Fork of mbed-dsp. CMSIS-DSP library of supporting NEON
Dependents: mbed-os-example-cmsis_dsp_neon
Fork of mbed-dsp by
Information
Japanese version is available in lower part of this page.
このページの後半に日本語版が用意されています.
CMSIS-DSP of supporting NEON
What is this ?
A library for CMSIS-DSP of supporting NEON.
We supported the NEON to CMSIS-DSP Ver1.4.3(CMSIS V4.1) that ARM supplied, has achieved the processing speed improvement.
If you use the mbed-dsp library, you can use to replace this library.
CMSIS-DSP of supporting NEON is provied as a library.
Library Creation environment
CMSIS-DSP library of supporting NEON was created by the following environment.
- Compiler
ARMCC Version 5.03 - Compile option switch[C Compiler]
-DARM_MATH_MATRIX_CHECK -DARM_MATH_ROUNDING -O3 -Otime --cpu=Cortex-A9 --littleend --arm --apcs=/interwork --no_unaligned_access --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp --vectorize --asm
- Compile option switch[Assembler]
--cpreproc --cpu=Cortex-A9 --littleend --arm --apcs=/interwork --no_unaligned_access --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp
Effects of NEON support
In the data which passes to each function, large size will be expected more effective than small size.
Also if the data is a multiple of 16, effect will be expected in every function in the CMSIS-DSP.
NEON対応CMSIS-DSP
概要
NEON対応したCMSIS-DSPのライブラリです。
ARM社提供のCMSIS-DSP Ver1.4.3(CMSIS V4.1)をターゲットにNEON対応を行ない、処理速度向上を実現しております。
mbed-dspライブラリを使用している場合は、本ライブラリに置き換えて使用することができます。
NEON対応したCMSIS-DSPはライブラリで提供します。
ライブラリ作成環境
NEON対応CMSIS-DSPライブラリは、以下の環境で作成しています。
- コンパイラ
ARMCC Version 5.03 - コンパイルオプションスイッチ[C Compiler]
-DARM_MATH_MATRIX_CHECK -DARM_MATH_ROUNDING -O3 -Otime --cpu=Cortex-A9 --littleend --arm --apcs=/interwork --no_unaligned_access --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp --vectorize --asm
- コンパイルオプションスイッチ[Assembler]
--cpreproc --cpu=Cortex-A9 --littleend --arm --apcs=/interwork --no_unaligned_access --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp
NEON対応による効果について
CMSIS-DSP内の各関数へ渡すデータは、小さいサイズよりも大きいサイズの方が効果が見込めます。
また、16の倍数のデータであれば、CMSIS-DSP内のどの関数でも効果が見込めます。
cmsis_dsp/MatrixFunctions/arm_mat_mult_q15.c
- Committer:
- emilmont
- Date:
- 2013-05-30
- Revision:
- 2:da51fb522205
- Parent:
- 1:fdd22bb7aa52
- Child:
- 3:7a284390b0ce
File content as of revision 2:da51fb522205:
/* ---------------------------------------------------------------------- * Copyright (C) 2010 ARM Limited. All rights reserved. * * $Date: 15. February 2012 * $Revision: V1.1.0 * * Project: CMSIS DSP Library * Title: arm_mat_mult_q15.c * * Description: Q15 matrix multiplication. * * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0 * * Version 1.1.0 2012/02/15 * Updated with more optimizations, bug fixes and minor API changes. * * Version 1.0.10 2011/7/15 * Big Endian support added and Merged M0 and M3/M4 Source code. * * Version 1.0.3 2010/11/29 * Re-organized the CMSIS folders and updated documentation. * * Version 1.0.2 2010/11/11 * Documentation updated. * * Version 1.0.1 2010/10/05 * Production release and review comments incorporated. * * Version 1.0.0 2010/09/20 * Production release and review comments incorporated. * * Version 0.0.5 2010/04/26 * incorporated review comments and updated with latest CMSIS layer * * Version 0.0.3 2010/03/10 * Initial version * -------------------------------------------------------------------- */ #include "arm_math.h" /** * @ingroup groupMatrix */ /** * @addtogroup MatrixMult * @{ */ /** * @brief Q15 matrix multiplication * @param[in] *pSrcA points to the first input matrix structure * @param[in] *pSrcB points to the second input matrix structure * @param[out] *pDst points to output matrix structure * @param[in] *pState points to the array for storing intermediate results * @return The function returns either * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking. * * @details * <b>Scaling and Overflow Behavior:</b> * * \par * The function is implemented using a 64-bit internal accumulator. The inputs to the * multiplications are in 1.15 format and multiplications yield a 2.30 result. * The 2.30 intermediate * results are accumulated in a 64-bit accumulator in 34.30 format. This approach * provides 33 guard bits and there is no risk of overflow. The 34.30 result is then * truncated to 34.15 format by discarding the low 15 bits and then saturated to * 1.15 format. * * \par * Refer to <code>arm_mat_mult_fast_q15()</code> for a faster but less precise version of this function for Cortex-M3 and Cortex-M4. * */ arm_status arm_mat_mult_q15( const arm_matrix_instance_q15 * pSrcA, const arm_matrix_instance_q15 * pSrcB, arm_matrix_instance_q15 * pDst, q15_t * pState) { q63_t sum; /* accumulator */ #ifndef ARM_MATH_CM0 /* Run the below code for Cortex-M4 and Cortex-M3 */ q15_t *pSrcBT = pState; /* input data matrix pointer for transpose */ q15_t *pInA = pSrcA->pData; /* input data matrix pointer A of Q15 type */ q15_t *pInB = pSrcB->pData; /* input data matrix pointer B of Q15 type */ q15_t *px; /* Temporary output data matrix pointer */ 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 */ uint16_t numRowsB = pSrcB->numRows; /* number of rows of input matrix A */ uint16_t col, i = 0u, row = numRowsB, colCnt; /* loop counters */ arm_status status; /* status of matrix multiplication */ #ifndef UNALIGNED_SUPPORT_DISABLE q31_t in; /* Temporary variable to hold the input value */ q31_t pSourceA1, pSourceB1, pSourceA2, pSourceB2; #else q15_t in; /* Temporary variable to hold the input value */ q15_t inA1, inB1, inA2, inB2; #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */ #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 */ { /* Matrix transpose */ do { /* Apply loop unrolling and exchange the columns with row elements */ col = numColsB >> 2; /* The pointer px is set to starting address of the column being processed */ px = pSrcBT + i; /* First part of the processing with loop unrolling. Compute 4 outputs at a time. ** a second loop below computes the remaining 1 to 3 samples. */ while(col > 0u) { #ifndef UNALIGNED_SUPPORT_DISABLE /* Read two elements from the row */ in = *__SIMD32(pInB)++; /* Unpack and store one element in the destination */ #ifndef ARM_MATH_BIG_ENDIAN *px = (q15_t) in; #else *px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16); #endif /* #ifndef ARM_MATH_BIG_ENDIAN */ /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; /* Unpack and store the second element in the destination */ #ifndef ARM_MATH_BIG_ENDIAN *px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16); #else *px = (q15_t) in; #endif /* #ifndef ARM_MATH_BIG_ENDIAN */ /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; /* Read two elements from the row */ in = *__SIMD32(pInB)++; /* Unpack and store one element in the destination */ #ifndef ARM_MATH_BIG_ENDIAN *px = (q15_t) in; #else *px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16); #endif /* #ifndef ARM_MATH_BIG_ENDIAN */ /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; /* Unpack and store the second element in the destination */ #ifndef ARM_MATH_BIG_ENDIAN *px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16); #else *px = (q15_t) in; #endif /* #ifndef ARM_MATH_BIG_ENDIAN */ /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; #else /* Read one element from the row */ in = *pInB++; /* Store one element in the destination */ *px = in; /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; /* Read one element from the row */ in = *pInB++; /* Store one element in the destination */ *px = in; /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; /* Read one element from the row */ in = *pInB++; /* Store one element in the destination */ *px = in; /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; /* Read one element from the row */ in = *pInB++; /* Store one element in the destination */ *px = in; /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */ /* Decrement the column loop counter */ col--; } /* If the columns of pSrcB is not a multiple of 4, compute any remaining output samples here. ** No loop unrolling is used. */ col = numColsB % 0x4u; while(col > 0u) { /* Read and store the input element in the destination */ *px = *pInB++; /* Update the pointer px to point to the next row of the transposed matrix */ px += numRowsB; /* Decrement the column loop counter */ col--; } i++; /* Decrement the row loop counter */ row--; } while(row > 0u); /* Reset the variables for the usage in the following multiplication process */ row = numRowsA; i = 0u; px = pDst->pData; /* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */ /* row loop */ do { /* 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 transposed pSrcB data */ pInB = pSrcBT; /* column loop */ do { /* Set the variable sum, that acts as accumulator, to zero */ sum = 0; /* Apply loop unrolling and compute 2 MACs simultaneously. */ colCnt = numColsA >> 2; /* Initiate the pointer pIn1 to point to the starting address of the column being processed */ pInA = pSrcA->pData + i; /* 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) */ #ifndef UNALIGNED_SUPPORT_DISABLE /* read real and imag values from pSrcA and pSrcB buffer */ pSourceA1 = *__SIMD32(pInA)++; pSourceB1 = *__SIMD32(pInB)++; pSourceA2 = *__SIMD32(pInA)++; pSourceB2 = *__SIMD32(pInB)++; /* Multiply and Accumlates */ sum = __SMLALD(pSourceA1, pSourceB1, sum); sum = __SMLALD(pSourceA2, pSourceB2, sum); #else /* read real and imag values from pSrcA and pSrcB buffer */ inA1 = *pInA++; inB1 = *pInB++; inA2 = *pInA++; /* Multiply and Accumlates */ sum += inA1 * inB1; inB2 = *pInB++; inA1 = *pInA++; inB1 = *pInB++; /* Multiply and Accumlates */ sum += inA2 * inB2; inA2 = *pInA++; inB2 = *pInB++; /* Multiply and Accumlates */ sum += inA1 * inB1; sum += inA2 * inB2; #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */ /* Decrement the loop counter */ colCnt--; } /* process remaining column samples */ colCnt = numColsA & 3u; 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 += *pInA++ * *pInB++; /* Decrement the loop counter */ colCnt--; } /* Saturate and store the result in the destination buffer */ *px = (q15_t) (__SSAT((sum >> 15), 16)); px++; /* Decrement the column loop counter */ col--; } while(col > 0u); i = i + numColsA; /* Decrement the row loop counter */ row--; } while(row > 0u); #else /* Run the below code for Cortex-M0 */ q15_t *pIn1 = pSrcA->pData; /* input data matrix pointer A */ q15_t *pIn2 = pSrcB->pData; /* input data matrix pointer B */ q15_t *pInA = pSrcA->pData; /* input data matrix pointer A of Q15 type */ q15_t *pInB = pSrcB->pData; /* input data matrix pointer B of Q15 type */ q15_t *pOut = pDst->pData; /* output data matrix pointer */ q15_t *px; /* Temporary output data matrix pointer */ uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */ uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */ uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */ 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 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; /* column loop */ do { /* Set the variable sum, that acts as accumulator, to zero */ sum = 0; /* Initiate the pointer pIn1 to point to the starting address of pSrcA */ pIn1 = pInA; /* Matrix A columns number of MAC operations are to be performed */ colCnt = numColsA; /* 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) */ /* Perform the multiply-accumulates */ sum += (q31_t) * pIn1++ * *pIn2; pIn2 += numColsB; /* Decrement the loop counter */ colCnt--; } /* Convert the result from 34.30 to 1.15 format and store the saturated value in destination buffer */ /* Saturate and store the result in the destination buffer */ *px++ = (q15_t) __SSAT((sum >> 15), 16); /* 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); /* Update the pointer pSrcA 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); #endif /* #ifndef ARM_MATH_CM0 */ /* set status as ARM_MATH_SUCCESS */ status = ARM_MATH_SUCCESS; } /* Return to application */ return (status); } /** * @} end of MatrixMult group */