arm_add_q7.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_add_q7.c 00009 * 00010 * Description: Q7 vector addition. 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 00053 void arm_add_q7( 00054 q7_t * pSrcA, 00055 q7_t * pSrcB, 00056 q7_t * pDst, 00057 uint32_t blockSize) 00058 { 00059 uint32_t blkCnt; /* loop counter */ 00060 q31_t inA1, inA2, inB2, inB1; /* temporary input variables */ 00061 q31_t out1, out2, out3, out4; /* temporary output variables */ 00062 00063 00064 /*loop Unrolling */ 00065 blkCnt = blockSize >> 4u; 00066 00067 /* First part of the processing with loop unrolling. Compute 16 outputs at a time. 00068 ** a second loop below computes the remaining 1 to 15 samples. */ 00069 while(blkCnt) 00070 { 00071 /* C = A + B */ 00072 /* Add and then store the results in the destination buffer. */ 00073 /* read four input samples at a time from sourceA*/ 00074 inA1 = _SIMD32_OFFSET(pSrcA); 00075 /* read four input samples at a time from sourceB*/ 00076 inB1 = _SIMD32_OFFSET(pSrcB); 00077 /* read four input samples at a time from sourceA*/ 00078 inA2 = _SIMD32_OFFSET(pSrcA + 4); 00079 00080 /* add, saturate and store four samples to destination at a time*/ 00081 out1 = __QADD8(inA1, inB1); 00082 00083 /* read four input samples at a time */ 00084 inB2 = _SIMD32_OFFSET(pSrcB + 4); 00085 00086 /* store four samples at a time to destination */ 00087 _SIMD32_OFFSET(pDst) = out1; 00088 00089 /* add, saturate and store four samples to destination at a time*/ 00090 out2 = __QADD8(inA2, inB2); 00091 00092 /* read four input samples at a time from sourceA*/ 00093 inA1 = _SIMD32_OFFSET(pSrcA + 8); 00094 /* read four input samples at a time from sourceA*/ 00095 inB1 = _SIMD32_OFFSET(pSrcB + 8); 00096 /* read four input samples at a time from sourceA*/ 00097 inA2 = _SIMD32_OFFSET(pSrcA + 12); 00098 00099 /* add, saturate and store four samples to destination at a time*/ 00100 out3 = __QADD8(inA1, inB1); 00101 00102 /* read four input samples at a time from sourceB*/ 00103 inB2 = _SIMD32_OFFSET(pSrcB + 12); 00104 00105 /* increment sourceA pointer by 16 */ 00106 pSrcA += 16u; 00107 00108 /* store four samples at a time to destination */ 00109 _SIMD32_OFFSET(pDst + 4) = out2; 00110 00111 /* add, saturate and store four samples to destination at a time*/ 00112 out4 = __QADD8(inA2, inB2); 00113 00114 /* store four samples at a time to destination */ 00115 _SIMD32_OFFSET(pDst + 8) = out3; 00116 00117 /* increment sourecB pointer by 16 */ 00118 pSrcB += 16u; 00119 00120 /* store four samples at a time to destination */ 00121 _SIMD32_OFFSET(pDst + 12) = out4; 00122 00123 /* incremnet destination pointer by 16 */ 00124 pDst += 16u; 00125 00126 /* Decrement the loop counter */ 00127 blkCnt--; 00128 } 00129 00130 /* If the blockSize is not a multiple of 16, compute any remaining output samples here. 00131 ** No loop unrolling is used. */ 00132 blkCnt = blockSize % 0x10u; 00133 00134 while(blkCnt > 0u) 00135 { 00136 /* C = A + B */ 00137 /* Add and then store the results in the destination buffer. */ 00138 #ifdef CCS 00139 *pDst++ = (q7_t) __SSATA(*pSrcA++ + *pSrcB++, 0, 8); 00140 #else 00141 *pDst++ = (q7_t) __SSAT(*pSrcA++ + *pSrcB++, 8); 00142 #endif // #ifdef CCS 00143 00144 /* Decrement the loop counter */ 00145 blkCnt--; 00146 } 00147 00148 } 00149
