Rev 2 | Details | Compare with Previous | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 2 | mjames | 1 | /* ---------------------------------------------------------------------- |
| 2 | * Project: CMSIS DSP Library |
||
| 3 | * Title: arm_mat_scale_q31.c |
||
| 4 | * Description: Multiplies a Q31 matrix by a scalar |
||
| 5 | * |
||
| 6 | * $Date: 27. January 2017 |
||
| 7 | * $Revision: V.1.5.1 |
||
| 8 | * |
||
| 9 | * Target Processor: Cortex-M cores |
||
| 10 | * -------------------------------------------------------------------- */ |
||
| 11 | /* |
||
| 12 | * Copyright (C) 2010-2017 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 "arm_math.h" |
||
| 30 | |||
| 31 | /** |
||
| 32 | * @ingroup groupMatrix |
||
| 33 | */ |
||
| 34 | |||
| 35 | /** |
||
| 36 | * @addtogroup MatrixScale |
||
| 37 | * @{ |
||
| 38 | */ |
||
| 39 | |||
| 40 | /** |
||
| 41 | * @brief Q31 matrix scaling. |
||
| 42 | * @param[in] *pSrc points to input matrix |
||
| 43 | * @param[in] scaleFract fractional portion of the scale factor |
||
| 44 | * @param[in] shift number of bits to shift the result by |
||
| 45 | * @param[out] *pDst points to output matrix structure |
||
| 46 | * @return The function returns either |
||
| 47 | * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking. |
||
| 48 | * |
||
| 49 | * @details |
||
| 50 | * <b>Scaling and Overflow Behavior:</b> |
||
| 51 | * \par |
||
| 52 | * The input data <code>*pSrc</code> and <code>scaleFract</code> are in 1.31 format. |
||
| 53 | * These are multiplied to yield a 2.62 intermediate result and this is shifted with saturation to 1.31 format. |
||
| 54 | */ |
||
| 55 | |||
| 56 | arm_status arm_mat_scale_q31( |
||
| 57 | const arm_matrix_instance_q31 * pSrc, |
||
| 58 | q31_t scaleFract, |
||
| 59 | int32_t shift, |
||
| 60 | arm_matrix_instance_q31 * pDst) |
||
| 61 | { |
||
| 62 | q31_t *pIn = pSrc->pData; /* input data matrix pointer */ |
||
| 63 | q31_t *pOut = pDst->pData; /* output data matrix pointer */ |
||
| 64 | uint32_t numSamples; /* total number of elements in the matrix */ |
||
| 65 | int32_t totShift = shift + 1; /* shift to apply after scaling */ |
||
| 66 | uint32_t blkCnt; /* loop counters */ |
||
| 67 | arm_status status; /* status of matrix scaling */ |
||
| 68 | q31_t in1, in2, out1; /* temporary variabels */ |
||
| 69 | |||
| 70 | #if defined (ARM_MATH_DSP) |
||
| 71 | |||
| 72 | q31_t in3, in4, out2, out3, out4; /* temporary variables */ |
||
| 73 | |||
| 74 | #endif // #ifndef ARM_MAT_CM0 |
||
| 75 | |||
| 76 | #ifdef ARM_MATH_MATRIX_CHECK |
||
| 77 | /* Check for matrix mismatch */ |
||
| 78 | if ((pSrc->numRows != pDst->numRows) || (pSrc->numCols != pDst->numCols)) |
||
| 79 | { |
||
| 80 | /* Set status as ARM_MATH_SIZE_MISMATCH */ |
||
| 81 | status = ARM_MATH_SIZE_MISMATCH; |
||
| 82 | } |
||
| 83 | else |
||
| 84 | #endif // #ifdef ARM_MATH_MATRIX_CHECK |
||
| 85 | { |
||
| 86 | /* Total number of samples in the input matrix */ |
||
| 87 | numSamples = (uint32_t) pSrc->numRows * pSrc->numCols; |
||
| 88 | |||
| 89 | #if defined (ARM_MATH_DSP) |
||
| 90 | |||
| 91 | /* Run the below code for Cortex-M4 and Cortex-M3 */ |
||
| 92 | |||
| 93 | /* Loop Unrolling */ |
||
| 94 | blkCnt = numSamples >> 2U; |
||
| 95 | |||
| 96 | /* First part of the processing with loop unrolling. Compute 4 outputs at a time. |
||
| 97 | ** a second loop below computes the remaining 1 to 3 samples. */ |
||
| 98 | while (blkCnt > 0U) |
||
| 99 | { |
||
| 100 | /* C(m,n) = A(m,n) * k */ |
||
| 101 | /* Read values from input */ |
||
| 102 | in1 = *pIn; |
||
| 103 | in2 = *(pIn + 1); |
||
| 104 | in3 = *(pIn + 2); |
||
| 105 | in4 = *(pIn + 3); |
||
| 106 | |||
| 107 | /* multiply input with scaler value */ |
||
| 108 | in1 = ((q63_t) in1 * scaleFract) >> 32; |
||
| 109 | in2 = ((q63_t) in2 * scaleFract) >> 32; |
||
| 110 | in3 = ((q63_t) in3 * scaleFract) >> 32; |
||
| 111 | in4 = ((q63_t) in4 * scaleFract) >> 32; |
||
| 112 | |||
| 113 | /* apply shifting */ |
||
| 114 | out1 = in1 << totShift; |
||
| 115 | out2 = in2 << totShift; |
||
| 116 | |||
| 117 | /* saturate the results. */ |
||
| 118 | if (in1 != (out1 >> totShift)) |
||
| 119 | out1 = 0x7FFFFFFF ^ (in1 >> 31); |
||
| 120 | |||
| 121 | if (in2 != (out2 >> totShift)) |
||
| 122 | out2 = 0x7FFFFFFF ^ (in2 >> 31); |
||
| 123 | |||
| 124 | out3 = in3 << totShift; |
||
| 125 | out4 = in4 << totShift; |
||
| 126 | |||
| 127 | *pOut = out1; |
||
| 128 | *(pOut + 1) = out2; |
||
| 129 | |||
| 130 | if (in3 != (out3 >> totShift)) |
||
| 131 | out3 = 0x7FFFFFFF ^ (in3 >> 31); |
||
| 132 | |||
| 133 | if (in4 != (out4 >> totShift)) |
||
| 134 | out4 = 0x7FFFFFFF ^ (in4 >> 31); |
||
| 135 | |||
| 136 | |||
| 137 | *(pOut + 2) = out3; |
||
| 138 | *(pOut + 3) = out4; |
||
| 139 | |||
| 140 | /* update pointers to process next sampels */ |
||
| 141 | pIn += 4U; |
||
| 142 | pOut += 4U; |
||
| 143 | |||
| 144 | |||
| 145 | /* Decrement the numSamples loop counter */ |
||
| 146 | blkCnt--; |
||
| 147 | } |
||
| 148 | |||
| 149 | /* If the numSamples is not a multiple of 4, compute any remaining output samples here. |
||
| 150 | ** No loop unrolling is used. */ |
||
| 151 | blkCnt = numSamples % 0x4U; |
||
| 152 | |||
| 153 | #else |
||
| 154 | |||
| 155 | /* Run the below code for Cortex-M0 */ |
||
| 156 | |||
| 157 | /* Initialize blkCnt with number of samples */ |
||
| 158 | blkCnt = numSamples; |
||
| 159 | |||
| 160 | #endif /* #if defined (ARM_MATH_DSP) */ |
||
| 161 | |||
| 162 | while (blkCnt > 0U) |
||
| 163 | { |
||
| 164 | /* C(m,n) = A(m,n) * k */ |
||
| 165 | /* Scale, saturate and then store the results in the destination buffer. */ |
||
| 166 | in1 = *pIn++; |
||
| 167 | |||
| 168 | in2 = ((q63_t) in1 * scaleFract) >> 32; |
||
| 169 | |||
| 170 | out1 = in2 << totShift; |
||
| 171 | |||
| 172 | if (in2 != (out1 >> totShift)) |
||
| 173 | out1 = 0x7FFFFFFF ^ (in2 >> 31); |
||
| 174 | |||
| 175 | *pOut++ = out1; |
||
| 176 | |||
| 177 | /* Decrement the numSamples loop counter */ |
||
| 178 | blkCnt--; |
||
| 179 | } |
||
| 180 | |||
| 181 | /* Set status as ARM_MATH_SUCCESS */ |
||
| 182 | status = ARM_MATH_SUCCESS; |
||
| 183 | } |
||
| 184 | |||
| 185 | /* Return to application */ |
||
| 186 | return (status); |
||
| 187 | } |
||
| 188 | |||
| 189 | /** |
||
| 190 | * @} end of MatrixScale group |
||
| 191 | */ |