arm_offset_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_offset_q7.c 00009 * 00010 * Description: Q7 vector offset. 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_offset_q7( 00054 q7_t * pSrc, 00055 q7_t offset, 00056 q7_t * pDst, 00057 uint32_t blockSize) 00058 { 00059 uint32_t blkCnt; /* loop counter */ 00060 q31_t offset_packed; /* Offset packed to 32 bit */ 00061 q31_t in1, in2, in3, in4; /* Temporary variabels */ 00062 q7_t in; /* Temporary variable */ 00063 00064 /*loop Unrolling */ 00065 blkCnt = blockSize >> 4u; 00066 00067 /* Offset is packed to 32 bit in order to use SIMD32 for addition */ 00068 offset_packed = __PACKq7(offset, offset, offset, offset); 00069 00070 /* First part of the processing with loop unrolling. Compute 8 outputs at a time. 00071 ** a second loop below computes the remaining 1 to 7 samples. */ 00072 while(blkCnt > 0u) 00073 { 00074 /* C = A + offset */ 00075 /* Add offset and then store the results in the destination bufferfor 8 samples at a time. */ 00076 /* read 4 samples at a time from the source */ 00077 in1 = _SIMD32_OFFSET(pSrc); 00078 in2 = _SIMD32_OFFSET(pSrc + 4); 00079 00080 /* added offset to the four samples at a time */ 00081 in1 = __QADD8(in1, offset_packed); 00082 00083 /* read 4 samples at a time from the source */ 00084 in3 = _SIMD32_OFFSET(pSrc + 8); 00085 00086 /* added offset to the four samples at a time */ 00087 in2 = __QADD8(in2, offset_packed); 00088 00089 /* read 4 samples at a time from the source */ 00090 in4 = _SIMD32_OFFSET(pSrc + 12); 00091 00092 /* store result to the destination four samples at a time */ 00093 _SIMD32_OFFSET(pDst) = in1; 00094 00095 /* added offset to the four samples at a time */ 00096 in3 = __QADD8(in3, offset_packed); 00097 00098 /* store result to the destination four samples at a time */ 00099 _SIMD32_OFFSET(pDst + 4) = in2; 00100 00101 /* added offset to the four samples at a time */ 00102 in4 = __QADD8(in4, offset_packed); 00103 00104 /* store result to the destination four samples at a time */ 00105 _SIMD32_OFFSET(pDst + 8) = in3; 00106 00107 /* increment source by 16 to process next samples */ 00108 pSrc += 16u; 00109 00110 /* store result to the destination four samples at a time */ 00111 _SIMD32_OFFSET(pDst + 12) = in4; 00112 00113 /* increment destination by 16 */ 00114 pDst += 16u; 00115 00116 /* Decrement the loop counter */ 00117 blkCnt--; 00118 } 00119 00120 /* If the blockSize is not a multiple of 8, compute any remaining output samples here. 00121 ** No loop unrolling is used. */ 00122 blkCnt = blockSize % 0x10u; 00123 00124 while(blkCnt > 0u) 00125 { 00126 /* C = A + offset */ 00127 /* Add offset and then store the result in the destination buffer. */ 00128 in = *pSrc++; 00129 *pDst++ = (q7_t)(__QADD8(in, offset)); 00130 00131 /* Decrement the loop counter */ 00132 blkCnt--; 00133 } 00134 } 00135
