123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274 |
- /* ----------------------------------------------------------------------
- * Project: CMSIS DSP Library
- * Title: arm_mat_mult_f32.c
- * Description: Floating-point matrix multiplication
- *
- * $Date: 27. January 2017
- * $Revision: V.1.5.1
- *
- * Target Processor: Cortex-M cores
- * -------------------------------------------------------------------- */
- /*
- * Copyright (C) 2010-2017 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 groupMatrix
- */
- /**
- * @defgroup MatrixMult Matrix Multiplication
- *
- * Multiplies two matrices.
- *
- * \image html MatrixMultiplication.gif "Multiplication of two 3 x 3 matrices"
- * Matrix multiplication is only defined if the number of columns of the
- * first matrix equals the number of rows of the second matrix.
- * Multiplying an <code>M x N</code> matrix with an <code>N x P</code> matrix results
- * in an <code>M x P</code> matrix.
- * When matrix size checking is enabled, the functions check: (1) that the inner dimensions of
- * <code>pSrcA</code> and <code>pSrcB</code> are equal; and (2) that the size of the output
- * matrix equals the outer dimensions of <code>pSrcA</code> and <code>pSrcB</code>.
- */
- /**
- * @addtogroup MatrixMult
- * @{
- */
- /**
- * @brief Floating-point matrix multiplication.
- * @param[in] *pSrcA points to the first input matrix structure
- * @param[in] *pSrcB points to the second input matrix structure
- * @param[out] *pDst points to output matrix structure
- * @return The function returns either
- * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
- */
- arm_status arm_mat_mult_f32(
- const arm_matrix_instance_f32 * pSrcA,
- const arm_matrix_instance_f32 * pSrcB,
- arm_matrix_instance_f32 * pDst)
- {
- float32_t *pIn1 = pSrcA->pData; /* input data matrix pointer A */
- float32_t *pIn2 = pSrcB->pData; /* input data matrix pointer B */
- float32_t *pInA = pSrcA->pData; /* input data matrix pointer A */
- float32_t *pOut = pDst->pData; /* output data matrix pointer */
- float32_t *px; /* Temporary output data matrix pointer */
- float32_t sum; /* Accumulator */
- uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */
- uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */
- uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */
- #if defined (ARM_MATH_DSP)
- /* Run the below code for Cortex-M4 and Cortex-M3 */
- float32_t in1, in2, in3, in4;
- uint16_t col, i = 0U, j, row = numRowsA, colCnt; /* loop counters */
- arm_status status; /* status of matrix multiplication */
- #ifdef ARM_MATH_MATRIX_CHECK
- /* Check for matrix mismatch condition */
- if ((pSrcA->numCols != pSrcB->numRows) ||
- (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
- {
- /* Set status as ARM_MATH_SIZE_MISMATCH */
- status = ARM_MATH_SIZE_MISMATCH;
- }
- else
- #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
- {
- /* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */
- /* row loop */
- do
- {
- /* Output pointer is set to starting address of the row being processed */
- px = pOut + i;
- /* For every row wise process, the column loop counter is to be initiated */
- col = numColsB;
- /* For every row wise process, the pIn2 pointer is set
- ** to the starting address of the pSrcB data */
- pIn2 = pSrcB->pData;
- j = 0U;
- /* column loop */
- do
- {
- /* Set the variable sum, that acts as accumulator, to zero */
- sum = 0.0f;
- /* Initiate the pointer pIn1 to point to the starting address of the column being processed */
- pIn1 = pInA;
- /* Apply loop unrolling and compute 4 MACs simultaneously. */
- colCnt = numColsA >> 2U;
- /* matrix multiplication */
- while (colCnt > 0U)
- {
- /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
- in3 = *pIn2;
- pIn2 += numColsB;
- in1 = pIn1[0];
- in2 = pIn1[1];
- sum += in1 * in3;
- in4 = *pIn2;
- pIn2 += numColsB;
- sum += in2 * in4;
- in3 = *pIn2;
- pIn2 += numColsB;
- in1 = pIn1[2];
- in2 = pIn1[3];
- sum += in1 * in3;
- in4 = *pIn2;
- pIn2 += numColsB;
- sum += in2 * in4;
- pIn1 += 4U;
- /* Decrement the loop count */
- colCnt--;
- }
- /* If the columns of pSrcA is not a multiple of 4, compute any remaining MACs here.
- ** No loop unrolling is used. */
- colCnt = numColsA % 0x4U;
- while (colCnt > 0U)
- {
- /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
- sum += *pIn1++ * (*pIn2);
- pIn2 += numColsB;
- /* Decrement the loop counter */
- colCnt--;
- }
- /* Store the result in the destination buffer */
- *px++ = sum;
- /* Update the pointer pIn2 to point to the starting address of the next column */
- j++;
- pIn2 = pSrcB->pData + j;
- /* Decrement the column loop counter */
- col--;
- } while (col > 0U);
- #else
- /* Run the below code for Cortex-M0 */
- float32_t *pInB = pSrcB->pData; /* input data matrix pointer B */
- uint16_t col, i = 0U, row = numRowsA, colCnt; /* loop counters */
- arm_status status; /* status of matrix multiplication */
- #ifdef ARM_MATH_MATRIX_CHECK
- /* Check for matrix mismatch condition */
- if ((pSrcA->numCols != pSrcB->numRows) ||
- (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
- {
- /* Set status as ARM_MATH_SIZE_MISMATCH */
- status = ARM_MATH_SIZE_MISMATCH;
- }
- else
- #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
- {
- /* The following loop performs the dot-product of each row in pInA with each column in pInB */
- /* row loop */
- do
- {
- /* Output pointer is set to starting address of the row being processed */
- px = pOut + i;
- /* For every row wise process, the column loop counter is to be initiated */
- col = numColsB;
- /* For every row wise process, the pIn2 pointer is set
- ** to the starting address of the pSrcB data */
- pIn2 = pSrcB->pData;
- /* column loop */
- do
- {
- /* Set the variable sum, that acts as accumulator, to zero */
- sum = 0.0f;
- /* Initialize the pointer pIn1 to point to the starting address of the row being processed */
- pIn1 = pInA;
- /* Matrix A columns number of MAC operations are to be performed */
- colCnt = numColsA;
- while (colCnt > 0U)
- {
- /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
- sum += *pIn1++ * (*pIn2);
- pIn2 += numColsB;
- /* Decrement the loop counter */
- colCnt--;
- }
- /* Store the result in the destination buffer */
- *px++ = sum;
- /* Decrement the column loop counter */
- col--;
- /* Update the pointer pIn2 to point to the starting address of the next column */
- pIn2 = pInB + (numColsB - col);
- } while (col > 0U);
- #endif /* #if defined (ARM_MATH_DSP) */
- /* Update the pointer pInA to point to the starting address of the next row */
- i = i + numColsB;
- pInA = pInA + numColsA;
- /* Decrement the row loop counter */
- row--;
- } while (row > 0U);
- /* Set status as ARM_MATH_SUCCESS */
- status = ARM_MATH_SUCCESS;
- }
- /* Return to application */
- return (status);
- }
- /**
- * @} end of MatrixMult group
- */
|