arm_scale_q15.c
Go to the documentation of this file.
00001 /* ---------------------------------------------------------------------- 00002 * Copyright (C) 2011 ARM Limited. All rights reserved. 00003 * 00004 * $Date: 15. December 2011 00005 * $Revision: V2.0.0 00006 * 00007 * Project: Cortex-R DSP Library 00008 * Title: arm_scale_q15.c 00009 * 00010 * Description: Multiplies a Q15 vector by a scalar. 00011 * 00012 * Target Processor: Cortex-R4/R5 00013 * 00014 * Version 1.0.0 2011/03/08 00015 * Alpha release. 00016 * 00017 * Version 1.0.1 2011/09/30 00018 * Beta release. 00019 * 00020 * Version 2.0.0 2011/12/15 00021 * Final release. 00022 * 00023 * -------------------------------------------------------------------- */ 00024 #include "arm_math.h" 00025 00055 void arm_scale_q15( 00056 q15_t * pSrc, 00057 q15_t scaleFract, 00058 int8_t shift, 00059 q15_t * pDst, 00060 uint32_t blockSize) 00061 { 00062 int8_t kShift = 15 - shift; /* shift to apply after scaling */ 00063 uint32_t blkCnt; /* loop counter */ 00064 q15_t in1, in2, in3, in4; 00065 q31_t inA1, inA2; /* Temporary variables */ 00066 q31_t out1, out2, out3, out4; 00067 00068 00069 /*loop Unrolling */ 00070 blkCnt = blockSize >> 2u; 00071 00072 /* First part of the processing with loop unrolling. Compute 4 outputs at a time. 00073 ** a second loop below computes the remaining 1 to 3 samples. */ 00074 while(blkCnt > 0u) 00075 { 00076 /* Reading 2 inputs from memory */ 00077 inA1 = *__SIMD32(pSrc)++; 00078 inA2 = *__SIMD32(pSrc)++; 00079 00080 /* C = A * scale */ 00081 /* Scale the inputs and then store the 2 results in the destination buffer 00082 * in single cycle by packing the outputs */ 00083 out1 = (q31_t)((q15_t)(inA1 >> 16) * scaleFract); 00084 out2 = (q31_t)((q15_t)inA1 * scaleFract); 00085 out3 = (q31_t)((q15_t)(inA2 >> 16) * scaleFract); 00086 out4 = (q31_t)((q15_t)inA2 * scaleFract); 00087 00088 /* apply shifting */ 00089 out1 = out1 >> kShift; 00090 out2 = out2 >> kShift; 00091 out3 = out3 >> kShift; 00092 out4 = out4 >> kShift; 00093 00094 /* saturate the output */ 00095 #ifdef CCS 00096 00097 in1 = (q15_t)(__SSATA(out1, 0, 16)); 00098 in2 = (q15_t)(__SSATA(out2, 0, 16)); 00099 in3 = (q15_t)(__SSATA(out3, 0, 16)); 00100 in4 = (q15_t)(__SSATA(out4, 0, 16)); 00101 00102 #else 00103 00104 in1 = (q15_t)(__SSAT(out1, 16)); 00105 in2 = (q15_t)(__SSAT(out2, 16)); 00106 in3 = (q15_t)(__SSAT(out3, 16)); 00107 in4 = (q15_t)(__SSAT(out4, 16)); 00108 00109 #endif 00110 00111 /* store the result to destination */ 00112 *__SIMD32(pDst)++ = __PKHBT(in2, in1,16); 00113 *__SIMD32(pDst)++ = __PKHBT(in4, in3,16); 00114 00115 /* Decrement the loop counter */ 00116 blkCnt--; 00117 } 00118 00119 /* If the blockSize is not a multiple of 4, compute any remaining output samples here. 00120 ** No loop unrolling is used. */ 00121 blkCnt = blockSize % 0x4u; 00122 00123 while(blkCnt > 0u) 00124 { 00125 /* C = A * scale */ 00126 /* Scale the input and then store the result in the destination buffer. */ 00127 #ifdef CCS 00128 *pDst++ = (q15_t) (__SSATA(((*pSrc++) * scaleFract) >> kShift, 0, 16)); 00129 #else 00130 *pDst++ = (q15_t) (__SSAT(((*pSrc++) * scaleFract) >> kShift, 16)); 00131 #endif 00132 /* Decrement the loop counter */ 00133 blkCnt--; 00134 } 00135 } 00136
