212 }
emilmont 1:fdd22bb7aa52 213 else
emilmont 1:fdd22bb7aa52 214 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
emilmont 1:fdd22bb7aa52 215
emilmont 1:fdd22bb7aa52 216 {
emilmont 1:fdd22bb7aa52 217 /* The following loop performs the dot-product of each row in pInA with each column in pInB */
emilmont 1:fdd22bb7aa52 218 /* row loop */
emilmont 1:fdd22bb7aa52 219 do
emilmont 1:fdd22bb7aa52 220 {
emilmont 1:fdd22bb7aa52 221 /* Output pointer is set to starting address of the row being processed */
emilmont 1:fdd22bb7aa52 222 px = pOut + i;
emilmont 1:fdd22bb7aa52 223
emilmont 1:fdd22bb7aa52 224 /* For every row wise process, the column loop counter is to be initiated */
emilmont 1:fdd22bb7aa52 225 col = numColsB;
emilmont 1:fdd22bb7aa52 226
emilmont 1:fdd22bb7aa52 227 /* For every row wise process, the pIn2 pointer is set
emilmont 1:fdd22bb7aa52 228 ** to the starting address of the pSrcB data */
emilmont 1:fdd22bb7aa52 229 pIn2 = pSrcB->pData;
emilmont 1:fdd22bb7aa52 230
emilmont 1:fdd22bb7aa52 231 /* column loop */
emilmont 1:fdd22bb7aa52 232 do
emilmont 1:fdd22bb7aa52 233 {
emilmont 1:fdd22bb7aa52 234 /* Set the variable sum, that acts as accumulator, to zero */
emilmont 1:fdd22bb7aa52 235 sum = 0.0f;
emilmont 1:fdd22bb7aa52 236
emilmont 1:fdd22bb7aa52 237 /* Initialize the pointer pIn1 to point to the starting address of the row being processed */
emilmont 1:fdd22bb7aa52 238 pIn1 = pInA;
emilmont 1:fdd22bb7aa52 239
emilmont 1:fdd22bb7aa52 240 /* Matrix A columns number of MAC operations are to be performed */
emilmont 1:fdd22bb7aa52 241 colCnt = numColsA;
emilmont 1:fdd22bb7aa52 242
emilmont 1:fdd22bb7aa52 243 while(colCnt > 0u)
emilmont 1:fdd22bb7aa52 244 {
emilmont 1:fdd22bb7aa52 245 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
emilmont 1:fdd22bb7aa52 246 sum += *pIn1++ * (*pIn2);
emilmont 1:fdd22bb7aa52 247 pIn2 += numColsB;
emilmont 1:fdd22bb7aa52 248
emilmont 1:fdd22bb7aa52 249 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 250 colCnt--;
emilmont 1:fdd22bb7aa52 251 }
emilmont 1:fdd22bb7aa52 252
emilmont 1:fdd22bb7aa52 253 /* Store the result in the destination buffer */
emilmont 1:fdd22bb7aa52 254 *px++ = sum;
emilmont 1:fdd22bb7aa52 255
emilmont 1:fdd22bb7aa52 256 /* Decrement the column loop counter */
emilmont 1:fdd22bb7aa52 257 col--;
emilmont 1:fdd22bb7aa52 258
emilmont 1:fdd22bb7aa52 259 /* Update the pointer pIn2 to point to the starting address of the next column */
emilmont 1:fdd22bb7aa52 260 pIn2 = pInB + (numColsB - col);
emilmont 1:fdd22bb7aa52 261
emilmont 1:fdd22bb7aa52 262 } while(col > 0u);
emilmont 1:fdd22bb7aa52 263
emilmont 1:fdd22bb7aa52 264 #endif /* #ifndef ARM_MATH_CM0 */
emilmont 1:fdd22bb7aa52 265
emilmont 1:fdd22bb7aa52 266 /* Update the pointer pInA to point to the starting address of the next row */
emilmont 1:fdd22bb7aa52 267 i = i + numColsB;
emilmont 1:fdd22bb7aa52 268 pInA = pInA + numColsA;
emilmont 1:fdd22bb7aa52 269
emilmont 1:fdd22bb7aa52 270 /* Decrement the row loop counter */
emilmont 1:fdd22bb7aa52 271 row--;
emilmont 1:fdd22bb7aa52 272
emilmont 1:fdd22bb7aa52 273 } while(row > 0u);
emilmont 1:fdd22bb7aa52 274 /* Set status as ARM_MATH_SUCCESS */
emilmont 1:fdd22bb7aa52 275 status = ARM_MATH_SUCCESS;
emilmont 1:fdd22bb7aa52 276 }
emilmont 1:fdd22bb7aa52 277
emilmont 1:fdd22bb7aa52 278 /* Return to application */
emilmont 1:fdd22bb7aa52 279 return (status);
emilmont 1:fdd22bb7aa52 280 }
emilmont 1:fdd22bb7aa52 281
emilmont 1:fdd22bb7aa52 282 /**
emilmont 1:fdd22bb7aa52 283 * @} end of MatrixMult group
emilmont 1:fdd22bb7aa52 284 */