| <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> |
| <html xmlns="http://www.w3.org/1999/xhtml"> |
| <head> |
| <meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/> |
| <title>CMSIS DSP Software Library: arm_correlate_f32.c Source File</title> |
| <link href="tabs.css" rel="stylesheet" type="text/css"/> |
| <link href="search/search.css" rel="stylesheet" type="text/css"/> |
| <script type="text/javaScript" src="search/search.js"></script> |
| <link href="doxygen.css" rel="stylesheet" type="text/css"/> |
| </head> |
| <body onload='searchBox.OnSelectItem(0);'> |
| <!-- Generated by Doxygen 1.7.2 --> |
| <script type="text/javascript"><!-- |
| var searchBox = new SearchBox("searchBox", "search",false,'Search'); |
| --></script> |
| <div class="navigation" id="top"> |
| <div class="tabs"> |
| <ul class="tablist"> |
| <li><a href="index.html"><span>Main Page</span></a></li> |
| <li><a href="modules.html"><span>Modules</span></a></li> |
| <li><a href="annotated.html"><span>Data Structures</span></a></li> |
| <li class="current"><a href="files.html"><span>Files</span></a></li> |
| <li><a href="examples.html"><span>Examples</span></a></li> |
| <li id="searchli"> |
| <div id="MSearchBox" class="MSearchBoxInactive"> |
| <span class="left"> |
| <img id="MSearchSelect" src="search/mag_sel.png" |
| onmouseover="return searchBox.OnSearchSelectShow()" |
| onmouseout="return searchBox.OnSearchSelectHide()" |
| alt=""/> |
| <input type="text" id="MSearchField" value="Search" accesskey="S" |
| onfocus="searchBox.OnSearchFieldFocus(true)" |
| onblur="searchBox.OnSearchFieldFocus(false)" |
| onkeyup="searchBox.OnSearchFieldChange(event)"/> |
| </span><span class="right"> |
| <a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a> |
| </span> |
| </div> |
| </li> |
| </ul> |
| </div> |
| <div class="tabs2"> |
| <ul class="tablist"> |
| <li><a href="files.html"><span>File List</span></a></li> |
| <li><a href="globals.html"><span>Globals</span></a></li> |
| </ul> |
| </div> |
| <div class="header"> |
| <div class="headertitle"> |
| <h1>arm_correlate_f32.c</h1> </div> |
| </div> |
| <div class="contents"> |
| <a href="arm__correlate__f32_8c.html">Go to the documentation of this file.</a><div class="fragment"><pre class="fragment"><a name="l00001"></a>00001 <span class="comment">/* ---------------------------------------------------------------------------- </span> |
| <a name="l00002"></a>00002 <span class="comment">* Copyright (C) 2010 ARM Limited. All rights reserved. </span> |
| <a name="l00003"></a>00003 <span class="comment">* </span> |
| <a name="l00004"></a>00004 <span class="comment">* $Date: 15. July 2011 </span> |
| <a name="l00005"></a>00005 <span class="comment">* $Revision: V1.0.10 </span> |
| <a name="l00006"></a>00006 <span class="comment">* </span> |
| <a name="l00007"></a>00007 <span class="comment">* Project: CMSIS DSP Library </span> |
| <a name="l00008"></a>00008 <span class="comment">* Title: arm_correlate_f32.c </span> |
| <a name="l00009"></a>00009 <span class="comment">* </span> |
| <a name="l00010"></a>00010 <span class="comment">* Description: Correlation of floating-point sequences. </span> |
| <a name="l00011"></a>00011 <span class="comment">* </span> |
| <a name="l00012"></a>00012 <span class="comment">* Target Processor: Cortex-M4/Cortex-M3/Cortex-M0</span> |
| <a name="l00013"></a>00013 <span class="comment">* </span> |
| <a name="l00014"></a>00014 <span class="comment">* Version 1.0.10 2011/7/15 </span> |
| <a name="l00015"></a>00015 <span class="comment">* Big Endian support added and Merged M0 and M3/M4 Source code. </span> |
| <a name="l00016"></a>00016 <span class="comment">* </span> |
| <a name="l00017"></a>00017 <span class="comment">* Version 1.0.3 2010/11/29 </span> |
| <a name="l00018"></a>00018 <span class="comment">* Re-organized the CMSIS folders and updated documentation. </span> |
| <a name="l00019"></a>00019 <span class="comment">* </span> |
| <a name="l00020"></a>00020 <span class="comment">* Version 1.0.2 2010/11/11 </span> |
| <a name="l00021"></a>00021 <span class="comment">* Documentation updated. </span> |
| <a name="l00022"></a>00022 <span class="comment">* </span> |
| <a name="l00023"></a>00023 <span class="comment">* Version 1.0.1 2010/10/05 </span> |
| <a name="l00024"></a>00024 <span class="comment">* Production release and review comments incorporated. </span> |
| <a name="l00025"></a>00025 <span class="comment">* </span> |
| <a name="l00026"></a>00026 <span class="comment">* Version 1.0.0 2010/09/20 </span> |
| <a name="l00027"></a>00027 <span class="comment">* Production release and review comments incorporated </span> |
| <a name="l00028"></a>00028 <span class="comment">* </span> |
| <a name="l00029"></a>00029 <span class="comment">* Version 0.0.7 2010/06/10 </span> |
| <a name="l00030"></a>00030 <span class="comment">* Misra-C changes done </span> |
| <a name="l00031"></a>00031 <span class="comment">* </span> |
| <a name="l00032"></a>00032 <span class="comment">* -------------------------------------------------------------------------- */</span> |
| <a name="l00033"></a>00033 |
| <a name="l00034"></a>00034 <span class="preprocessor">#include "<a class="code" href="arm__math_8h.html">arm_math.h</a>"</span> |
| <a name="l00035"></a>00035 |
| <a name="l00095"></a><a class="code" href="group___corr.html#ga22021e4222773f01e9960358a531cfb8">00095</a> <span class="keywordtype">void</span> <a class="code" href="group___corr.html#ga22021e4222773f01e9960358a531cfb8" title="Correlation of floating-point sequences.">arm_correlate_f32</a>( |
| <a name="l00096"></a>00096 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> * pSrcA, |
| <a name="l00097"></a>00097 uint32_t <a class="code" href="arm__convolution__example__f32_8c.html#ace48ed566e2cd6a680f0681192e6af28">srcALen</a>, |
| <a name="l00098"></a>00098 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> * pSrcB, |
| <a name="l00099"></a>00099 uint32_t <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>, |
| <a name="l00100"></a>00100 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> * pDst) |
| <a name="l00101"></a>00101 { |
| <a name="l00102"></a>00102 |
| <a name="l00103"></a>00103 |
| <a name="l00104"></a>00104 <span class="preprocessor">#ifndef ARM_MATH_CM0</span> |
| <a name="l00105"></a>00105 <span class="preprocessor"></span> |
| <a name="l00106"></a>00106 <span class="comment">/* Run the below code for Cortex-M4 and Cortex-M3 */</span> |
| <a name="l00107"></a>00107 |
| <a name="l00108"></a>00108 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> *pIn1; <span class="comment">/* inputA pointer */</span> |
| <a name="l00109"></a>00109 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> *pIn2; <span class="comment">/* inputB pointer */</span> |
| <a name="l00110"></a>00110 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> *pOut = pDst; <span class="comment">/* output pointer */</span> |
| <a name="l00111"></a>00111 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> *px; <span class="comment">/* Intermediate inputA pointer */</span> |
| <a name="l00112"></a>00112 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> *py; <span class="comment">/* Intermediate inputB pointer */</span> |
| <a name="l00113"></a>00113 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> *pSrc1; <span class="comment">/* Intermediate pointers */</span> |
| <a name="l00114"></a>00114 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> sum, acc0, acc1, acc2, acc3; <span class="comment">/* Accumulators */</span> |
| <a name="l00115"></a>00115 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> x0, x1, x2, x3, c0; <span class="comment">/* temporary variables for holding input and coefficient values */</span> |
| <a name="l00116"></a>00116 uint32_t j, k = 0u, count, blkCnt, outBlockSize, blockSize1, blockSize2, blockSize3; <span class="comment">/* loop counters */</span> |
| <a name="l00117"></a>00117 int32_t inc = 1; <span class="comment">/* Destination address modifier */</span> |
| <a name="l00118"></a>00118 |
| <a name="l00119"></a>00119 |
| <a name="l00120"></a>00120 <span class="comment">/* The algorithm implementation is based on the lengths of the inputs. */</span> |
| <a name="l00121"></a>00121 <span class="comment">/* srcB is always made to slide across srcA. */</span> |
| <a name="l00122"></a>00122 <span class="comment">/* So srcBLen is always considered as shorter or equal to srcALen */</span> |
| <a name="l00123"></a>00123 <span class="comment">/* But CORR(x, y) is reverse of CORR(y, x) */</span> |
| <a name="l00124"></a>00124 <span class="comment">/* So, when srcBLen > srcALen, output pointer is made to point to the end of the output buffer */</span> |
| <a name="l00125"></a>00125 <span class="comment">/* and the destination pointer modifier, inc is set to -1 */</span> |
| <a name="l00126"></a>00126 <span class="comment">/* If srcALen > srcBLen, zero pad has to be done to srcB to make the two inputs of same length */</span> |
| <a name="l00127"></a>00127 <span class="comment">/* But to improve the performance, </span> |
| <a name="l00128"></a>00128 <span class="comment"> * we include zeroes in the output instead of zero padding either of the the inputs*/</span> |
| <a name="l00129"></a>00129 <span class="comment">/* If srcALen > srcBLen, </span> |
| <a name="l00130"></a>00130 <span class="comment"> * (srcALen - srcBLen) zeroes has to included in the starting of the output buffer */</span> |
| <a name="l00131"></a>00131 <span class="comment">/* If srcALen < srcBLen, </span> |
| <a name="l00132"></a>00132 <span class="comment"> * (srcALen - srcBLen) zeroes has to included in the ending of the output buffer */</span> |
| <a name="l00133"></a>00133 <span class="keywordflow">if</span>(srcALen >= srcBLen) |
| <a name="l00134"></a>00134 { |
| <a name="l00135"></a>00135 <span class="comment">/* Initialization of inputA pointer */</span> |
| <a name="l00136"></a>00136 pIn1 = pSrcA; |
| <a name="l00137"></a>00137 |
| <a name="l00138"></a>00138 <span class="comment">/* Initialization of inputB pointer */</span> |
| <a name="l00139"></a>00139 pIn2 = pSrcB; |
| <a name="l00140"></a>00140 |
| <a name="l00141"></a>00141 <span class="comment">/* Number of output samples is calculated */</span> |
| <a name="l00142"></a>00142 outBlockSize = (2u * <a class="code" href="arm__convolution__example__f32_8c.html#ace48ed566e2cd6a680f0681192e6af28">srcALen</a>) - 1u; |
| <a name="l00143"></a>00143 |
| <a name="l00144"></a>00144 <span class="comment">/* When srcALen > srcBLen, zero padding has to be done to srcB </span> |
| <a name="l00145"></a>00145 <span class="comment"> * to make their lengths equal. </span> |
| <a name="l00146"></a>00146 <span class="comment"> * Instead, (outBlockSize - (srcALen + srcBLen - 1)) </span> |
| <a name="l00147"></a>00147 <span class="comment"> * number of output samples are made zero */</span> |
| <a name="l00148"></a>00148 j = outBlockSize - (srcALen + (srcBLen - 1u)); |
| <a name="l00149"></a>00149 |
| <a name="l00150"></a>00150 <span class="comment">/* Updating the pointer position to non zero value */</span> |
| <a name="l00151"></a>00151 pOut += j; |
| <a name="l00152"></a>00152 |
| <a name="l00153"></a>00153 <span class="comment">//while(j > 0u) </span> |
| <a name="l00154"></a>00154 <span class="comment">//{ </span> |
| <a name="l00155"></a>00155 <span class="comment">// /* Zero is stored in the destination buffer */ </span> |
| <a name="l00156"></a>00156 <span class="comment">// *pOut++ = 0.0f; </span> |
| <a name="l00157"></a>00157 |
| <a name="l00158"></a>00158 <span class="comment">// /* Decrement the loop counter */ </span> |
| <a name="l00159"></a>00159 <span class="comment">// j--; </span> |
| <a name="l00160"></a>00160 <span class="comment">//} </span> |
| <a name="l00161"></a>00161 |
| <a name="l00162"></a>00162 } |
| <a name="l00163"></a>00163 <span class="keywordflow">else</span> |
| <a name="l00164"></a>00164 { |
| <a name="l00165"></a>00165 <span class="comment">/* Initialization of inputA pointer */</span> |
| <a name="l00166"></a>00166 pIn1 = pSrcB; |
| <a name="l00167"></a>00167 |
| <a name="l00168"></a>00168 <span class="comment">/* Initialization of inputB pointer */</span> |
| <a name="l00169"></a>00169 pIn2 = pSrcA; |
| <a name="l00170"></a>00170 |
| <a name="l00171"></a>00171 <span class="comment">/* srcBLen is always considered as shorter or equal to srcALen */</span> |
| <a name="l00172"></a>00172 j = <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>; |
| <a name="l00173"></a>00173 srcBLen = <a class="code" href="arm__convolution__example__f32_8c.html#ace48ed566e2cd6a680f0681192e6af28">srcALen</a>; |
| <a name="l00174"></a>00174 srcALen = j; |
| <a name="l00175"></a>00175 |
| <a name="l00176"></a>00176 <span class="comment">/* CORR(x, y) = Reverse order(CORR(y, x)) */</span> |
| <a name="l00177"></a>00177 <span class="comment">/* Hence set the destination pointer to point to the last output sample */</span> |
| <a name="l00178"></a>00178 pOut = pDst + ((srcALen + <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>) - 2u); |
| <a name="l00179"></a>00179 |
| <a name="l00180"></a>00180 <span class="comment">/* Destination address modifier is set to -1 */</span> |
| <a name="l00181"></a>00181 inc = -1; |
| <a name="l00182"></a>00182 |
| <a name="l00183"></a>00183 } |
| <a name="l00184"></a>00184 |
| <a name="l00185"></a>00185 <span class="comment">/* The function is internally </span> |
| <a name="l00186"></a>00186 <span class="comment"> * divided into three parts according to the number of multiplications that has to be </span> |
| <a name="l00187"></a>00187 <span class="comment"> * taken place between inputA samples and inputB samples. In the first part of the </span> |
| <a name="l00188"></a>00188 <span class="comment"> * algorithm, the multiplications increase by one for every iteration. </span> |
| <a name="l00189"></a>00189 <span class="comment"> * In the second part of the algorithm, srcBLen number of multiplications are done. </span> |
| <a name="l00190"></a>00190 <span class="comment"> * In the third part of the algorithm, the multiplications decrease by one </span> |
| <a name="l00191"></a>00191 <span class="comment"> * for every iteration.*/</span> |
| <a name="l00192"></a>00192 <span class="comment">/* The algorithm is implemented in three stages. </span> |
| <a name="l00193"></a>00193 <span class="comment"> * The loop counters of each stage is initiated here. */</span> |
| <a name="l00194"></a>00194 blockSize1 = srcBLen - 1u; |
| <a name="l00195"></a>00195 blockSize2 = srcALen - (srcBLen - 1u); |
| <a name="l00196"></a>00196 blockSize3 = blockSize1; |
| <a name="l00197"></a>00197 |
| <a name="l00198"></a>00198 <span class="comment">/* -------------------------- </span> |
| <a name="l00199"></a>00199 <span class="comment"> * Initializations of stage1 </span> |
| <a name="l00200"></a>00200 <span class="comment"> * -------------------------*/</span> |
| <a name="l00201"></a>00201 |
| <a name="l00202"></a>00202 <span class="comment">/* sum = x[0] * y[srcBlen - 1] </span> |
| <a name="l00203"></a>00203 <span class="comment"> * sum = x[0] * y[srcBlen-2] + x[1] * y[srcBlen - 1] </span> |
| <a name="l00204"></a>00204 <span class="comment"> * .... </span> |
| <a name="l00205"></a>00205 <span class="comment"> * sum = x[0] * y[0] + x[1] * y[1] +...+ x[srcBLen - 1] * y[srcBLen - 1] </span> |
| <a name="l00206"></a>00206 <span class="comment"> */</span> |
| <a name="l00207"></a>00207 |
| <a name="l00208"></a>00208 <span class="comment">/* In this stage the MAC operations are increased by 1 for every iteration. </span> |
| <a name="l00209"></a>00209 <span class="comment"> The count variable holds the number of MAC operations performed */</span> |
| <a name="l00210"></a>00210 count = 1u; |
| <a name="l00211"></a>00211 |
| <a name="l00212"></a>00212 <span class="comment">/* Working pointer of inputA */</span> |
| <a name="l00213"></a>00213 px = pIn1; |
| <a name="l00214"></a>00214 |
| <a name="l00215"></a>00215 <span class="comment">/* Working pointer of inputB */</span> |
| <a name="l00216"></a>00216 pSrc1 = pIn2 + (srcBLen - 1u); |
| <a name="l00217"></a>00217 py = pSrc1; |
| <a name="l00218"></a>00218 |
| <a name="l00219"></a>00219 <span class="comment">/* ------------------------ </span> |
| <a name="l00220"></a>00220 <span class="comment"> * Stage1 process </span> |
| <a name="l00221"></a>00221 <span class="comment"> * ----------------------*/</span> |
| <a name="l00222"></a>00222 |
| <a name="l00223"></a>00223 <span class="comment">/* The first stage starts here */</span> |
| <a name="l00224"></a>00224 <span class="keywordflow">while</span>(blockSize1 > 0u) |
| <a name="l00225"></a>00225 { |
| <a name="l00226"></a>00226 <span class="comment">/* Accumulator is made zero for every iteration */</span> |
| <a name="l00227"></a>00227 sum = 0.0f; |
| <a name="l00228"></a>00228 |
| <a name="l00229"></a>00229 <span class="comment">/* Apply loop unrolling and compute 4 MACs simultaneously. */</span> |
| <a name="l00230"></a>00230 k = count >> 2u; |
| <a name="l00231"></a>00231 |
| <a name="l00232"></a>00232 <span class="comment">/* First part of the processing with loop unrolling. Compute 4 MACs at a time. </span> |
| <a name="l00233"></a>00233 <span class="comment"> ** a second loop below computes MACs for the remaining 1 to 3 samples. */</span> |
| <a name="l00234"></a>00234 <span class="keywordflow">while</span>(k > 0u) |
| <a name="l00235"></a>00235 { |
| <a name="l00236"></a>00236 <span class="comment">/* x[0] * y[srcBLen - 4] */</span> |
| <a name="l00237"></a>00237 sum += *px++ * *py++; |
| <a name="l00238"></a>00238 <span class="comment">/* x[1] * y[srcBLen - 3] */</span> |
| <a name="l00239"></a>00239 sum += *px++ * *py++; |
| <a name="l00240"></a>00240 <span class="comment">/* x[2] * y[srcBLen - 2] */</span> |
| <a name="l00241"></a>00241 sum += *px++ * *py++; |
| <a name="l00242"></a>00242 <span class="comment">/* x[3] * y[srcBLen - 1] */</span> |
| <a name="l00243"></a>00243 sum += *px++ * *py++; |
| <a name="l00244"></a>00244 |
| <a name="l00245"></a>00245 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00246"></a>00246 k--; |
| <a name="l00247"></a>00247 } |
| <a name="l00248"></a>00248 |
| <a name="l00249"></a>00249 <span class="comment">/* If the count is not a multiple of 4, compute any remaining MACs here. </span> |
| <a name="l00250"></a>00250 <span class="comment"> ** No loop unrolling is used. */</span> |
| <a name="l00251"></a>00251 k = count % 0x4u; |
| <a name="l00252"></a>00252 |
| <a name="l00253"></a>00253 <span class="keywordflow">while</span>(k > 0u) |
| <a name="l00254"></a>00254 { |
| <a name="l00255"></a>00255 <span class="comment">/* Perform the multiply-accumulate */</span> |
| <a name="l00256"></a>00256 <span class="comment">/* x[0] * y[srcBLen - 1] */</span> |
| <a name="l00257"></a>00257 sum += *px++ * *py++; |
| <a name="l00258"></a>00258 |
| <a name="l00259"></a>00259 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00260"></a>00260 k--; |
| <a name="l00261"></a>00261 } |
| <a name="l00262"></a>00262 |
| <a name="l00263"></a>00263 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span> |
| <a name="l00264"></a>00264 *pOut = sum; |
| <a name="l00265"></a>00265 <span class="comment">/* Destination pointer is updated according to the address modifier, inc */</span> |
| <a name="l00266"></a>00266 pOut += inc; |
| <a name="l00267"></a>00267 |
| <a name="l00268"></a>00268 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span> |
| <a name="l00269"></a>00269 py = pSrc1 - count; |
| <a name="l00270"></a>00270 px = pIn1; |
| <a name="l00271"></a>00271 |
| <a name="l00272"></a>00272 <span class="comment">/* Increment the MAC count */</span> |
| <a name="l00273"></a>00273 count++; |
| <a name="l00274"></a>00274 |
| <a name="l00275"></a>00275 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00276"></a>00276 blockSize1--; |
| <a name="l00277"></a>00277 } |
| <a name="l00278"></a>00278 |
| <a name="l00279"></a>00279 <span class="comment">/* -------------------------- </span> |
| <a name="l00280"></a>00280 <span class="comment"> * Initializations of stage2 </span> |
| <a name="l00281"></a>00281 <span class="comment"> * ------------------------*/</span> |
| <a name="l00282"></a>00282 |
| <a name="l00283"></a>00283 <span class="comment">/* sum = x[0] * y[0] + x[1] * y[1] +...+ x[srcBLen-1] * y[srcBLen-1] </span> |
| <a name="l00284"></a>00284 <span class="comment"> * sum = x[1] * y[0] + x[2] * y[1] +...+ x[srcBLen] * y[srcBLen-1] </span> |
| <a name="l00285"></a>00285 <span class="comment"> * .... </span> |
| <a name="l00286"></a>00286 <span class="comment"> * sum = x[srcALen-srcBLen-2] * y[0] + x[srcALen-srcBLen-1] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] </span> |
| <a name="l00287"></a>00287 <span class="comment"> */</span> |
| <a name="l00288"></a>00288 |
| <a name="l00289"></a>00289 <span class="comment">/* Working pointer of inputA */</span> |
| <a name="l00290"></a>00290 px = pIn1; |
| <a name="l00291"></a>00291 |
| <a name="l00292"></a>00292 <span class="comment">/* Working pointer of inputB */</span> |
| <a name="l00293"></a>00293 py = pIn2; |
| <a name="l00294"></a>00294 |
| <a name="l00295"></a>00295 <span class="comment">/* count is index by which the pointer pIn1 to be incremented */</span> |
| <a name="l00296"></a>00296 count = 1u; |
| <a name="l00297"></a>00297 |
| <a name="l00298"></a>00298 <span class="comment">/* ------------------- </span> |
| <a name="l00299"></a>00299 <span class="comment"> * Stage2 process </span> |
| <a name="l00300"></a>00300 <span class="comment"> * ------------------*/</span> |
| <a name="l00301"></a>00301 |
| <a name="l00302"></a>00302 <span class="comment">/* Stage2 depends on srcBLen as in this stage srcBLen number of MACS are performed. </span> |
| <a name="l00303"></a>00303 <span class="comment"> * So, to loop unroll over blockSize2, </span> |
| <a name="l00304"></a>00304 <span class="comment"> * srcBLen should be greater than or equal to 4, to loop unroll the srcBLen loop */</span> |
| <a name="l00305"></a>00305 <span class="keywordflow">if</span>(srcBLen >= 4u) |
| <a name="l00306"></a>00306 { |
| <a name="l00307"></a>00307 <span class="comment">/* Loop unroll over blockSize2, by 4 */</span> |
| <a name="l00308"></a>00308 blkCnt = blockSize2 >> 2u; |
| <a name="l00309"></a>00309 |
| <a name="l00310"></a>00310 <span class="keywordflow">while</span>(blkCnt > 0u) |
| <a name="l00311"></a>00311 { |
| <a name="l00312"></a>00312 <span class="comment">/* Set all accumulators to zero */</span> |
| <a name="l00313"></a>00313 acc0 = 0.0f; |
| <a name="l00314"></a>00314 acc1 = 0.0f; |
| <a name="l00315"></a>00315 acc2 = 0.0f; |
| <a name="l00316"></a>00316 acc3 = 0.0f; |
| <a name="l00317"></a>00317 |
| <a name="l00318"></a>00318 <span class="comment">/* read x[0], x[1], x[2] samples */</span> |
| <a name="l00319"></a>00319 x0 = *(px++); |
| <a name="l00320"></a>00320 x1 = *(px++); |
| <a name="l00321"></a>00321 x2 = *(px++); |
| <a name="l00322"></a>00322 |
| <a name="l00323"></a>00323 <span class="comment">/* Apply loop unrolling and compute 4 MACs simultaneously. */</span> |
| <a name="l00324"></a>00324 k = srcBLen >> 2u; |
| <a name="l00325"></a>00325 |
| <a name="l00326"></a>00326 <span class="comment">/* First part of the processing with loop unrolling. Compute 4 MACs at a time. </span> |
| <a name="l00327"></a>00327 <span class="comment"> ** a second loop below computes MACs for the remaining 1 to 3 samples. */</span> |
| <a name="l00328"></a>00328 <span class="keywordflow">do</span> |
| <a name="l00329"></a>00329 { |
| <a name="l00330"></a>00330 <span class="comment">/* Read y[0] sample */</span> |
| <a name="l00331"></a>00331 c0 = *(py++); |
| <a name="l00332"></a>00332 |
| <a name="l00333"></a>00333 <span class="comment">/* Read x[3] sample */</span> |
| <a name="l00334"></a>00334 x3 = *(px++); |
| <a name="l00335"></a>00335 |
| <a name="l00336"></a>00336 <span class="comment">/* Perform the multiply-accumulate */</span> |
| <a name="l00337"></a>00337 <span class="comment">/* acc0 += x[0] * y[0] */</span> |
| <a name="l00338"></a>00338 acc0 += x0 * c0; |
| <a name="l00339"></a>00339 <span class="comment">/* acc1 += x[1] * y[0] */</span> |
| <a name="l00340"></a>00340 acc1 += x1 * c0; |
| <a name="l00341"></a>00341 <span class="comment">/* acc2 += x[2] * y[0] */</span> |
| <a name="l00342"></a>00342 acc2 += x2 * c0; |
| <a name="l00343"></a>00343 <span class="comment">/* acc3 += x[3] * y[0] */</span> |
| <a name="l00344"></a>00344 acc3 += x3 * c0; |
| <a name="l00345"></a>00345 |
| <a name="l00346"></a>00346 <span class="comment">/* Read y[1] sample */</span> |
| <a name="l00347"></a>00347 c0 = *(py++); |
| <a name="l00348"></a>00348 |
| <a name="l00349"></a>00349 <span class="comment">/* Read x[4] sample */</span> |
| <a name="l00350"></a>00350 x0 = *(px++); |
| <a name="l00351"></a>00351 |
| <a name="l00352"></a>00352 <span class="comment">/* Perform the multiply-accumulate */</span> |
| <a name="l00353"></a>00353 <span class="comment">/* acc0 += x[1] * y[1] */</span> |
| <a name="l00354"></a>00354 acc0 += x1 * c0; |
| <a name="l00355"></a>00355 <span class="comment">/* acc1 += x[2] * y[1] */</span> |
| <a name="l00356"></a>00356 acc1 += x2 * c0; |
| <a name="l00357"></a>00357 <span class="comment">/* acc2 += x[3] * y[1] */</span> |
| <a name="l00358"></a>00358 acc2 += x3 * c0; |
| <a name="l00359"></a>00359 <span class="comment">/* acc3 += x[4] * y[1] */</span> |
| <a name="l00360"></a>00360 acc3 += x0 * c0; |
| <a name="l00361"></a>00361 |
| <a name="l00362"></a>00362 <span class="comment">/* Read y[2] sample */</span> |
| <a name="l00363"></a>00363 c0 = *(py++); |
| <a name="l00364"></a>00364 |
| <a name="l00365"></a>00365 <span class="comment">/* Read x[5] sample */</span> |
| <a name="l00366"></a>00366 x1 = *(px++); |
| <a name="l00367"></a>00367 |
| <a name="l00368"></a>00368 <span class="comment">/* Perform the multiply-accumulates */</span> |
| <a name="l00369"></a>00369 <span class="comment">/* acc0 += x[2] * y[2] */</span> |
| <a name="l00370"></a>00370 acc0 += x2 * c0; |
| <a name="l00371"></a>00371 <span class="comment">/* acc1 += x[3] * y[2] */</span> |
| <a name="l00372"></a>00372 acc1 += x3 * c0; |
| <a name="l00373"></a>00373 <span class="comment">/* acc2 += x[4] * y[2] */</span> |
| <a name="l00374"></a>00374 acc2 += x0 * c0; |
| <a name="l00375"></a>00375 <span class="comment">/* acc3 += x[5] * y[2] */</span> |
| <a name="l00376"></a>00376 acc3 += x1 * c0; |
| <a name="l00377"></a>00377 |
| <a name="l00378"></a>00378 <span class="comment">/* Read y[3] sample */</span> |
| <a name="l00379"></a>00379 c0 = *(py++); |
| <a name="l00380"></a>00380 |
| <a name="l00381"></a>00381 <span class="comment">/* Read x[6] sample */</span> |
| <a name="l00382"></a>00382 x2 = *(px++); |
| <a name="l00383"></a>00383 |
| <a name="l00384"></a>00384 <span class="comment">/* Perform the multiply-accumulates */</span> |
| <a name="l00385"></a>00385 <span class="comment">/* acc0 += x[3] * y[3] */</span> |
| <a name="l00386"></a>00386 acc0 += x3 * c0; |
| <a name="l00387"></a>00387 <span class="comment">/* acc1 += x[4] * y[3] */</span> |
| <a name="l00388"></a>00388 acc1 += x0 * c0; |
| <a name="l00389"></a>00389 <span class="comment">/* acc2 += x[5] * y[3] */</span> |
| <a name="l00390"></a>00390 acc2 += x1 * c0; |
| <a name="l00391"></a>00391 <span class="comment">/* acc3 += x[6] * y[3] */</span> |
| <a name="l00392"></a>00392 acc3 += x2 * c0; |
| <a name="l00393"></a>00393 |
| <a name="l00394"></a>00394 |
| <a name="l00395"></a>00395 } <span class="keywordflow">while</span>(--k); |
| <a name="l00396"></a>00396 |
| <a name="l00397"></a>00397 <span class="comment">/* If the srcBLen is not a multiple of 4, compute any remaining MACs here. </span> |
| <a name="l00398"></a>00398 <span class="comment"> ** No loop unrolling is used. */</span> |
| <a name="l00399"></a>00399 k = srcBLen % 0x4u; |
| <a name="l00400"></a>00400 |
| <a name="l00401"></a>00401 <span class="keywordflow">while</span>(k > 0u) |
| <a name="l00402"></a>00402 { |
| <a name="l00403"></a>00403 <span class="comment">/* Read y[4] sample */</span> |
| <a name="l00404"></a>00404 c0 = *(py++); |
| <a name="l00405"></a>00405 |
| <a name="l00406"></a>00406 <span class="comment">/* Read x[7] sample */</span> |
| <a name="l00407"></a>00407 x3 = *(px++); |
| <a name="l00408"></a>00408 |
| <a name="l00409"></a>00409 <span class="comment">/* Perform the multiply-accumulates */</span> |
| <a name="l00410"></a>00410 <span class="comment">/* acc0 += x[4] * y[4] */</span> |
| <a name="l00411"></a>00411 acc0 += x0 * c0; |
| <a name="l00412"></a>00412 <span class="comment">/* acc1 += x[5] * y[4] */</span> |
| <a name="l00413"></a>00413 acc1 += x1 * c0; |
| <a name="l00414"></a>00414 <span class="comment">/* acc2 += x[6] * y[4] */</span> |
| <a name="l00415"></a>00415 acc2 += x2 * c0; |
| <a name="l00416"></a>00416 <span class="comment">/* acc3 += x[7] * y[4] */</span> |
| <a name="l00417"></a>00417 acc3 += x3 * c0; |
| <a name="l00418"></a>00418 |
| <a name="l00419"></a>00419 <span class="comment">/* Reuse the present samples for the next MAC */</span> |
| <a name="l00420"></a>00420 x0 = x1; |
| <a name="l00421"></a>00421 x1 = x2; |
| <a name="l00422"></a>00422 x2 = x3; |
| <a name="l00423"></a>00423 |
| <a name="l00424"></a>00424 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00425"></a>00425 k--; |
| <a name="l00426"></a>00426 } |
| <a name="l00427"></a>00427 |
| <a name="l00428"></a>00428 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span> |
| <a name="l00429"></a>00429 *pOut = acc0; |
| <a name="l00430"></a>00430 <span class="comment">/* Destination pointer is updated according to the address modifier, inc */</span> |
| <a name="l00431"></a>00431 pOut += inc; |
| <a name="l00432"></a>00432 |
| <a name="l00433"></a>00433 *pOut = acc1; |
| <a name="l00434"></a>00434 pOut += inc; |
| <a name="l00435"></a>00435 |
| <a name="l00436"></a>00436 *pOut = acc2; |
| <a name="l00437"></a>00437 pOut += inc; |
| <a name="l00438"></a>00438 |
| <a name="l00439"></a>00439 *pOut = acc3; |
| <a name="l00440"></a>00440 pOut += inc; |
| <a name="l00441"></a>00441 |
| <a name="l00442"></a>00442 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span> |
| <a name="l00443"></a>00443 px = pIn1 + (count * 4u); |
| <a name="l00444"></a>00444 py = pIn2; |
| <a name="l00445"></a>00445 |
| <a name="l00446"></a>00446 <span class="comment">/* Increment the pointer pIn1 index, count by 1 */</span> |
| <a name="l00447"></a>00447 count++; |
| <a name="l00448"></a>00448 |
| <a name="l00449"></a>00449 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00450"></a>00450 blkCnt--; |
| <a name="l00451"></a>00451 } |
| <a name="l00452"></a>00452 |
| <a name="l00453"></a>00453 <span class="comment">/* If the blockSize2 is not a multiple of 4, compute any remaining output samples here. </span> |
| <a name="l00454"></a>00454 <span class="comment"> ** No loop unrolling is used. */</span> |
| <a name="l00455"></a>00455 blkCnt = blockSize2 % 0x4u; |
| <a name="l00456"></a>00456 |
| <a name="l00457"></a>00457 <span class="keywordflow">while</span>(blkCnt > 0u) |
| <a name="l00458"></a>00458 { |
| <a name="l00459"></a>00459 <span class="comment">/* Accumulator is made zero for every iteration */</span> |
| <a name="l00460"></a>00460 sum = 0.0f; |
| <a name="l00461"></a>00461 |
| <a name="l00462"></a>00462 <span class="comment">/* Apply loop unrolling and compute 4 MACs simultaneously. */</span> |
| <a name="l00463"></a>00463 k = srcBLen >> 2u; |
| <a name="l00464"></a>00464 |
| <a name="l00465"></a>00465 <span class="comment">/* First part of the processing with loop unrolling. Compute 4 MACs at a time. </span> |
| <a name="l00466"></a>00466 <span class="comment"> ** a second loop below computes MACs for the remaining 1 to 3 samples. */</span> |
| <a name="l00467"></a>00467 <span class="keywordflow">while</span>(k > 0u) |
| <a name="l00468"></a>00468 { |
| <a name="l00469"></a>00469 <span class="comment">/* Perform the multiply-accumulates */</span> |
| <a name="l00470"></a>00470 sum += *px++ * *py++; |
| <a name="l00471"></a>00471 sum += *px++ * *py++; |
| <a name="l00472"></a>00472 sum += *px++ * *py++; |
| <a name="l00473"></a>00473 sum += *px++ * *py++; |
| <a name="l00474"></a>00474 |
| <a name="l00475"></a>00475 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00476"></a>00476 k--; |
| <a name="l00477"></a>00477 } |
| <a name="l00478"></a>00478 |
| <a name="l00479"></a>00479 <span class="comment">/* If the srcBLen is not a multiple of 4, compute any remaining MACs here. </span> |
| <a name="l00480"></a>00480 <span class="comment"> ** No loop unrolling is used. */</span> |
| <a name="l00481"></a>00481 k = srcBLen % 0x4u; |
| <a name="l00482"></a>00482 |
| <a name="l00483"></a>00483 <span class="keywordflow">while</span>(k > 0u) |
| <a name="l00484"></a>00484 { |
| <a name="l00485"></a>00485 <span class="comment">/* Perform the multiply-accumulate */</span> |
| <a name="l00486"></a>00486 sum += *px++ * *py++; |
| <a name="l00487"></a>00487 |
| <a name="l00488"></a>00488 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00489"></a>00489 k--; |
| <a name="l00490"></a>00490 } |
| <a name="l00491"></a>00491 |
| <a name="l00492"></a>00492 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span> |
| <a name="l00493"></a>00493 *pOut = sum; |
| <a name="l00494"></a>00494 <span class="comment">/* Destination pointer is updated according to the address modifier, inc */</span> |
| <a name="l00495"></a>00495 pOut += inc; |
| <a name="l00496"></a>00496 |
| <a name="l00497"></a>00497 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span> |
| <a name="l00498"></a>00498 px = pIn1 + count; |
| <a name="l00499"></a>00499 py = pIn2; |
| <a name="l00500"></a>00500 |
| <a name="l00501"></a>00501 <span class="comment">/* Increment the pointer pIn1 index, count by 1 */</span> |
| <a name="l00502"></a>00502 count++; |
| <a name="l00503"></a>00503 |
| <a name="l00504"></a>00504 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00505"></a>00505 blkCnt--; |
| <a name="l00506"></a>00506 } |
| <a name="l00507"></a>00507 } |
| <a name="l00508"></a>00508 <span class="keywordflow">else</span> |
| <a name="l00509"></a>00509 { |
| <a name="l00510"></a>00510 <span class="comment">/* If the srcBLen is not a multiple of 4, </span> |
| <a name="l00511"></a>00511 <span class="comment"> * the blockSize2 loop cannot be unrolled by 4 */</span> |
| <a name="l00512"></a>00512 blkCnt = blockSize2; |
| <a name="l00513"></a>00513 |
| <a name="l00514"></a>00514 <span class="keywordflow">while</span>(blkCnt > 0u) |
| <a name="l00515"></a>00515 { |
| <a name="l00516"></a>00516 <span class="comment">/* Accumulator is made zero for every iteration */</span> |
| <a name="l00517"></a>00517 sum = 0.0f; |
| <a name="l00518"></a>00518 |
| <a name="l00519"></a>00519 <span class="comment">/* Loop over srcBLen */</span> |
| <a name="l00520"></a>00520 k = <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>; |
| <a name="l00521"></a>00521 |
| <a name="l00522"></a>00522 <span class="keywordflow">while</span>(k > 0u) |
| <a name="l00523"></a>00523 { |
| <a name="l00524"></a>00524 <span class="comment">/* Perform the multiply-accumulate */</span> |
| <a name="l00525"></a>00525 sum += *px++ * *py++; |
| <a name="l00526"></a>00526 |
| <a name="l00527"></a>00527 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00528"></a>00528 k--; |
| <a name="l00529"></a>00529 } |
| <a name="l00530"></a>00530 |
| <a name="l00531"></a>00531 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span> |
| <a name="l00532"></a>00532 *pOut = sum; |
| <a name="l00533"></a>00533 <span class="comment">/* Destination pointer is updated according to the address modifier, inc */</span> |
| <a name="l00534"></a>00534 pOut += inc; |
| <a name="l00535"></a>00535 |
| <a name="l00536"></a>00536 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span> |
| <a name="l00537"></a>00537 px = pIn1 + count; |
| <a name="l00538"></a>00538 py = pIn2; |
| <a name="l00539"></a>00539 |
| <a name="l00540"></a>00540 <span class="comment">/* Increment the pointer pIn1 index, count by 1 */</span> |
| <a name="l00541"></a>00541 count++; |
| <a name="l00542"></a>00542 |
| <a name="l00543"></a>00543 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00544"></a>00544 blkCnt--; |
| <a name="l00545"></a>00545 } |
| <a name="l00546"></a>00546 } |
| <a name="l00547"></a>00547 |
| <a name="l00548"></a>00548 <span class="comment">/* -------------------------- </span> |
| <a name="l00549"></a>00549 <span class="comment"> * Initializations of stage3 </span> |
| <a name="l00550"></a>00550 <span class="comment"> * -------------------------*/</span> |
| <a name="l00551"></a>00551 |
| <a name="l00552"></a>00552 <span class="comment">/* sum += x[srcALen-srcBLen+1] * y[0] + x[srcALen-srcBLen+2] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] </span> |
| <a name="l00553"></a>00553 <span class="comment"> * sum += x[srcALen-srcBLen+2] * y[0] + x[srcALen-srcBLen+3] * y[1] +...+ x[srcALen-1] * y[srcBLen-1] </span> |
| <a name="l00554"></a>00554 <span class="comment"> * .... </span> |
| <a name="l00555"></a>00555 <span class="comment"> * sum += x[srcALen-2] * y[0] + x[srcALen-1] * y[1] </span> |
| <a name="l00556"></a>00556 <span class="comment"> * sum += x[srcALen-1] * y[0] </span> |
| <a name="l00557"></a>00557 <span class="comment"> */</span> |
| <a name="l00558"></a>00558 |
| <a name="l00559"></a>00559 <span class="comment">/* In this stage the MAC operations are decreased by 1 for every iteration. </span> |
| <a name="l00560"></a>00560 <span class="comment"> The count variable holds the number of MAC operations performed */</span> |
| <a name="l00561"></a>00561 count = srcBLen - 1u; |
| <a name="l00562"></a>00562 |
| <a name="l00563"></a>00563 <span class="comment">/* Working pointer of inputA */</span> |
| <a name="l00564"></a>00564 pSrc1 = pIn1 + (srcALen - (srcBLen - 1u)); |
| <a name="l00565"></a>00565 px = pSrc1; |
| <a name="l00566"></a>00566 |
| <a name="l00567"></a>00567 <span class="comment">/* Working pointer of inputB */</span> |
| <a name="l00568"></a>00568 py = pIn2; |
| <a name="l00569"></a>00569 |
| <a name="l00570"></a>00570 <span class="comment">/* ------------------- </span> |
| <a name="l00571"></a>00571 <span class="comment"> * Stage3 process </span> |
| <a name="l00572"></a>00572 <span class="comment"> * ------------------*/</span> |
| <a name="l00573"></a>00573 |
| <a name="l00574"></a>00574 <span class="keywordflow">while</span>(blockSize3 > 0u) |
| <a name="l00575"></a>00575 { |
| <a name="l00576"></a>00576 <span class="comment">/* Accumulator is made zero for every iteration */</span> |
| <a name="l00577"></a>00577 sum = 0.0f; |
| <a name="l00578"></a>00578 |
| <a name="l00579"></a>00579 <span class="comment">/* Apply loop unrolling and compute 4 MACs simultaneously. */</span> |
| <a name="l00580"></a>00580 k = count >> 2u; |
| <a name="l00581"></a>00581 |
| <a name="l00582"></a>00582 <span class="comment">/* First part of the processing with loop unrolling. Compute 4 MACs at a time. </span> |
| <a name="l00583"></a>00583 <span class="comment"> ** a second loop below computes MACs for the remaining 1 to 3 samples. */</span> |
| <a name="l00584"></a>00584 <span class="keywordflow">while</span>(k > 0u) |
| <a name="l00585"></a>00585 { |
| <a name="l00586"></a>00586 <span class="comment">/* Perform the multiply-accumulates */</span> |
| <a name="l00587"></a>00587 <span class="comment">/* sum += x[srcALen - srcBLen + 4] * y[3] */</span> |
| <a name="l00588"></a>00588 sum += *px++ * *py++; |
| <a name="l00589"></a>00589 <span class="comment">/* sum += x[srcALen - srcBLen + 3] * y[2] */</span> |
| <a name="l00590"></a>00590 sum += *px++ * *py++; |
| <a name="l00591"></a>00591 <span class="comment">/* sum += x[srcALen - srcBLen + 2] * y[1] */</span> |
| <a name="l00592"></a>00592 sum += *px++ * *py++; |
| <a name="l00593"></a>00593 <span class="comment">/* sum += x[srcALen - srcBLen + 1] * y[0] */</span> |
| <a name="l00594"></a>00594 sum += *px++ * *py++; |
| <a name="l00595"></a>00595 |
| <a name="l00596"></a>00596 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00597"></a>00597 k--; |
| <a name="l00598"></a>00598 } |
| <a name="l00599"></a>00599 |
| <a name="l00600"></a>00600 <span class="comment">/* If the count is not a multiple of 4, compute any remaining MACs here. </span> |
| <a name="l00601"></a>00601 <span class="comment"> ** No loop unrolling is used. */</span> |
| <a name="l00602"></a>00602 k = count % 0x4u; |
| <a name="l00603"></a>00603 |
| <a name="l00604"></a>00604 <span class="keywordflow">while</span>(k > 0u) |
| <a name="l00605"></a>00605 { |
| <a name="l00606"></a>00606 <span class="comment">/* Perform the multiply-accumulates */</span> |
| <a name="l00607"></a>00607 sum += *px++ * *py++; |
| <a name="l00608"></a>00608 |
| <a name="l00609"></a>00609 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00610"></a>00610 k--; |
| <a name="l00611"></a>00611 } |
| <a name="l00612"></a>00612 |
| <a name="l00613"></a>00613 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span> |
| <a name="l00614"></a>00614 *pOut = sum; |
| <a name="l00615"></a>00615 <span class="comment">/* Destination pointer is updated according to the address modifier, inc */</span> |
| <a name="l00616"></a>00616 pOut += inc; |
| <a name="l00617"></a>00617 |
| <a name="l00618"></a>00618 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span> |
| <a name="l00619"></a>00619 px = ++pSrc1; |
| <a name="l00620"></a>00620 py = pIn2; |
| <a name="l00621"></a>00621 |
| <a name="l00622"></a>00622 <span class="comment">/* Decrement the MAC count */</span> |
| <a name="l00623"></a>00623 count--; |
| <a name="l00624"></a>00624 |
| <a name="l00625"></a>00625 <span class="comment">/* Decrement the loop counter */</span> |
| <a name="l00626"></a>00626 blockSize3--; |
| <a name="l00627"></a>00627 } |
| <a name="l00628"></a>00628 |
| <a name="l00629"></a>00629 <span class="preprocessor">#else</span> |
| <a name="l00630"></a>00630 <span class="preprocessor"></span> |
| <a name="l00631"></a>00631 <span class="comment">/* Run the below code for Cortex-M0 */</span> |
| <a name="l00632"></a>00632 |
| <a name="l00633"></a>00633 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> *pIn1 = pSrcA; <span class="comment">/* inputA pointer */</span> |
| <a name="l00634"></a>00634 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> *pIn2 = pSrcB + (srcBLen - 1u); <span class="comment">/* inputB pointer */</span> |
| <a name="l00635"></a>00635 <a class="code" href="arm__math_8h.html#a4611b605e45ab401f02cab15c5e38715" title="32-bit floating-point type definition.">float32_t</a> sum; <span class="comment">/* Accumulator */</span> |
| <a name="l00636"></a>00636 uint32_t i = 0u, j; <span class="comment">/* loop counters */</span> |
| <a name="l00637"></a>00637 uint32_t inv = 0u; <span class="comment">/* Reverse order flag */</span> |
| <a name="l00638"></a>00638 uint32_t tot = 0u; <span class="comment">/* Length */</span> |
| <a name="l00639"></a>00639 |
| <a name="l00640"></a>00640 <span class="comment">/* The algorithm implementation is based on the lengths of the inputs. */</span> |
| <a name="l00641"></a>00641 <span class="comment">/* srcB is always made to slide across srcA. */</span> |
| <a name="l00642"></a>00642 <span class="comment">/* So srcBLen is always considered as shorter or equal to srcALen */</span> |
| <a name="l00643"></a>00643 <span class="comment">/* But CORR(x, y) is reverse of CORR(y, x) */</span> |
| <a name="l00644"></a>00644 <span class="comment">/* So, when srcBLen > srcALen, output pointer is made to point to the end of the output buffer */</span> |
| <a name="l00645"></a>00645 <span class="comment">/* and a varaible, inv is set to 1 */</span> |
| <a name="l00646"></a>00646 <span class="comment">/* If lengths are not equal then zero pad has to be done to make the two </span> |
| <a name="l00647"></a>00647 <span class="comment"> * inputs of same length. But to improve the performance, we include zeroes </span> |
| <a name="l00648"></a>00648 <span class="comment"> * in the output instead of zero padding either of the the inputs*/</span> |
| <a name="l00649"></a>00649 <span class="comment">/* If srcALen > srcBLen, (srcALen - srcBLen) zeroes has to included in the </span> |
| <a name="l00650"></a>00650 <span class="comment"> * starting of the output buffer */</span> |
| <a name="l00651"></a>00651 <span class="comment">/* If srcALen < srcBLen, (srcALen - srcBLen) zeroes has to included in the </span> |
| <a name="l00652"></a>00652 <span class="comment"> * ending of the output buffer */</span> |
| <a name="l00653"></a>00653 <span class="comment">/* Once the zero padding is done the remaining of the output is calcualted </span> |
| <a name="l00654"></a>00654 <span class="comment"> * using convolution but with the shorter signal time shifted. */</span> |
| <a name="l00655"></a>00655 |
| <a name="l00656"></a>00656 <span class="comment">/* Calculate the length of the remaining sequence */</span> |
| <a name="l00657"></a>00657 tot = ((srcALen + <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>) - 2u); |
| <a name="l00658"></a>00658 |
| <a name="l00659"></a>00659 <span class="keywordflow">if</span>(srcALen > srcBLen) |
| <a name="l00660"></a>00660 { |
| <a name="l00661"></a>00661 <span class="comment">/* Calculating the number of zeros to be padded to the output */</span> |
| <a name="l00662"></a>00662 j = srcALen - <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>; |
| <a name="l00663"></a>00663 |
| <a name="l00664"></a>00664 <span class="comment">/* Initialise the pointer after zero padding */</span> |
| <a name="l00665"></a>00665 pDst += j; |
| <a name="l00666"></a>00666 } |
| <a name="l00667"></a>00667 |
| <a name="l00668"></a>00668 <span class="keywordflow">else</span> <span class="keywordflow">if</span>(srcALen < srcBLen) |
| <a name="l00669"></a>00669 { |
| <a name="l00670"></a>00670 <span class="comment">/* Initialization to inputB pointer */</span> |
| <a name="l00671"></a>00671 pIn1 = pSrcB; |
| <a name="l00672"></a>00672 |
| <a name="l00673"></a>00673 <span class="comment">/* Initialization to the end of inputA pointer */</span> |
| <a name="l00674"></a>00674 pIn2 = pSrcA + (srcALen - 1u); |
| <a name="l00675"></a>00675 |
| <a name="l00676"></a>00676 <span class="comment">/* Initialisation of the pointer after zero padding */</span> |
| <a name="l00677"></a>00677 pDst = pDst + tot; |
| <a name="l00678"></a>00678 |
| <a name="l00679"></a>00679 <span class="comment">/* Swapping the lengths */</span> |
| <a name="l00680"></a>00680 j = <a class="code" href="arm__convolution__example__f32_8c.html#ace48ed566e2cd6a680f0681192e6af28">srcALen</a>; |
| <a name="l00681"></a>00681 srcALen = <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>; |
| <a name="l00682"></a>00682 srcBLen = j; |
| <a name="l00683"></a>00683 |
| <a name="l00684"></a>00684 <span class="comment">/* Setting the reverse flag */</span> |
| <a name="l00685"></a>00685 inv = 1; |
| <a name="l00686"></a>00686 |
| <a name="l00687"></a>00687 } |
| <a name="l00688"></a>00688 |
| <a name="l00689"></a>00689 <span class="comment">/* Loop to calculate convolution for output length number of times */</span> |
| <a name="l00690"></a>00690 <span class="keywordflow">for</span> (i = 0u; i <= tot; i++) |
| <a name="l00691"></a>00691 { |
| <a name="l00692"></a>00692 <span class="comment">/* Initialize sum with zero to carry on MAC operations */</span> |
| <a name="l00693"></a>00693 sum = 0.0f; |
| <a name="l00694"></a>00694 |
| <a name="l00695"></a>00695 <span class="comment">/* Loop to perform MAC operations according to convolution equation */</span> |
| <a name="l00696"></a>00696 <span class="keywordflow">for</span> (j = 0u; j <= i; j++) |
| <a name="l00697"></a>00697 { |
| <a name="l00698"></a>00698 <span class="comment">/* Check the array limitations */</span> |
| <a name="l00699"></a>00699 <span class="keywordflow">if</span>((((i - j) < <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>) && (j < srcALen))) |
| <a name="l00700"></a>00700 { |
| <a name="l00701"></a>00701 <span class="comment">/* z[i] += x[i-j] * y[j] */</span> |
| <a name="l00702"></a>00702 sum += pIn1[j] * pIn2[-((int32_t) i - j)]; |
| <a name="l00703"></a>00703 } |
| <a name="l00704"></a>00704 } |
| <a name="l00705"></a>00705 <span class="comment">/* Store the output in the destination buffer */</span> |
| <a name="l00706"></a>00706 <span class="keywordflow">if</span>(inv == 1) |
| <a name="l00707"></a>00707 *pDst-- = sum; |
| <a name="l00708"></a>00708 <span class="keywordflow">else</span> |
| <a name="l00709"></a>00709 *pDst++ = sum; |
| <a name="l00710"></a>00710 } |
| <a name="l00711"></a>00711 |
| <a name="l00712"></a>00712 <span class="preprocessor">#endif </span><span class="comment">/* #ifndef ARM_MATH_CM0 */</span> |
| <a name="l00713"></a>00713 |
| <a name="l00714"></a>00714 } |
| <a name="l00715"></a>00715 |
| </pre></div></div> |
| </div> |
| <!--- window showing the filter options --> |
| <div id="MSearchSelectWindow" |
| onmouseover="return searchBox.OnSearchSelectShow()" |
| onmouseout="return searchBox.OnSearchSelectHide()" |
| onkeydown="return searchBox.OnSearchSelectKey(event)"> |
| <a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(0)"><span class="SelectionMark"> </span>All</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(1)"><span class="SelectionMark"> </span>Data Structures</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(2)"><span class="SelectionMark"> </span>Files</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(3)"><span class="SelectionMark"> </span>Functions</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(4)"><span class="SelectionMark"> </span>Variables</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(5)"><span class="SelectionMark"> </span>Typedefs</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(6)"><span class="SelectionMark"> </span>Enumerations</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(7)"><span class="SelectionMark"> </span>Enumerator</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(8)"><span class="SelectionMark"> </span>Defines</a></div> |
| |
| <!-- iframe showing the search results (closed by default) --> |
| <div id="MSearchResultsWindow"> |
| <iframe src="" frameborder="0" |
| name="MSearchResults" id="MSearchResults"> |
| </iframe> |
| </div> |
| |
| <hr class="footer"/><address class="footer"><small>Generated on Fri Jul 15 2011 13:16:16 for CMSIS DSP Software Library by  |
| <a href="http://www.doxygen.org/index.html"> |
| <img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.7.2 </small></address> |
| </body> |
| </html> |
