1 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
2 <html xmlns="http://www.w3.org/1999/xhtml">
4 <meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
5 <title>CMSIS DSP Software Library: arm_conv_partial_fast_q31.c Source File</title>
6 <link href="tabs.css" rel="stylesheet" type="text/css"/>
7 <link href="search/search.css" rel="stylesheet" type="text/css"/>
8 <script type="text/javaScript" src="search/search.js"></script>
9 <link href="doxygen.css" rel="stylesheet" type="text/css"/>
11 <body onload='searchBox.OnSelectItem(0);'>
12 <!-- Generated by Doxygen 1.7.2 -->
13 <script type="text/javascript"><!--
14 var searchBox = new SearchBox("searchBox", "search",false,'Search');
16 <div class="navigation" id="top">
19 <li><a href="index.html"><span>Main Page</span></a></li>
20 <li><a href="modules.html"><span>Modules</span></a></li>
21 <li><a href="annotated.html"><span>Data Structures</span></a></li>
22 <li class="current"><a href="files.html"><span>Files</span></a></li>
23 <li><a href="examples.html"><span>Examples</span></a></li>
25 <div id="MSearchBox" class="MSearchBoxInactive">
27 <img id="MSearchSelect" src="search/mag_sel.png"
28 onmouseover="return searchBox.OnSearchSelectShow()"
29 onmouseout="return searchBox.OnSearchSelectHide()"
31 <input type="text" id="MSearchField" value="Search" accesskey="S"
32 onfocus="searchBox.OnSearchFieldFocus(true)"
33 onblur="searchBox.OnSearchFieldFocus(false)"
34 onkeyup="searchBox.OnSearchFieldChange(event)"/>
35 </span><span class="right">
36 <a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
44 <li><a href="files.html"><span>File List</span></a></li>
45 <li><a href="globals.html"><span>Globals</span></a></li>
49 <div class="headertitle">
50 <h1>arm_conv_partial_fast_q31.c</h1> </div>
52 <div class="contents">
53 <a href="arm__conv__partial__fast__q31_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>
54 <a name="l00002"></a>00002 <span class="comment">* Copyright (C) 2010 ARM Limited. All rights reserved. </span>
55 <a name="l00003"></a>00003 <span class="comment">* </span>
56 <a name="l00004"></a>00004 <span class="comment">* $Date: 15. July 2011 </span>
57 <a name="l00005"></a>00005 <span class="comment">* $Revision: V1.0.10 </span>
58 <a name="l00006"></a>00006 <span class="comment">* </span>
59 <a name="l00007"></a>00007 <span class="comment">* Project: CMSIS DSP Library </span>
60 <a name="l00008"></a>00008 <span class="comment">* Title: arm_conv_partial_fast_q31.c </span>
61 <a name="l00009"></a>00009 <span class="comment">* </span>
62 <a name="l00010"></a>00010 <span class="comment">* Description: Fast Q31 Partial convolution. </span>
63 <a name="l00011"></a>00011 <span class="comment">* </span>
64 <a name="l00012"></a>00012 <span class="comment">* Target Processor: Cortex-M4/Cortex-M3</span>
65 <a name="l00013"></a>00013 <span class="comment">* </span>
66 <a name="l00014"></a>00014 <span class="comment">* Version 1.0.10 2011/7/15 </span>
67 <a name="l00015"></a>00015 <span class="comment">* Big Endian support added and Merged M0 and M3/M4 Source code. </span>
68 <a name="l00016"></a>00016 <span class="comment">* </span>
69 <a name="l00017"></a>00017 <span class="comment">* Version 1.0.3 2010/11/29 </span>
70 <a name="l00018"></a>00018 <span class="comment">* Re-organized the CMSIS folders and updated documentation. </span>
71 <a name="l00019"></a>00019 <span class="comment">* </span>
72 <a name="l00020"></a>00020 <span class="comment">* Version 1.0.2 2010/11/11 </span>
73 <a name="l00021"></a>00021 <span class="comment">* Documentation updated. </span>
74 <a name="l00022"></a>00022 <span class="comment">* </span>
75 <a name="l00023"></a>00023 <span class="comment">* Version 1.0.1 2010/10/05 </span>
76 <a name="l00024"></a>00024 <span class="comment">* Production release and review comments incorporated. </span>
77 <a name="l00025"></a>00025 <span class="comment">* </span>
78 <a name="l00026"></a>00026 <span class="comment">* Version 1.0.0 2010/09/20 </span>
79 <a name="l00027"></a>00027 <span class="comment">* Production release and review comments incorporated. </span>
80 <a name="l00028"></a>00028 <span class="comment">* -------------------------------------------------------------------- */</span>
81 <a name="l00029"></a>00029
82 <a name="l00030"></a>00030 <span class="preprocessor">#include "<a class="code" href="arm__math_8h.html">arm_math.h</a>"</span>
83 <a name="l00031"></a>00031
84 <a name="l00056"></a><a class="code" href="group___partial_conv.html#ga10c5294cda8c4985386f4e3944be7650">00056</a> <a class="code" href="arm__math_8h.html#a5e459c6409dfcd2927bb8a57491d7cf6" title="Error status returned by some functions in the library.">arm_status</a> <a class="code" href="group___partial_conv.html#ga10c5294cda8c4985386f4e3944be7650" title="Partial convolution of Q31 sequences (fast version) for Cortex-M3 and Cortex-M4.">arm_conv_partial_fast_q31</a>(
85 <a name="l00057"></a>00057 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> * pSrcA,
86 <a name="l00058"></a>00058 uint32_t <a class="code" href="arm__convolution__example__f32_8c.html#ace48ed566e2cd6a680f0681192e6af28">srcALen</a>,
87 <a name="l00059"></a>00059 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> * pSrcB,
88 <a name="l00060"></a>00060 uint32_t <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>,
89 <a name="l00061"></a>00061 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> * pDst,
90 <a name="l00062"></a>00062 uint32_t firstIndex,
91 <a name="l00063"></a>00063 uint32_t numPoints)
92 <a name="l00064"></a>00064 {
93 <a name="l00065"></a>00065 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> *pIn1; <span class="comment">/* inputA pointer */</span>
94 <a name="l00066"></a>00066 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> *pIn2; <span class="comment">/* inputB pointer */</span>
95 <a name="l00067"></a>00067 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> *pOut = pDst; <span class="comment">/* output pointer */</span>
96 <a name="l00068"></a>00068 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> *px; <span class="comment">/* Intermediate inputA pointer */</span>
97 <a name="l00069"></a>00069 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> *py; <span class="comment">/* Intermediate inputB pointer */</span>
98 <a name="l00070"></a>00070 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> *pSrc1, *pSrc2; <span class="comment">/* Intermediate pointers */</span>
99 <a name="l00071"></a>00071 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> sum, acc0, acc1, acc2, acc3; <span class="comment">/* Accumulators */</span>
100 <a name="l00072"></a>00072 <a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a> x0, x1, x2, x3, c0;
101 <a name="l00073"></a>00073 uint32_t j, k, count, check, blkCnt;
102 <a name="l00074"></a>00074 int32_t blockSize1, blockSize2, blockSize3; <span class="comment">/* loop counters */</span>
103 <a name="l00075"></a>00075 <a class="code" href="arm__math_8h.html#a5e459c6409dfcd2927bb8a57491d7cf6" title="Error status returned by some functions in the library.">arm_status</a> <a class="code" href="arm__dotproduct__example__f32_8c.html#a88ccb294236ab22b00310c47164c53c3">status</a>; <span class="comment">/* status of Partial convolution */</span>
104 <a name="l00076"></a>00076
105 <a name="l00077"></a>00077
106 <a name="l00078"></a>00078 <span class="comment">/* Check for range of output samples to be calculated */</span>
107 <a name="l00079"></a>00079 <span class="keywordflow">if</span>((firstIndex + numPoints) > ((srcALen + (srcBLen - 1u))))
108 <a name="l00080"></a>00080 {
109 <a name="l00081"></a>00081 <span class="comment">/* Set status as ARM_MATH_ARGUMENT_ERROR */</span>
110 <a name="l00082"></a>00082 status = <a class="code" href="arm__math_8h.html#a5e459c6409dfcd2927bb8a57491d7cf6a74897e18d4b8f62b12a7d8a01dd2bb35">ARM_MATH_ARGUMENT_ERROR</a>;
111 <a name="l00083"></a>00083 }
112 <a name="l00084"></a>00084 <span class="keywordflow">else</span>
113 <a name="l00085"></a>00085 {
114 <a name="l00086"></a>00086
115 <a name="l00087"></a>00087 <span class="comment">/* The algorithm implementation is based on the lengths of the inputs. */</span>
116 <a name="l00088"></a>00088 <span class="comment">/* srcB is always made to slide across srcA. */</span>
117 <a name="l00089"></a>00089 <span class="comment">/* So srcBLen is always considered as shorter or equal to srcALen */</span>
118 <a name="l00090"></a>00090 <span class="keywordflow">if</span>(srcALen >= srcBLen)
119 <a name="l00091"></a>00091 {
120 <a name="l00092"></a>00092 <span class="comment">/* Initialization of inputA pointer */</span>
121 <a name="l00093"></a>00093 pIn1 = pSrcA;
122 <a name="l00094"></a>00094
123 <a name="l00095"></a>00095 <span class="comment">/* Initialization of inputB pointer */</span>
124 <a name="l00096"></a>00096 pIn2 = pSrcB;
125 <a name="l00097"></a>00097 }
126 <a name="l00098"></a>00098 <span class="keywordflow">else</span>
127 <a name="l00099"></a>00099 {
128 <a name="l00100"></a>00100 <span class="comment">/* Initialization of inputA pointer */</span>
129 <a name="l00101"></a>00101 pIn1 = pSrcB;
130 <a name="l00102"></a>00102
131 <a name="l00103"></a>00103 <span class="comment">/* Initialization of inputB pointer */</span>
132 <a name="l00104"></a>00104 pIn2 = pSrcA;
133 <a name="l00105"></a>00105
134 <a name="l00106"></a>00106 <span class="comment">/* srcBLen is always considered as shorter or equal to srcALen */</span>
135 <a name="l00107"></a>00107 j = <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>;
136 <a name="l00108"></a>00108 srcBLen = <a class="code" href="arm__convolution__example__f32_8c.html#ace48ed566e2cd6a680f0681192e6af28">srcALen</a>;
137 <a name="l00109"></a>00109 srcALen = j;
138 <a name="l00110"></a>00110 }
139 <a name="l00111"></a>00111
140 <a name="l00112"></a>00112 <span class="comment">/* Conditions to check which loopCounter holds </span>
141 <a name="l00113"></a>00113 <span class="comment"> * the first and last indices of the output samples to be calculated. */</span>
142 <a name="l00114"></a>00114 check = firstIndex + numPoints;
143 <a name="l00115"></a>00115 blockSize3 = ((int32_t) check - (int32_t) <a class="code" href="arm__convolution__example__f32_8c.html#ace48ed566e2cd6a680f0681192e6af28">srcALen</a>);
144 <a name="l00116"></a>00116 blockSize3 = (blockSize3 > 0) ? blockSize3 : 0;
145 <a name="l00117"></a>00117 blockSize1 = (((int32_t) srcBLen - 1) - (int32_t) firstIndex);
146 <a name="l00118"></a>00118 blockSize1 = (blockSize1 > 0) ? ((check > (srcBLen - 1u)) ? blockSize1 :
147 <a name="l00119"></a>00119 (int32_t) numPoints) : 0;
148 <a name="l00120"></a>00120 blockSize2 = (int32_t) check - ((blockSize3 + blockSize1) +
149 <a name="l00121"></a>00121 (int32_t) firstIndex);
150 <a name="l00122"></a>00122 blockSize2 = (blockSize2 > 0) ? blockSize2 : 0;
151 <a name="l00123"></a>00123
152 <a name="l00124"></a>00124 <span class="comment">/* conv(x,y) at n = x[n] * y[0] + x[n-1] * y[1] + x[n-2] * y[2] + ...+ x[n-N+1] * y[N -1] */</span>
153 <a name="l00125"></a>00125 <span class="comment">/* The function is internally </span>
154 <a name="l00126"></a>00126 <span class="comment"> * divided into three stages according to the number of multiplications that has to be </span>
155 <a name="l00127"></a>00127 <span class="comment"> * taken place between inputA samples and inputB samples. In the first stage of the </span>
156 <a name="l00128"></a>00128 <span class="comment"> * algorithm, the multiplications increase by one for every iteration. </span>
157 <a name="l00129"></a>00129 <span class="comment"> * In the second stage of the algorithm, srcBLen number of multiplications are done. </span>
158 <a name="l00130"></a>00130 <span class="comment"> * In the third stage of the algorithm, the multiplications decrease by one </span>
159 <a name="l00131"></a>00131 <span class="comment"> * for every iteration. */</span>
160 <a name="l00132"></a>00132
161 <a name="l00133"></a>00133 <span class="comment">/* Set the output pointer to point to the firstIndex </span>
162 <a name="l00134"></a>00134 <span class="comment"> * of the output sample to be calculated. */</span>
163 <a name="l00135"></a>00135 pOut = pDst + firstIndex;
164 <a name="l00136"></a>00136
165 <a name="l00137"></a>00137 <span class="comment">/* -------------------------- </span>
166 <a name="l00138"></a>00138 <span class="comment"> * Initializations of stage1 </span>
167 <a name="l00139"></a>00139 <span class="comment"> * -------------------------*/</span>
168 <a name="l00140"></a>00140
169 <a name="l00141"></a>00141 <span class="comment">/* sum = x[0] * y[0] </span>
170 <a name="l00142"></a>00142 <span class="comment"> * sum = x[0] * y[1] + x[1] * y[0] </span>
171 <a name="l00143"></a>00143 <span class="comment"> * .... </span>
172 <a name="l00144"></a>00144 <span class="comment"> * sum = x[0] * y[srcBlen - 1] + x[1] * y[srcBlen - 2] +...+ x[srcBLen - 1] * y[0] </span>
173 <a name="l00145"></a>00145 <span class="comment"> */</span>
174 <a name="l00146"></a>00146
175 <a name="l00147"></a>00147 <span class="comment">/* In this stage the MAC operations are increased by 1 for every iteration. </span>
176 <a name="l00148"></a>00148 <span class="comment"> The count variable holds the number of MAC operations performed. </span>
177 <a name="l00149"></a>00149 <span class="comment"> Since the partial convolution starts from firstIndex </span>
178 <a name="l00150"></a>00150 <span class="comment"> Number of Macs to be performed is firstIndex + 1 */</span>
179 <a name="l00151"></a>00151 count = 1u + firstIndex;
180 <a name="l00152"></a>00152
181 <a name="l00153"></a>00153 <span class="comment">/* Working pointer of inputA */</span>
182 <a name="l00154"></a>00154 px = pIn1;
183 <a name="l00155"></a>00155
184 <a name="l00156"></a>00156 <span class="comment">/* Working pointer of inputB */</span>
185 <a name="l00157"></a>00157 pSrc2 = pIn2 + firstIndex;
186 <a name="l00158"></a>00158 py = pSrc2;
187 <a name="l00159"></a>00159
188 <a name="l00160"></a>00160 <span class="comment">/* ------------------------ </span>
189 <a name="l00161"></a>00161 <span class="comment"> * Stage1 process </span>
190 <a name="l00162"></a>00162 <span class="comment"> * ----------------------*/</span>
191 <a name="l00163"></a>00163
192 <a name="l00164"></a>00164 <span class="comment">/* The first loop starts here */</span>
193 <a name="l00165"></a>00165 <span class="keywordflow">while</span>(blockSize1 > 0)
194 <a name="l00166"></a>00166 {
195 <a name="l00167"></a>00167 <span class="comment">/* Accumulator is made zero for every iteration */</span>
196 <a name="l00168"></a>00168 sum = 0;
197 <a name="l00169"></a>00169
198 <a name="l00170"></a>00170 <span class="comment">/* Apply loop unrolling and compute 4 MACs simultaneously. */</span>
199 <a name="l00171"></a>00171 k = count >> 2u;
200 <a name="l00172"></a>00172
201 <a name="l00173"></a>00173 <span class="comment">/* First part of the processing with loop unrolling. Compute 4 MACs at a time. </span>
202 <a name="l00174"></a>00174 <span class="comment"> ** a second loop below computes MACs for the remaining 1 to 3 samples. */</span>
203 <a name="l00175"></a>00175 <span class="keywordflow">while</span>(k > 0u)
204 <a name="l00176"></a>00176 {
205 <a name="l00177"></a>00177 <span class="comment">/* x[0] * y[srcBLen - 1] */</span>
206 <a name="l00178"></a>00178 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
207 <a name="l00179"></a>00179 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
208 <a name="l00180"></a>00180
209 <a name="l00181"></a>00181 <span class="comment">/* x[1] * y[srcBLen - 2] */</span>
210 <a name="l00182"></a>00182 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
211 <a name="l00183"></a>00183 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
212 <a name="l00184"></a>00184
213 <a name="l00185"></a>00185 <span class="comment">/* x[2] * y[srcBLen - 3] */</span>
214 <a name="l00186"></a>00186 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
215 <a name="l00187"></a>00187 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
216 <a name="l00188"></a>00188
217 <a name="l00189"></a>00189 <span class="comment">/* x[3] * y[srcBLen - 4] */</span>
218 <a name="l00190"></a>00190 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
219 <a name="l00191"></a>00191 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
220 <a name="l00192"></a>00192
221 <a name="l00193"></a>00193 <span class="comment">/* Decrement the loop counter */</span>
222 <a name="l00194"></a>00194 k--;
223 <a name="l00195"></a>00195 }
224 <a name="l00196"></a>00196
225 <a name="l00197"></a>00197 <span class="comment">/* If the count is not a multiple of 4, compute any remaining MACs here. </span>
226 <a name="l00198"></a>00198 <span class="comment"> ** No loop unrolling is used. */</span>
227 <a name="l00199"></a>00199 k = count % 0x4u;
228 <a name="l00200"></a>00200
229 <a name="l00201"></a>00201 <span class="keywordflow">while</span>(k > 0u)
230 <a name="l00202"></a>00202 {
231 <a name="l00203"></a>00203 <span class="comment">/* Perform the multiply-accumulates */</span>
232 <a name="l00204"></a>00204 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
233 <a name="l00205"></a>00205 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
234 <a name="l00206"></a>00206
235 <a name="l00207"></a>00207 <span class="comment">/* Decrement the loop counter */</span>
236 <a name="l00208"></a>00208 k--;
237 <a name="l00209"></a>00209 }
238 <a name="l00210"></a>00210
239 <a name="l00211"></a>00211 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span>
240 <a name="l00212"></a>00212 *pOut++ = sum << 1;
241 <a name="l00213"></a>00213
242 <a name="l00214"></a>00214 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span>
243 <a name="l00215"></a>00215 py = ++pSrc2;
244 <a name="l00216"></a>00216 px = pIn1;
245 <a name="l00217"></a>00217
246 <a name="l00218"></a>00218 <span class="comment">/* Increment the MAC count */</span>
247 <a name="l00219"></a>00219 count++;
248 <a name="l00220"></a>00220
249 <a name="l00221"></a>00221 <span class="comment">/* Decrement the loop counter */</span>
250 <a name="l00222"></a>00222 blockSize1--;
251 <a name="l00223"></a>00223 }
252 <a name="l00224"></a>00224
253 <a name="l00225"></a>00225 <span class="comment">/* -------------------------- </span>
254 <a name="l00226"></a>00226 <span class="comment"> * Initializations of stage2 </span>
255 <a name="l00227"></a>00227 <span class="comment"> * ------------------------*/</span>
256 <a name="l00228"></a>00228
257 <a name="l00229"></a>00229 <span class="comment">/* sum = x[0] * y[srcBLen-1] + x[1] * y[srcBLen-2] +...+ x[srcBLen-1] * y[0] </span>
258 <a name="l00230"></a>00230 <span class="comment"> * sum = x[1] * y[srcBLen-1] + x[2] * y[srcBLen-2] +...+ x[srcBLen] * y[0] </span>
259 <a name="l00231"></a>00231 <span class="comment"> * .... </span>
260 <a name="l00232"></a>00232 <span class="comment"> * sum = x[srcALen-srcBLen-2] * y[srcBLen-1] + x[srcALen] * y[srcBLen-2] +...+ x[srcALen-1] * y[0] </span>
261 <a name="l00233"></a>00233 <span class="comment"> */</span>
262 <a name="l00234"></a>00234
263 <a name="l00235"></a>00235 <span class="comment">/* Working pointer of inputA */</span>
264 <a name="l00236"></a>00236 px = pIn1;
265 <a name="l00237"></a>00237
266 <a name="l00238"></a>00238 <span class="comment">/* Working pointer of inputB */</span>
267 <a name="l00239"></a>00239 pSrc2 = pIn2 + (srcBLen - 1u);
268 <a name="l00240"></a>00240 py = pSrc2;
269 <a name="l00241"></a>00241
270 <a name="l00242"></a>00242 <span class="comment">/* count is index by which the pointer pIn1 to be incremented */</span>
271 <a name="l00243"></a>00243 count = 1u;
272 <a name="l00244"></a>00244
273 <a name="l00245"></a>00245 <span class="comment">/* ------------------- </span>
274 <a name="l00246"></a>00246 <span class="comment"> * Stage2 process </span>
275 <a name="l00247"></a>00247 <span class="comment"> * ------------------*/</span>
276 <a name="l00248"></a>00248
277 <a name="l00249"></a>00249 <span class="comment">/* Stage2 depends on srcBLen as in this stage srcBLen number of MACS are performed. </span>
278 <a name="l00250"></a>00250 <span class="comment"> * So, to loop unroll over blockSize2, </span>
279 <a name="l00251"></a>00251 <span class="comment"> * srcBLen should be greater than or equal to 4 */</span>
280 <a name="l00252"></a>00252 <span class="keywordflow">if</span>(srcBLen >= 4u)
281 <a name="l00253"></a>00253 {
282 <a name="l00254"></a>00254 <span class="comment">/* Loop unroll over blockSize2 */</span>
283 <a name="l00255"></a>00255 blkCnt = ((uint32_t) blockSize2 >> 2u);
284 <a name="l00256"></a>00256
285 <a name="l00257"></a>00257 <span class="keywordflow">while</span>(blkCnt > 0u)
286 <a name="l00258"></a>00258 {
287 <a name="l00259"></a>00259 <span class="comment">/* Set all accumulators to zero */</span>
288 <a name="l00260"></a>00260 acc0 = 0;
289 <a name="l00261"></a>00261 acc1 = 0;
290 <a name="l00262"></a>00262 acc2 = 0;
291 <a name="l00263"></a>00263 acc3 = 0;
292 <a name="l00264"></a>00264
293 <a name="l00265"></a>00265 <span class="comment">/* read x[0], x[1], x[2] samples */</span>
294 <a name="l00266"></a>00266 x0 = *(px++);
295 <a name="l00267"></a>00267 x1 = *(px++);
296 <a name="l00268"></a>00268 x2 = *(px++);
297 <a name="l00269"></a>00269
298 <a name="l00270"></a>00270 <span class="comment">/* Apply loop unrolling and compute 4 MACs simultaneously. */</span>
299 <a name="l00271"></a>00271 k = srcBLen >> 2u;
300 <a name="l00272"></a>00272
301 <a name="l00273"></a>00273 <span class="comment">/* First part of the processing with loop unrolling. Compute 4 MACs at a time. </span>
302 <a name="l00274"></a>00274 <span class="comment"> ** a second loop below computes MACs for the remaining 1 to 3 samples. */</span>
303 <a name="l00275"></a>00275 <span class="keywordflow">do</span>
304 <a name="l00276"></a>00276 {
305 <a name="l00277"></a>00277 <span class="comment">/* Read y[srcBLen - 1] sample */</span>
306 <a name="l00278"></a>00278 c0 = *(py--);
307 <a name="l00279"></a>00279
308 <a name="l00280"></a>00280 <span class="comment">/* Read x[3] sample */</span>
309 <a name="l00281"></a>00281 x3 = *(px++);
310 <a name="l00282"></a>00282
311 <a name="l00283"></a>00283 <span class="comment">/* Perform the multiply-accumulate */</span>
312 <a name="l00284"></a>00284 <span class="comment">/* acc0 += x[0] * y[srcBLen - 1] */</span>
313 <a name="l00285"></a>00285 acc0 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc0 << 32) + ((q63_t) x0 * c0)) >> 32);
314 <a name="l00286"></a>00286
315 <a name="l00287"></a>00287 <span class="comment">/* acc1 += x[1] * y[srcBLen - 1] */</span>
316 <a name="l00288"></a>00288 acc1 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc1 << 32) + ((q63_t) x1 * c0)) >> 32);
317 <a name="l00289"></a>00289
318 <a name="l00290"></a>00290 <span class="comment">/* acc2 += x[2] * y[srcBLen - 1] */</span>
319 <a name="l00291"></a>00291 acc2 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc2 << 32) + ((q63_t) x2 * c0)) >> 32);
320 <a name="l00292"></a>00292
321 <a name="l00293"></a>00293 <span class="comment">/* acc3 += x[3] * y[srcBLen - 1] */</span>
322 <a name="l00294"></a>00294 acc3 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc3 << 32) + ((q63_t) x3 * c0)) >> 32);
323 <a name="l00295"></a>00295
324 <a name="l00296"></a>00296 <span class="comment">/* Read y[srcBLen - 2] sample */</span>
325 <a name="l00297"></a>00297 c0 = *(py--);
326 <a name="l00298"></a>00298
327 <a name="l00299"></a>00299 <span class="comment">/* Read x[4] sample */</span>
328 <a name="l00300"></a>00300 x0 = *(px++);
329 <a name="l00301"></a>00301
330 <a name="l00302"></a>00302 <span class="comment">/* Perform the multiply-accumulate */</span>
331 <a name="l00303"></a>00303 <span class="comment">/* acc0 += x[1] * y[srcBLen - 2] */</span>
332 <a name="l00304"></a>00304 acc0 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc0 << 32) + ((q63_t) x1 * c0)) >> 32);
333 <a name="l00305"></a>00305 <span class="comment">/* acc1 += x[2] * y[srcBLen - 2] */</span>
334 <a name="l00306"></a>00306 acc1 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc1 << 32) + ((q63_t) x2 * c0)) >> 32);
335 <a name="l00307"></a>00307 <span class="comment">/* acc2 += x[3] * y[srcBLen - 2] */</span>
336 <a name="l00308"></a>00308 acc2 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc2 << 32) + ((q63_t) x3 * c0)) >> 32);
337 <a name="l00309"></a>00309 <span class="comment">/* acc3 += x[4] * y[srcBLen - 2] */</span>
338 <a name="l00310"></a>00310 acc3 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc3 << 32) + ((q63_t) x0 * c0)) >> 32);
339 <a name="l00311"></a>00311
340 <a name="l00312"></a>00312 <span class="comment">/* Read y[srcBLen - 3] sample */</span>
341 <a name="l00313"></a>00313 c0 = *(py--);
342 <a name="l00314"></a>00314
343 <a name="l00315"></a>00315 <span class="comment">/* Read x[5] sample */</span>
344 <a name="l00316"></a>00316 x1 = *(px++);
345 <a name="l00317"></a>00317
346 <a name="l00318"></a>00318 <span class="comment">/* Perform the multiply-accumulates */</span>
347 <a name="l00319"></a>00319 <span class="comment">/* acc0 += x[2] * y[srcBLen - 3] */</span>
348 <a name="l00320"></a>00320 acc0 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc0 << 32) + ((q63_t) x2 * c0)) >> 32);
349 <a name="l00321"></a>00321 <span class="comment">/* acc1 += x[3] * y[srcBLen - 2] */</span>
350 <a name="l00322"></a>00322 acc1 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc1 << 32) + ((q63_t) x3 * c0)) >> 32);
351 <a name="l00323"></a>00323 <span class="comment">/* acc2 += x[4] * y[srcBLen - 2] */</span>
352 <a name="l00324"></a>00324 acc2 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc2 << 32) + ((q63_t) x0 * c0)) >> 32);
353 <a name="l00325"></a>00325 <span class="comment">/* acc3 += x[5] * y[srcBLen - 2] */</span>
354 <a name="l00326"></a>00326 acc3 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc3 << 32) + ((q63_t) x1 * c0)) >> 32);
355 <a name="l00327"></a>00327
356 <a name="l00328"></a>00328 <span class="comment">/* Read y[srcBLen - 4] sample */</span>
357 <a name="l00329"></a>00329 c0 = *(py--);
358 <a name="l00330"></a>00330
359 <a name="l00331"></a>00331 <span class="comment">/* Read x[6] sample */</span>
360 <a name="l00332"></a>00332 x2 = *(px++);
361 <a name="l00333"></a>00333
362 <a name="l00334"></a>00334 <span class="comment">/* Perform the multiply-accumulates */</span>
363 <a name="l00335"></a>00335 <span class="comment">/* acc0 += x[3] * y[srcBLen - 4] */</span>
364 <a name="l00336"></a>00336 acc0 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc0 << 32) + ((q63_t) x3 * c0)) >> 32);
365 <a name="l00337"></a>00337 <span class="comment">/* acc1 += x[4] * y[srcBLen - 4] */</span>
366 <a name="l00338"></a>00338 acc1 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc1 << 32) + ((q63_t) x0 * c0)) >> 32);
367 <a name="l00339"></a>00339 <span class="comment">/* acc2 += x[5] * y[srcBLen - 4] */</span>
368 <a name="l00340"></a>00340 acc2 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc2 << 32) + ((q63_t) x1 * c0)) >> 32);
369 <a name="l00341"></a>00341 <span class="comment">/* acc3 += x[6] * y[srcBLen - 4] */</span>
370 <a name="l00342"></a>00342 acc3 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc3 << 32) + ((q63_t) x2 * c0)) >> 32);
371 <a name="l00343"></a>00343
372 <a name="l00344"></a>00344
373 <a name="l00345"></a>00345 } <span class="keywordflow">while</span>(--k);
374 <a name="l00346"></a>00346
375 <a name="l00347"></a>00347 <span class="comment">/* If the srcBLen is not a multiple of 4, compute any remaining MACs here. </span>
376 <a name="l00348"></a>00348 <span class="comment"> ** No loop unrolling is used. */</span>
377 <a name="l00349"></a>00349 k = srcBLen % 0x4u;
378 <a name="l00350"></a>00350
379 <a name="l00351"></a>00351 <span class="keywordflow">while</span>(k > 0u)
380 <a name="l00352"></a>00352 {
381 <a name="l00353"></a>00353 <span class="comment">/* Read y[srcBLen - 5] sample */</span>
382 <a name="l00354"></a>00354 c0 = *(py--);
383 <a name="l00355"></a>00355
384 <a name="l00356"></a>00356 <span class="comment">/* Read x[7] sample */</span>
385 <a name="l00357"></a>00357 x3 = *(px++);
386 <a name="l00358"></a>00358
387 <a name="l00359"></a>00359 <span class="comment">/* Perform the multiply-accumulates */</span>
388 <a name="l00360"></a>00360 <span class="comment">/* acc0 += x[4] * y[srcBLen - 5] */</span>
389 <a name="l00361"></a>00361 acc0 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc0 << 32) + ((q63_t) x0 * c0)) >> 32);
390 <a name="l00362"></a>00362 <span class="comment">/* acc1 += x[5] * y[srcBLen - 5] */</span>
391 <a name="l00363"></a>00363 acc1 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc1 << 32) + ((q63_t) x1 * c0)) >> 32);
392 <a name="l00364"></a>00364 <span class="comment">/* acc2 += x[6] * y[srcBLen - 5] */</span>
393 <a name="l00365"></a>00365 acc2 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc2 << 32) + ((q63_t) x2 * c0)) >> 32);
394 <a name="l00366"></a>00366 <span class="comment">/* acc3 += x[7] * y[srcBLen - 5] */</span>
395 <a name="l00367"></a>00367 acc3 = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) acc3 << 32) + ((q63_t) x3 * c0)) >> 32);
396 <a name="l00368"></a>00368
397 <a name="l00369"></a>00369 <span class="comment">/* Reuse the present samples for the next MAC */</span>
398 <a name="l00370"></a>00370 x0 = x1;
399 <a name="l00371"></a>00371 x1 = x2;
400 <a name="l00372"></a>00372 x2 = x3;
401 <a name="l00373"></a>00373
402 <a name="l00374"></a>00374 <span class="comment">/* Decrement the loop counter */</span>
403 <a name="l00375"></a>00375 k--;
404 <a name="l00376"></a>00376 }
405 <a name="l00377"></a>00377
406 <a name="l00378"></a>00378 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span>
407 <a name="l00379"></a>00379 *pOut++ = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) (acc0 << 1);
408 <a name="l00380"></a>00380 *pOut++ = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) (acc1 << 1);
409 <a name="l00381"></a>00381 *pOut++ = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) (acc2 << 1);
410 <a name="l00382"></a>00382 *pOut++ = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) (acc3 << 1);
411 <a name="l00383"></a>00383
412 <a name="l00384"></a>00384 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span>
413 <a name="l00385"></a>00385 px = pIn1 + (count * 4u);
414 <a name="l00386"></a>00386 py = pSrc2;
415 <a name="l00387"></a>00387
416 <a name="l00388"></a>00388 <span class="comment">/* Increment the pointer pIn1 index, count by 1 */</span>
417 <a name="l00389"></a>00389 count++;
418 <a name="l00390"></a>00390
419 <a name="l00391"></a>00391 <span class="comment">/* Decrement the loop counter */</span>
420 <a name="l00392"></a>00392 blkCnt--;
421 <a name="l00393"></a>00393 }
422 <a name="l00394"></a>00394
423 <a name="l00395"></a>00395 <span class="comment">/* If the blockSize2 is not a multiple of 4, compute any remaining output samples here. </span>
424 <a name="l00396"></a>00396 <span class="comment"> ** No loop unrolling is used. */</span>
425 <a name="l00397"></a>00397 blkCnt = (uint32_t) blockSize2 % 0x4u;
426 <a name="l00398"></a>00398
427 <a name="l00399"></a>00399 <span class="keywordflow">while</span>(blkCnt > 0u)
428 <a name="l00400"></a>00400 {
429 <a name="l00401"></a>00401 <span class="comment">/* Accumulator is made zero for every iteration */</span>
430 <a name="l00402"></a>00402 sum = 0;
431 <a name="l00403"></a>00403
432 <a name="l00404"></a>00404 <span class="comment">/* Apply loop unrolling and compute 4 MACs simultaneously. */</span>
433 <a name="l00405"></a>00405 k = srcBLen >> 2u;
434 <a name="l00406"></a>00406
435 <a name="l00407"></a>00407 <span class="comment">/* First part of the processing with loop unrolling. Compute 4 MACs at a time. </span>
436 <a name="l00408"></a>00408 <span class="comment"> ** a second loop below computes MACs for the remaining 1 to 3 samples. */</span>
437 <a name="l00409"></a>00409 <span class="keywordflow">while</span>(k > 0u)
438 <a name="l00410"></a>00410 {
439 <a name="l00411"></a>00411 <span class="comment">/* Perform the multiply-accumulates */</span>
440 <a name="l00412"></a>00412 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
441 <a name="l00413"></a>00413 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
442 <a name="l00414"></a>00414 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
443 <a name="l00415"></a>00415 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
444 <a name="l00416"></a>00416 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
445 <a name="l00417"></a>00417 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
446 <a name="l00418"></a>00418 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
447 <a name="l00419"></a>00419 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
448 <a name="l00420"></a>00420
449 <a name="l00421"></a>00421 <span class="comment">/* Decrement the loop counter */</span>
450 <a name="l00422"></a>00422 k--;
451 <a name="l00423"></a>00423 }
452 <a name="l00424"></a>00424
453 <a name="l00425"></a>00425 <span class="comment">/* If the srcBLen is not a multiple of 4, compute any remaining MACs here. </span>
454 <a name="l00426"></a>00426 <span class="comment"> ** No loop unrolling is used. */</span>
455 <a name="l00427"></a>00427 k = srcBLen % 0x4u;
456 <a name="l00428"></a>00428
457 <a name="l00429"></a>00429 <span class="keywordflow">while</span>(k > 0u)
458 <a name="l00430"></a>00430 {
459 <a name="l00431"></a>00431 <span class="comment">/* Perform the multiply-accumulate */</span>
460 <a name="l00432"></a>00432 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
461 <a name="l00433"></a>00433 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
462 <a name="l00434"></a>00434
463 <a name="l00435"></a>00435 <span class="comment">/* Decrement the loop counter */</span>
464 <a name="l00436"></a>00436 k--;
465 <a name="l00437"></a>00437 }
466 <a name="l00438"></a>00438
467 <a name="l00439"></a>00439 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span>
468 <a name="l00440"></a>00440 *pOut++ = sum << 1;
469 <a name="l00441"></a>00441
470 <a name="l00442"></a>00442 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span>
471 <a name="l00443"></a>00443 px = pIn1 + count;
472 <a name="l00444"></a>00444 py = pSrc2;
473 <a name="l00445"></a>00445
474 <a name="l00446"></a>00446 <span class="comment">/* Increment the MAC count */</span>
475 <a name="l00447"></a>00447 count++;
476 <a name="l00448"></a>00448
477 <a name="l00449"></a>00449 <span class="comment">/* Decrement the loop counter */</span>
478 <a name="l00450"></a>00450 blkCnt--;
479 <a name="l00451"></a>00451 }
480 <a name="l00452"></a>00452 }
481 <a name="l00453"></a>00453 <span class="keywordflow">else</span>
482 <a name="l00454"></a>00454 {
483 <a name="l00455"></a>00455 <span class="comment">/* If the srcBLen is not a multiple of 4, </span>
484 <a name="l00456"></a>00456 <span class="comment"> * the blockSize2 loop cannot be unrolled by 4 */</span>
485 <a name="l00457"></a>00457 blkCnt = (uint32_t) blockSize2;
486 <a name="l00458"></a>00458
487 <a name="l00459"></a>00459 <span class="keywordflow">while</span>(blkCnt > 0u)
488 <a name="l00460"></a>00460 {
489 <a name="l00461"></a>00461 <span class="comment">/* Accumulator is made zero for every iteration */</span>
490 <a name="l00462"></a>00462 sum = 0;
491 <a name="l00463"></a>00463
492 <a name="l00464"></a>00464 <span class="comment">/* srcBLen number of MACS should be performed */</span>
493 <a name="l00465"></a>00465 k = <a class="code" href="arm__convolution__example__f32_8c.html#aea71286f498978c5ed3775609b974fc8">srcBLen</a>;
494 <a name="l00466"></a>00466
495 <a name="l00467"></a>00467 <span class="keywordflow">while</span>(k > 0u)
496 <a name="l00468"></a>00468 {
497 <a name="l00469"></a>00469 <span class="comment">/* Perform the multiply-accumulate */</span>
498 <a name="l00470"></a>00470 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
499 <a name="l00471"></a>00471 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
500 <a name="l00472"></a>00472
501 <a name="l00473"></a>00473 <span class="comment">/* Decrement the loop counter */</span>
502 <a name="l00474"></a>00474 k--;
503 <a name="l00475"></a>00475 }
504 <a name="l00476"></a>00476
505 <a name="l00477"></a>00477 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span>
506 <a name="l00478"></a>00478 *pOut++ = sum << 1;
507 <a name="l00479"></a>00479
508 <a name="l00480"></a>00480 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span>
509 <a name="l00481"></a>00481 px = pIn1 + count;
510 <a name="l00482"></a>00482 py = pSrc2;
511 <a name="l00483"></a>00483
512 <a name="l00484"></a>00484 <span class="comment">/* Increment the MAC count */</span>
513 <a name="l00485"></a>00485 count++;
514 <a name="l00486"></a>00486
515 <a name="l00487"></a>00487 <span class="comment">/* Decrement the loop counter */</span>
516 <a name="l00488"></a>00488 blkCnt--;
517 <a name="l00489"></a>00489 }
518 <a name="l00490"></a>00490 }
519 <a name="l00491"></a>00491
520 <a name="l00492"></a>00492
521 <a name="l00493"></a>00493 <span class="comment">/* -------------------------- </span>
522 <a name="l00494"></a>00494 <span class="comment"> * Initializations of stage3 </span>
523 <a name="l00495"></a>00495 <span class="comment"> * -------------------------*/</span>
524 <a name="l00496"></a>00496
525 <a name="l00497"></a>00497 <span class="comment">/* sum += x[srcALen-srcBLen+1] * y[srcBLen-1] + x[srcALen-srcBLen+2] * y[srcBLen-2] +...+ x[srcALen-1] * y[1] </span>
526 <a name="l00498"></a>00498 <span class="comment"> * sum += x[srcALen-srcBLen+2] * y[srcBLen-1] + x[srcALen-srcBLen+3] * y[srcBLen-2] +...+ x[srcALen-1] * y[2] </span>
527 <a name="l00499"></a>00499 <span class="comment"> * .... </span>
528 <a name="l00500"></a>00500 <span class="comment"> * sum += x[srcALen-2] * y[srcBLen-1] + x[srcALen-1] * y[srcBLen-2] </span>
529 <a name="l00501"></a>00501 <span class="comment"> * sum += x[srcALen-1] * y[srcBLen-1] </span>
530 <a name="l00502"></a>00502 <span class="comment"> */</span>
531 <a name="l00503"></a>00503
532 <a name="l00504"></a>00504 <span class="comment">/* In this stage the MAC operations are decreased by 1 for every iteration. </span>
533 <a name="l00505"></a>00505 <span class="comment"> The count variable holds the number of MAC operations performed */</span>
534 <a name="l00506"></a>00506 count = srcBLen - 1u;
535 <a name="l00507"></a>00507
536 <a name="l00508"></a>00508 <span class="comment">/* Working pointer of inputA */</span>
537 <a name="l00509"></a>00509 pSrc1 = (pIn1 + <a class="code" href="arm__convolution__example__f32_8c.html#ace48ed566e2cd6a680f0681192e6af28">srcALen</a>) - (srcBLen - 1u);
538 <a name="l00510"></a>00510 px = pSrc1;
539 <a name="l00511"></a>00511
540 <a name="l00512"></a>00512 <span class="comment">/* Working pointer of inputB */</span>
541 <a name="l00513"></a>00513 pSrc2 = pIn2 + (srcBLen - 1u);
542 <a name="l00514"></a>00514 py = pSrc2;
543 <a name="l00515"></a>00515
544 <a name="l00516"></a>00516 <span class="comment">/* ------------------- </span>
545 <a name="l00517"></a>00517 <span class="comment"> * Stage3 process </span>
546 <a name="l00518"></a>00518 <span class="comment"> * ------------------*/</span>
547 <a name="l00519"></a>00519
548 <a name="l00520"></a>00520 <span class="keywordflow">while</span>(blockSize3 > 0)
549 <a name="l00521"></a>00521 {
550 <a name="l00522"></a>00522 <span class="comment">/* Accumulator is made zero for every iteration */</span>
551 <a name="l00523"></a>00523 sum = 0;
552 <a name="l00524"></a>00524
553 <a name="l00525"></a>00525 <span class="comment">/* Apply loop unrolling and compute 4 MACs simultaneously. */</span>
554 <a name="l00526"></a>00526 k = count >> 2u;
555 <a name="l00527"></a>00527
556 <a name="l00528"></a>00528 <span class="comment">/* First part of the processing with loop unrolling. Compute 4 MACs at a time. </span>
557 <a name="l00529"></a>00529 <span class="comment"> ** a second loop below computes MACs for the remaining 1 to 3 samples. */</span>
558 <a name="l00530"></a>00530 <span class="keywordflow">while</span>(k > 0u)
559 <a name="l00531"></a>00531 {
560 <a name="l00532"></a>00532 <span class="comment">/* sum += x[srcALen - srcBLen + 1] * y[srcBLen - 1] */</span>
561 <a name="l00533"></a>00533 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
562 <a name="l00534"></a>00534 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
563 <a name="l00535"></a>00535
564 <a name="l00536"></a>00536 <span class="comment">/* sum += x[srcALen - srcBLen + 2] * y[srcBLen - 2] */</span>
565 <a name="l00537"></a>00537 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
566 <a name="l00538"></a>00538 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
567 <a name="l00539"></a>00539
568 <a name="l00540"></a>00540 <span class="comment">/* sum += x[srcALen - srcBLen + 3] * y[srcBLen - 3] */</span>
569 <a name="l00541"></a>00541 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
570 <a name="l00542"></a>00542 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
571 <a name="l00543"></a>00543
572 <a name="l00544"></a>00544 <span class="comment">/* sum += x[srcALen - srcBLen + 4] * y[srcBLen - 4] */</span>
573 <a name="l00545"></a>00545 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
574 <a name="l00546"></a>00546 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
575 <a name="l00547"></a>00547
576 <a name="l00548"></a>00548 <span class="comment">/* Decrement the loop counter */</span>
577 <a name="l00549"></a>00549 k--;
578 <a name="l00550"></a>00550 }
579 <a name="l00551"></a>00551
580 <a name="l00552"></a>00552 <span class="comment">/* If the count is not a multiple of 4, compute any remaining MACs here. </span>
581 <a name="l00553"></a>00553 <span class="comment"> ** No loop unrolling is used. */</span>
582 <a name="l00554"></a>00554 k = count % 0x4u;
583 <a name="l00555"></a>00555
584 <a name="l00556"></a>00556 <span class="keywordflow">while</span>(k > 0u)
585 <a name="l00557"></a>00557 {
586 <a name="l00558"></a>00558 <span class="comment">/* Perform the multiply-accumulates */</span>
587 <a name="l00559"></a>00559 <span class="comment">/* sum += x[srcALen-1] * y[srcBLen-1] */</span>
588 <a name="l00560"></a>00560 sum = (<a class="code" href="arm__math_8h.html#adc89a3547f5324b7b3b95adec3806bc0" title="32-bit fractional data type in 1.31 format.">q31_t</a>) ((((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) sum << 32) +
589 <a name="l00561"></a>00561 ((<a class="code" href="arm__math_8h.html#a5aea1cb12fc02d9d44c8abf217eaa5c6" title="64-bit fractional data type in 1.63 format.">q63_t</a>) * px++ * (*py--))) >> 32);
590 <a name="l00562"></a>00562
591 <a name="l00563"></a>00563 <span class="comment">/* Decrement the loop counter */</span>
592 <a name="l00564"></a>00564 k--;
593 <a name="l00565"></a>00565 }
594 <a name="l00566"></a>00566
595 <a name="l00567"></a>00567 <span class="comment">/* Store the result in the accumulator in the destination buffer. */</span>
596 <a name="l00568"></a>00568 *pOut++ = sum << 1;
597 <a name="l00569"></a>00569
598 <a name="l00570"></a>00570 <span class="comment">/* Update the inputA and inputB pointers for next MAC calculation */</span>
599 <a name="l00571"></a>00571 px = ++pSrc1;
600 <a name="l00572"></a>00572 py = pSrc2;
601 <a name="l00573"></a>00573
602 <a name="l00574"></a>00574 <span class="comment">/* Decrement the MAC count */</span>
603 <a name="l00575"></a>00575 count--;
604 <a name="l00576"></a>00576
605 <a name="l00577"></a>00577 <span class="comment">/* Decrement the loop counter */</span>
606 <a name="l00578"></a>00578 blockSize3--;
607 <a name="l00579"></a>00579
608 <a name="l00580"></a>00580 }
609 <a name="l00581"></a>00581
610 <a name="l00582"></a>00582 <span class="comment">/* set status as ARM_MATH_SUCCESS */</span>
611 <a name="l00583"></a>00583 status = <a class="code" href="arm__math_8h.html#a5e459c6409dfcd2927bb8a57491d7cf6a9f8b2a10bd827fb4600e77d455902eb0">ARM_MATH_SUCCESS</a>;
612 <a name="l00584"></a>00584 }
613 <a name="l00585"></a>00585
614 <a name="l00586"></a>00586 <span class="comment">/* Return to application */</span>
615 <a name="l00587"></a>00587 <span class="keywordflow">return</span> (status);
616 <a name="l00588"></a>00588
617 <a name="l00589"></a>00589 }
618 <a name="l00590"></a>00590
621 <!--- window showing the filter options -->
622 <div id="MSearchSelectWindow"
623 onmouseover="return searchBox.OnSearchSelectShow()"
624 onmouseout="return searchBox.OnSearchSelectHide()"
625 onkeydown="return searchBox.OnSearchSelectKey(event)">
626 <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>
628 <!-- iframe showing the search results (closed by default) -->
629 <div id="MSearchResultsWindow">
630 <iframe src="" frameborder="0"
631 name="MSearchResults" id="MSearchResults">
635 <hr class="footer"/><address class="footer"><small>Generated on Fri Jul 15 2011 13:16:16 for CMSIS DSP Software Library by 
636 <a href="http://www.doxygen.org/index.html">
637 <img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.7.2 </small></address>
projects / fw / stlink / blob