From 96d6da4e252b06dcfdc041e7df23e86161c33007 Mon Sep 17 00:00:00 2001 From: rihab kouki Date: Tue, 28 Jul 2020 11:24:49 +0100 Subject: Official ARM version: v5.6.0 --- DSP/Source/MatrixFunctions/arm_mat_scale_f32.c | 166 ++++++++++++++++--------- 1 file changed, 109 insertions(+), 57 deletions(-) (limited to 'DSP/Source/MatrixFunctions/arm_mat_scale_f32.c') diff --git a/DSP/Source/MatrixFunctions/arm_mat_scale_f32.c b/DSP/Source/MatrixFunctions/arm_mat_scale_f32.c index dbc385a..a0097b1 100644 --- a/DSP/Source/MatrixFunctions/arm_mat_scale_f32.c +++ b/DSP/Source/MatrixFunctions/arm_mat_scale_f32.c @@ -3,13 +3,13 @@ * Title: arm_mat_scale_f32.c * Description: Multiplies a floating-point matrix by a scalar * - * $Date: 27. January 2017 - * $Revision: V.1.5.1 + * $Date: 18. March 2019 + * $Revision: V1.6.0 * * Target Processor: Cortex-M cores * -------------------------------------------------------------------- */ /* - * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved. + * Copyright (C) 2010-2019 ARM Limited or its affiliates. All rights reserved. * * SPDX-License-Identifier: Apache-2.0 * @@ -29,42 +29,42 @@ #include "arm_math.h" /** - * @ingroup groupMatrix + @ingroup groupMatrix */ /** - * @defgroup MatrixScale Matrix Scale - * - * Multiplies a matrix by a scalar. This is accomplished by multiplying each element in the - * matrix by the scalar. For example: - * \image html MatrixScale.gif "Matrix Scaling of a 3 x 3 matrix" - * - * The function checks to make sure that the input and output matrices are of the same size. - * - * In the fixed-point Q15 and Q31 functions, scale is represented by - * a fractional multiplication scaleFract and an arithmetic shift shift. - * The shift allows the gain of the scaling operation to exceed 1.0. - * The overall scale factor applied to the fixed-point data is - * 
- *     scale = scaleFract * 2^shift.
- *
+ @defgroup MatrixScale Matrix Scale + + Multiplies a matrix by a scalar. This is accomplished by multiplying each element in the + matrix by the scalar. For example: + \image html MatrixScale.gif "Matrix Scaling of a 3 x 3 matrix" + + The function checks to make sure that the input and output matrices are of the same size. + + In the fixed-point Q15 and Q31 functions, scale is represented by + a fractional multiplication scaleFract and an arithmetic shift shift. + The shift allows the gain of the scaling operation to exceed 1.0. + The overall scale factor applied to the fixed-point data is + 
+      scale = scaleFract * 2^shift.
+
*/ /** - * @addtogroup MatrixScale - * @{ + @addtogroup MatrixScale + @{ */ /** - * @brief Floating-point matrix scaling. - * @param[in] *pSrc points to input matrix structure - * @param[in] scale scale factor to be applied - * @param[out] *pDst points to output matrix structure - * @return The function returns either ARM_MATH_SIZE_MISMATCH - * or ARM_MATH_SUCCESS based on the outcome of size checking. - * + @brief Floating-point matrix scaling. + @param[in] pSrc points to input matrix + @param[in] scale scale factor to be applied + @param[out] pDst points to output matrix structure + @return execution status + - \ref ARM_MATH_SUCCESS : Operation successful + - \ref ARM_MATH_SIZE_MISMATCH : Matrix size check failed */ - +#if defined(ARM_MATH_NEON_EXPERIMENTAL) arm_status arm_mat_scale_f32( const arm_matrix_instance_f32 * pSrc, float32_t scale, @@ -76,12 +76,10 @@ arm_status arm_mat_scale_f32( uint32_t blkCnt; /* loop counters */ arm_status status; /* status of matrix scaling */ -#if defined (ARM_MATH_DSP) float32_t in1, in2, in3, in4; /* temporary variables */ float32_t out1, out2, out3, out4; /* temporary variables */ -#endif // #if defined (ARM_MATH_DSP) #ifdef ARM_MATH_MATRIX_CHECK /* Check for matrix mismatch condition */ @@ -93,37 +91,23 @@ arm_status arm_mat_scale_f32( else #endif /* #ifdef ARM_MATH_MATRIX_CHECK */ { + float32x4_t vec1; + float32x4_t res; + /* Total number of samples in the input matrix */ numSamples = (uint32_t) pSrc->numRows * pSrc->numCols; -#if defined (ARM_MATH_DSP) - - /* Run the below code for Cortex-M4 and Cortex-M3 */ - - /* Loop Unrolling */ blkCnt = numSamples >> 2; - /* First part of the processing with loop unrolling. Compute 4 outputs at a time. + /* Compute 4 outputs at a time. ** a second loop below computes the remaining 1 to 3 samples. */ while (blkCnt > 0U) { /* C(m,n) = A(m,n) * scale */ /* Scaling and results are stored in the destination buffer. */ - in1 = pIn[0]; - in2 = pIn[1]; - in3 = pIn[2]; - in4 = pIn[3]; - - out1 = in1 * scale; - out2 = in2 * scale; - out3 = in3 * scale; - out4 = in4 * scale; - - - pOut[0] = out1; - pOut[1] = out2; - pOut[2] = out3; - pOut[3] = out4; + vec1 = vld1q_f32(pIn); + res = vmulq_f32(vec1, vdupq_n_f32(scale)); + vst1q_f32(pOut, res); /* update pointers to process next sampels */ pIn += 4U; @@ -137,22 +121,89 @@ arm_status arm_mat_scale_f32( ** No loop unrolling is used. */ blkCnt = numSamples % 0x4U; + while (blkCnt > 0U) + { + /* C(m,n) = A(m,n) * scale */ + /* The results are stored in the destination buffer. */ + *pOut++ = (*pIn++) * scale; + + /* Decrement the loop counter */ + blkCnt--; + } + + /* Set status as ARM_MATH_SUCCESS */ + status = ARM_MATH_SUCCESS; + } + + /* Return to application */ + return (status); +} #else +arm_status arm_mat_scale_f32( + const arm_matrix_instance_f32 * pSrc, + float32_t scale, + arm_matrix_instance_f32 * pDst) +{ + float32_t *pIn = pSrc->pData; /* Input data matrix pointer */ + float32_t *pOut = pDst->pData; /* Output data matrix pointer */ + uint32_t numSamples; /* Total number of elements in the matrix */ + uint32_t blkCnt; /* Loop counters */ + arm_status status; /* Status of matrix scaling */ - /* Run the below code for Cortex-M0 */ +#ifdef ARM_MATH_MATRIX_CHECK + + /* Check for matrix mismatch condition */ + if ((pSrc->numRows != pDst->numRows) || + (pSrc->numCols != pDst->numCols) ) + { + /* Set status as ARM_MATH_SIZE_MISMATCH */ + status = ARM_MATH_SIZE_MISMATCH; + } + else + +#endif /* #ifdef ARM_MATH_MATRIX_CHECK */ + + { + /* Total number of samples in input matrix */ + numSamples = (uint32_t) pSrc->numRows * pSrc->numCols; + +#if defined (ARM_MATH_LOOPUNROLL) + + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = numSamples >> 2U; + + while (blkCnt > 0U) + { + /* C(m,n) = A(m,n) * scale */ + + /* Scale and store result in destination buffer. */ + *pOut++ = (*pIn++) * scale; + *pOut++ = (*pIn++) * scale; + *pOut++ = (*pIn++) * scale; + *pOut++ = (*pIn++) * scale; + + /* Decrement loop counter */ + blkCnt--; + } + + /* Loop unrolling: Compute remaining outputs */ + blkCnt = numSamples % 0x4U; + +#else /* Initialize blkCnt with number of samples */ blkCnt = numSamples; -#endif /* #if defined (ARM_MATH_DSP) */ +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ while (blkCnt > 0U) { /* C(m,n) = A(m,n) * scale */ - /* The results are stored in the destination buffer. */ + + /* Scale and store result in destination buffer. */ *pOut++ = (*pIn++) * scale; - /* Decrement the loop counter */ + /* Decrement loop counter */ blkCnt--; } @@ -163,7 +214,8 @@ arm_status arm_mat_scale_f32( /* Return to application */ return (status); } +#endif /* #if defined(ARM_MATH_NEON) */ /** - * @} end of MatrixScale group + @} end of MatrixScale group */ -- cgit