mirror of
https://github.com/ml-explore/mlx.git
synced 2025-09-22 13:28:11 +08:00
rebase
This commit is contained in:
124
docs/build/html/scan_8h_source.html
vendored
124
docs/build/html/scan_8h_source.html
vendored
@@ -3,7 +3,7 @@
|
||||
<head>
|
||||
<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
|
||||
<meta http-equiv="X-UA-Compatible" content="IE=11"/>
|
||||
<meta name="generator" content="Doxygen 1.12.0"/>
|
||||
<meta name="generator" content="Doxygen 1.13.1"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1"/>
|
||||
<title>MLX: mlx/backend/metal/kernels/scan.h Source File</title>
|
||||
<link href="tabs.css" rel="stylesheet" type="text/css"/>
|
||||
@@ -11,11 +11,18 @@
|
||||
<script type="text/javascript" src="dynsections.js"></script>
|
||||
<script type="text/javascript" src="clipboard.js"></script>
|
||||
<link href="navtree.css" rel="stylesheet" type="text/css"/>
|
||||
<script type="text/javascript" src="navtreedata.js"></script>
|
||||
<script type="text/javascript" src="navtree.js"></script>
|
||||
<script type="text/javascript" src="resize.js"></script>
|
||||
<script type="text/javascript" src="cookie.js"></script>
|
||||
<link href="search/search.css" rel="stylesheet" type="text/css"/>
|
||||
<script type="text/javascript" src="search/searchdata.js"></script>
|
||||
<script type="text/javascript" src="search/search.js"></script>
|
||||
<script type="text/javascript">
|
||||
/* @license magnet:?xt=urn:btih:d3d9a9a6595521f9666a5e94cc830dab83b65699&dn=expat.txt MIT */
|
||||
$(function() { init_search(); });
|
||||
/* @license-end */
|
||||
</script>
|
||||
<link href="doxygen.css" rel="stylesheet" type="text/css" />
|
||||
</head>
|
||||
<body>
|
||||
@@ -28,12 +35,24 @@
|
||||
<div id="projectname">MLX
|
||||
</div>
|
||||
</td>
|
||||
<td> <div id="MSearchBox" class="MSearchBoxInactive">
|
||||
<span class="left">
|
||||
<span id="MSearchSelect" onmouseover="return searchBox.OnSearchSelectShow()" onmouseout="return searchBox.OnSearchSelectHide()"> </span>
|
||||
<input type="text" id="MSearchField" value="" placeholder="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.svg" alt=""/></a>
|
||||
</span>
|
||||
</div>
|
||||
</td>
|
||||
</tr>
|
||||
</tbody>
|
||||
</table>
|
||||
</div>
|
||||
<!-- end header part -->
|
||||
<!-- Generated by Doxygen 1.12.0 -->
|
||||
<!-- Generated by Doxygen 1.13.1 -->
|
||||
<script type="text/javascript">
|
||||
/* @license magnet:?xt=urn:btih:d3d9a9a6595521f9666a5e94cc830dab83b65699&dn=expat.txt MIT */
|
||||
var searchBox = new SearchBox("searchBox", "search/",'.html');
|
||||
@@ -44,22 +63,23 @@ var searchBox = new SearchBox("searchBox", "search/",'.html');
|
||||
$(function() { codefold.init(0); });
|
||||
/* @license-end */
|
||||
</script>
|
||||
<script type="text/javascript" src="menudata.js"></script>
|
||||
<script type="text/javascript" src="menu.js"></script>
|
||||
</div><!-- top -->
|
||||
<div id="side-nav" class="ui-resizable side-nav-resizable">
|
||||
<div id="nav-tree">
|
||||
<div id="nav-tree-contents">
|
||||
<div id="nav-sync" class="sync"></div>
|
||||
</div>
|
||||
</div>
|
||||
<div id="splitbar" style="-moz-user-select:none;"
|
||||
class="ui-resizable-handle">
|
||||
</div>
|
||||
</div>
|
||||
<script type="text/javascript">
|
||||
/* @license magnet:?xt=urn:btih:d3d9a9a6595521f9666a5e94cc830dab83b65699&dn=expat.txt MIT */
|
||||
$(function() {
|
||||
initMenu('',true,false,'search.php','Search',false);
|
||||
$(function() { init_search(); });
|
||||
});
|
||||
/* @license-end */
|
||||
</script>
|
||||
<div id="main-nav"></div>
|
||||
<script type="text/javascript">
|
||||
/* @license magnet:?xt=urn:btih:d3d9a9a6595521f9666a5e94cc830dab83b65699&dn=expat.txt MIT */
|
||||
$(function(){ initResizable(false); });
|
||||
$(function(){initNavTree('scan_8h_source.html',''); initResizable(true); });
|
||||
/* @license-end */
|
||||
</script>
|
||||
<div id="doc-content">
|
||||
<!-- window showing the filter options -->
|
||||
<div id="MSearchSelectWindow"
|
||||
onmouseover="return searchBox.OnSearchSelectShow()"
|
||||
@@ -81,12 +101,6 @@ $(function(){ initResizable(false); });
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="nav-path" class="navpath">
|
||||
<ul>
|
||||
<li class="navelem"><a class="el" href="dir_938ab0ecf10b8b860ff766c820f665fd.html">mlx</a></li><li class="navelem"><a class="el" href="dir_1d446c9bd3c99228254c9484e0bc5c06.html">backend</a></li><li class="navelem"><a class="el" href="dir_d0c977ea65824390717cdb7efc36c157.html">metal</a></li><li class="navelem"><a class="el" href="dir_70a37effa88bcbd6b791977fa1e64356.html">kernels</a></li> </ul>
|
||||
</div>
|
||||
</div><!-- top -->
|
||||
<div id="doc-content">
|
||||
<div class="header">
|
||||
<div class="headertitle"><div class="title">scan.h</div></div>
|
||||
</div><!--header-->
|
||||
@@ -173,20 +187,20 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00072" name="l00072"></a><span class="lineno"> 72</span> </div>
|
||||
<div class="line"><a id="l00073" name="l00073"></a><span class="lineno"> 73</span><span class="keyword">template</span> <></div>
|
||||
<div class="foldopen" id="foldopen00074" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00074" name="l00074"></a><span class="lineno"><a class="line" href="struct_cum_prod_3_01bool_01_4.html"> 74</a></span><span class="keyword">struct </span><a class="code hl_struct" href="struct_cum_prod.html">CumProd</a><bool> {</div>
|
||||
<div class="line"><a id="l00075" name="l00075"></a><span class="lineno"><a class="line" href="struct_cum_prod_3_01bool_01_4.html#ae7a8b0ba9e6898356b87b18766e76d2c"> 75</a></span> <span class="keyword">static</span> <span class="keyword">constexpr</span> constant <span class="keywordtype">bool</span> init = <span class="keyword">true</span>;</div>
|
||||
<div class="line"><a id="l00074" name="l00074"></a><span class="lineno"><a class="line" href="struct_cum_prod.html"> 74</a></span><span class="keyword">struct </span><a class="code hl_struct" href="struct_cum_prod.html">CumProd</a><bool> {</div>
|
||||
<div class="line"><a id="l00075" name="l00075"></a><span class="lineno"><a class="line" href="struct_cum_prod.html#ae7a8b0ba9e6898356b87b18766e76d2c"> 75</a></span> <span class="keyword">static</span> <span class="keyword">constexpr</span> constant <span class="keywordtype">bool</span> <a class="code hl_variable" href="struct_cum_prod.html#ae7a8b0ba9e6898356b87b18766e76d2c">init</a> = <span class="keyword">true</span>;</div>
|
||||
<div class="line"><a id="l00076" name="l00076"></a><span class="lineno"> 76</span> </div>
|
||||
<div class="line"><a id="l00077" name="l00077"></a><span class="lineno"> 77</span> <span class="keyword">template</span> <<span class="keyword">typename</span> T></div>
|
||||
<div class="foldopen" id="foldopen00078" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00078" name="l00078"></a><span class="lineno"><a class="line" href="struct_cum_prod_3_01bool_01_4.html#ad634be0b139d10ce6d21332eef0d936b"> 78</a></span> <span class="keywordtype">bool</span> <a class="code hl_function" href="struct_cum_prod_3_01bool_01_4.html#ad634be0b139d10ce6d21332eef0d936b">operator()</a>(<span class="keywordtype">bool</span> a, T b) {</div>
|
||||
<div class="line"><a id="l00078" name="l00078"></a><span class="lineno"><a class="line" href="struct_cum_prod.html#ad634be0b139d10ce6d21332eef0d936b"> 78</a></span> <span class="keywordtype">bool</span> <a class="code hl_function" href="struct_cum_prod.html#ad634be0b139d10ce6d21332eef0d936b">operator()</a>(<span class="keywordtype">bool</span> a, T b) {</div>
|
||||
<div class="line"><a id="l00079" name="l00079"></a><span class="lineno"> 79</span> <span class="keywordflow">return</span> a & <span class="keyword">static_cast<</span><span class="keywordtype">bool</span><span class="keyword">></span>(b);</div>
|
||||
<div class="line"><a id="l00080" name="l00080"></a><span class="lineno"> 80</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00081" name="l00081"></a><span class="lineno"> 81</span> </div>
|
||||
<div class="foldopen" id="foldopen00082" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00082" name="l00082"></a><span class="lineno"><a class="line" href="struct_cum_prod_3_01bool_01_4.html#abeb5ec4237b330e7219f4e881cf10d7a"> 82</a></span> <span class="keywordtype">bool</span> <a class="code hl_function" href="struct_cum_prod_3_01bool_01_4.html#abeb5ec4237b330e7219f4e881cf10d7a">simd_scan</a>(<span class="keywordtype">bool</span> x) {</div>
|
||||
<div class="line"><a id="l00082" name="l00082"></a><span class="lineno"><a class="line" href="struct_cum_prod.html#abeb5ec4237b330e7219f4e881cf10d7a"> 82</a></span> <span class="keywordtype">bool</span> <a class="code hl_function" href="struct_cum_prod.html#abeb5ec4237b330e7219f4e881cf10d7a">simd_scan</a>(<span class="keywordtype">bool</span> x) {</div>
|
||||
<div class="line"><a id="l00083" name="l00083"></a><span class="lineno"> 83</span> <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = 1; i <= 16; i *= 2) {</div>
|
||||
<div class="line"><a id="l00084" name="l00084"></a><span class="lineno"> 84</span> <span class="keywordtype">bool</span> other = <a class="code hl_function" href="backend_2metal_2kernels_2utils_8h.html#a5862d5ea154c9b76cf56a630cf6385b4">simd_shuffle_and_fill_up</a>(x, init, i);</div>
|
||||
<div class="line"><a id="l00084" name="l00084"></a><span class="lineno"> 84</span> <span class="keywordtype">bool</span> other = <a class="code hl_function" href="backend_2metal_2kernels_2utils_8h.html#a5862d5ea154c9b76cf56a630cf6385b4">simd_shuffle_and_fill_up</a>(x, <a class="code hl_variable" href="struct_cum_prod.html#ae7a8b0ba9e6898356b87b18766e76d2c">init</a>, i);</div>
|
||||
<div class="line"><a id="l00085" name="l00085"></a><span class="lineno"> 85</span> x &= other;</div>
|
||||
<div class="line"><a id="l00086" name="l00086"></a><span class="lineno"> 86</span> }</div>
|
||||
<div class="line"><a id="l00087" name="l00087"></a><span class="lineno"> 87</span> <span class="keywordflow">return</span> x;</div>
|
||||
@@ -194,9 +208,9 @@ $(function(){ initResizable(false); });
|
||||
</div>
|
||||
<div class="line"><a id="l00089" name="l00089"></a><span class="lineno"> 89</span> </div>
|
||||
<div class="foldopen" id="foldopen00090" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00090" name="l00090"></a><span class="lineno"><a class="line" href="struct_cum_prod_3_01bool_01_4.html#a1a86e9398bae24182b7be0a6577bf223"> 90</a></span> <span class="keywordtype">bool</span> <a class="code hl_function" href="struct_cum_prod_3_01bool_01_4.html#a1a86e9398bae24182b7be0a6577bf223">simd_exclusive_scan</a>(<span class="keywordtype">bool</span> x) {</div>
|
||||
<div class="line"><a id="l00091" name="l00091"></a><span class="lineno"> 91</span> x = simd_scan(x);</div>
|
||||
<div class="line"><a id="l00092" name="l00092"></a><span class="lineno"> 92</span> <span class="keywordflow">return</span> <a class="code hl_function" href="backend_2metal_2kernels_2utils_8h.html#a5862d5ea154c9b76cf56a630cf6385b4">simd_shuffle_and_fill_up</a>(x, init, 1);</div>
|
||||
<div class="line"><a id="l00090" name="l00090"></a><span class="lineno"><a class="line" href="struct_cum_prod.html#a1a86e9398bae24182b7be0a6577bf223"> 90</a></span> <span class="keywordtype">bool</span> <a class="code hl_function" href="struct_cum_prod.html#a1a86e9398bae24182b7be0a6577bf223">simd_exclusive_scan</a>(<span class="keywordtype">bool</span> x) {</div>
|
||||
<div class="line"><a id="l00091" name="l00091"></a><span class="lineno"> 91</span> x = <a class="code hl_function" href="struct_cum_prod.html#abeb5ec4237b330e7219f4e881cf10d7a">simd_scan</a>(x);</div>
|
||||
<div class="line"><a id="l00092" name="l00092"></a><span class="lineno"> 92</span> <span class="keywordflow">return</span> <a class="code hl_function" href="backend_2metal_2kernels_2utils_8h.html#a5862d5ea154c9b76cf56a630cf6385b4">simd_shuffle_and_fill_up</a>(x, <a class="code hl_variable" href="struct_cum_prod.html#ae7a8b0ba9e6898356b87b18766e76d2c">init</a>, 1);</div>
|
||||
<div class="line"><a id="l00093" name="l00093"></a><span class="lineno"> 93</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00094" name="l00094"></a><span class="lineno"> 94</span>};</div>
|
||||
@@ -205,7 +219,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00096" name="l00096"></a><span class="lineno"> 96</span><span class="keyword">template</span> <<span class="keyword">typename</span> U></div>
|
||||
<div class="foldopen" id="foldopen00097" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00097" name="l00097"></a><span class="lineno"><a class="line" href="struct_cum_max.html"> 97</a></span><span class="keyword">struct </span><a class="code hl_struct" href="struct_cum_max.html">CumMax</a> {</div>
|
||||
<div class="line"><a id="l00098" name="l00098"></a><span class="lineno"><a class="line" href="struct_cum_max.html#a16480052a2eeb4340e546838aab59cc4"> 98</a></span> <span class="keyword">static</span> <span class="keyword">constexpr</span> constant U <a class="code hl_variable" href="struct_cum_max.html#a16480052a2eeb4340e546838aab59cc4">init</a> = <a class="code hl_struct" href="struct_limits.html">Limits<U>::min</a>;</div>
|
||||
<div class="line"><a id="l00098" name="l00098"></a><span class="lineno"><a class="line" href="struct_cum_max.html#a16480052a2eeb4340e546838aab59cc4"> 98</a></span> <span class="keyword">static</span> <span class="keyword">constexpr</span> constant U <a class="code hl_variable" href="struct_cum_max.html#a16480052a2eeb4340e546838aab59cc4">init</a> = <a class="code hl_variable" href="struct_limits.html#a6e81584ba65a4dc6ff9366b458e3a20e">Limits<U>::min</a>;</div>
|
||||
<div class="line"><a id="l00099" name="l00099"></a><span class="lineno"> 99</span> </div>
|
||||
<div class="line"><a id="l00100" name="l00100"></a><span class="lineno"> 100</span> <span class="keyword">template</span> <<span class="keyword">typename</span> T></div>
|
||||
<div class="foldopen" id="foldopen00101" data-start="{" data-end="}">
|
||||
@@ -236,7 +250,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00119" name="l00119"></a><span class="lineno"> 119</span><span class="keyword">template</span> <<span class="keyword">typename</span> U></div>
|
||||
<div class="foldopen" id="foldopen00120" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00120" name="l00120"></a><span class="lineno"><a class="line" href="struct_cum_min.html"> 120</a></span><span class="keyword">struct </span><a class="code hl_struct" href="struct_cum_min.html">CumMin</a> {</div>
|
||||
<div class="line"><a id="l00121" name="l00121"></a><span class="lineno"><a class="line" href="struct_cum_min.html#a8b67f739c620d0cc194b533190990ab9"> 121</a></span> <span class="keyword">static</span> <span class="keyword">constexpr</span> constant U <a class="code hl_variable" href="struct_cum_min.html#a8b67f739c620d0cc194b533190990ab9">init</a> = <a class="code hl_struct" href="struct_limits.html">Limits<U>::max</a>;</div>
|
||||
<div class="line"><a id="l00121" name="l00121"></a><span class="lineno"><a class="line" href="struct_cum_min.html#a8b67f739c620d0cc194b533190990ab9"> 121</a></span> <span class="keyword">static</span> <span class="keyword">constexpr</span> constant U <a class="code hl_variable" href="struct_cum_min.html#a8b67f739c620d0cc194b533190990ab9">init</a> = <a class="code hl_variable" href="struct_limits.html#a2f0673b6f9da89ce1d64f9f3d74f50a8">Limits<U>::max</a>;</div>
|
||||
<div class="line"><a id="l00122" name="l00122"></a><span class="lineno"> 122</span> </div>
|
||||
<div class="line"><a id="l00123" name="l00123"></a><span class="lineno"> 123</span> <span class="keyword">template</span> <<span class="keyword">typename</span> T></div>
|
||||
<div class="foldopen" id="foldopen00124" data-start="{" data-end="}">
|
||||
@@ -353,7 +367,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00219" name="l00219"></a><span class="lineno"> 219</span> uint simd_lane_id [[thread_index_in_simdgroup]],</div>
|
||||
<div class="line"><a id="l00220" name="l00220"></a><span class="lineno"> 220</span> uint simd_group_id [[simdgroup_index_in_threadgroup]]) {</div>
|
||||
<div class="line"><a id="l00221" name="l00221"></a><span class="lineno"> 221</span> <span class="keyword">constexpr</span> <span class="keywordtype">int</span> <a class="code hl_variable" href="backend_2metal_2kernels_2reduction_2ops_8h.html#a515b75d563a93d3c09ee677948dc83e3">simd_size</a> = 32;</div>
|
||||
<div class="line"><a id="l00222" name="l00222"></a><span class="lineno"> 222</span> Op <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>;</div>
|
||||
<div class="line"><a id="l00222" name="l00222"></a><span class="lineno"> 222</span> Op op;</div>
|
||||
<div class="line"><a id="l00223" name="l00223"></a><span class="lineno"> 223</span> </div>
|
||||
<div class="line"><a id="l00224" name="l00224"></a><span class="lineno"> 224</span> <span class="comment">// Position the pointers</span></div>
|
||||
<div class="line"><a id="l00225" name="l00225"></a><span class="lineno"> 225</span> <span class="keywordtype">size_t</span> offset = (gid.y + gsize.y * size_t(gid.z)) * axis_size;</div>
|
||||
@@ -408,30 +422,30 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00274" name="l00274"></a><span class="lineno"> 274</span> </div>
|
||||
<div class="line"><a id="l00275" name="l00275"></a><span class="lineno"> 275</span> <span class="comment">// Compute an inclusive scan per thread</span></div>
|
||||
<div class="line"><a id="l00276" name="l00276"></a><span class="lineno"> 276</span> <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = 1; i < N_READS; i++) {</div>
|
||||
<div class="line"><a id="l00277" name="l00277"></a><span class="lineno"> 277</span> values[i] = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>(values[i], values[i - 1]);</div>
|
||||
<div class="line"><a id="l00277" name="l00277"></a><span class="lineno"> 277</span> values[i] = op(values[i], values[i - 1]);</div>
|
||||
<div class="line"><a id="l00278" name="l00278"></a><span class="lineno"> 278</span> }</div>
|
||||
<div class="line"><a id="l00279" name="l00279"></a><span class="lineno"> 279</span> </div>
|
||||
<div class="line"><a id="l00280" name="l00280"></a><span class="lineno"> 280</span> <span class="comment">// Compute exclusive scan of thread sums</span></div>
|
||||
<div class="line"><a id="l00281" name="l00281"></a><span class="lineno"> 281</span> U prev_thread = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>.simd_exclusive_scan(values[N_READS - 1]);</div>
|
||||
<div class="line"><a id="l00281" name="l00281"></a><span class="lineno"> 281</span> U prev_thread = op.simd_exclusive_scan(values[N_READS - 1]);</div>
|
||||
<div class="line"><a id="l00282" name="l00282"></a><span class="lineno"> 282</span> </div>
|
||||
<div class="line"><a id="l00283" name="l00283"></a><span class="lineno"> 283</span> <span class="comment">// Write simdgroup_sums to SM</span></div>
|
||||
<div class="line"><a id="l00284" name="l00284"></a><span class="lineno"> 284</span> <span class="keywordflow">if</span> (simd_lane_id == <a class="code hl_variable" href="backend_2metal_2kernels_2reduction_2ops_8h.html#a515b75d563a93d3c09ee677948dc83e3">simd_size</a> - 1) {</div>
|
||||
<div class="line"><a id="l00285" name="l00285"></a><span class="lineno"> 285</span> simdgroup_sums[simd_group_id] = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>(prev_thread, values[N_READS - 1]);</div>
|
||||
<div class="line"><a id="l00285" name="l00285"></a><span class="lineno"> 285</span> simdgroup_sums[simd_group_id] = op(prev_thread, values[N_READS - 1]);</div>
|
||||
<div class="line"><a id="l00286" name="l00286"></a><span class="lineno"> 286</span> }</div>
|
||||
<div class="line"><a id="l00287" name="l00287"></a><span class="lineno"> 287</span> threadgroup_barrier(mem_flags::mem_threadgroup);</div>
|
||||
<div class="line"><a id="l00288" name="l00288"></a><span class="lineno"> 288</span> </div>
|
||||
<div class="line"><a id="l00289" name="l00289"></a><span class="lineno"> 289</span> <span class="comment">// Compute exclusive scan of simdgroup_sums</span></div>
|
||||
<div class="line"><a id="l00290" name="l00290"></a><span class="lineno"> 290</span> <span class="keywordflow">if</span> (simd_group_id == 0) {</div>
|
||||
<div class="line"><a id="l00291" name="l00291"></a><span class="lineno"> 291</span> U prev_simdgroup = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>.simd_exclusive_scan(simdgroup_sums[simd_lane_id]);</div>
|
||||
<div class="line"><a id="l00291" name="l00291"></a><span class="lineno"> 291</span> U prev_simdgroup = op.simd_exclusive_scan(simdgroup_sums[simd_lane_id]);</div>
|
||||
<div class="line"><a id="l00292" name="l00292"></a><span class="lineno"> 292</span> simdgroup_sums[simd_lane_id] = prev_simdgroup;</div>
|
||||
<div class="line"><a id="l00293" name="l00293"></a><span class="lineno"> 293</span> }</div>
|
||||
<div class="line"><a id="l00294" name="l00294"></a><span class="lineno"> 294</span> threadgroup_barrier(mem_flags::mem_threadgroup);</div>
|
||||
<div class="line"><a id="l00295" name="l00295"></a><span class="lineno"> 295</span> </div>
|
||||
<div class="line"><a id="l00296" name="l00296"></a><span class="lineno"> 296</span> <span class="comment">// Compute the output</span></div>
|
||||
<div class="line"><a id="l00297" name="l00297"></a><span class="lineno"> 297</span> <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = 0; i < N_READS; i++) {</div>
|
||||
<div class="line"><a id="l00298" name="l00298"></a><span class="lineno"> 298</span> values[i] = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>(values[i], prefix);</div>
|
||||
<div class="line"><a id="l00299" name="l00299"></a><span class="lineno"> 299</span> values[i] = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>(values[i], simdgroup_sums[simd_group_id]);</div>
|
||||
<div class="line"><a id="l00300" name="l00300"></a><span class="lineno"> 300</span> values[i] = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>(values[i], prev_thread);</div>
|
||||
<div class="line"><a id="l00298" name="l00298"></a><span class="lineno"> 298</span> values[i] = op(values[i], prefix);</div>
|
||||
<div class="line"><a id="l00299" name="l00299"></a><span class="lineno"> 299</span> values[i] = op(values[i], simdgroup_sums[simd_group_id]);</div>
|
||||
<div class="line"><a id="l00300" name="l00300"></a><span class="lineno"> 300</span> values[i] = op(values[i], prev_thread);</div>
|
||||
<div class="line"><a id="l00301" name="l00301"></a><span class="lineno"> 301</span> }</div>
|
||||
<div class="line"><a id="l00302" name="l00302"></a><span class="lineno"> 302</span> </div>
|
||||
<div class="line"><a id="l00303" name="l00303"></a><span class="lineno"> 303</span> <span class="comment">// Write the values</span></div>
|
||||
@@ -516,7 +530,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00380" name="l00380"></a><span class="lineno"> 380</span> <span class="keyword">constexpr</span> <span class="keywordtype">int</span> BN_pad = 32 + 16 / <span class="keyword">sizeof</span>(U);</div>
|
||||
<div class="line"><a id="l00381" name="l00381"></a><span class="lineno"> 381</span> <span class="keyword">constexpr</span> <span class="keywordtype">int</span> n_simds = BN / N_READS;</div>
|
||||
<div class="line"><a id="l00382" name="l00382"></a><span class="lineno"> 382</span> <span class="keyword">constexpr</span> <span class="keywordtype">int</span> n_scans = BN / n_simds;</div>
|
||||
<div class="line"><a id="l00383" name="l00383"></a><span class="lineno"> 383</span> Op <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>;</div>
|
||||
<div class="line"><a id="l00383" name="l00383"></a><span class="lineno"> 383</span> Op op;</div>
|
||||
<div class="line"><a id="l00384" name="l00384"></a><span class="lineno"> 384</span> </div>
|
||||
<div class="line"><a id="l00385" name="l00385"></a><span class="lineno"> 385</span> threadgroup U read_buffer[BM * BN_pad];</div>
|
||||
<div class="line"><a id="l00386" name="l00386"></a><span class="lineno"> 386</span> U values[n_scans];</div>
|
||||
@@ -574,8 +588,8 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00438" name="l00438"></a><span class="lineno"> 438</span> </div>
|
||||
<div class="line"><a id="l00439" name="l00439"></a><span class="lineno"> 439</span> <span class="comment">// Perform the scan</span></div>
|
||||
<div class="line"><a id="l00440" name="l00440"></a><span class="lineno"> 440</span> <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = 0; i < n_scans; i++) {</div>
|
||||
<div class="line"><a id="l00441" name="l00441"></a><span class="lineno"> 441</span> values[i] = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>.simd_scan(values[i]);</div>
|
||||
<div class="line"><a id="l00442" name="l00442"></a><span class="lineno"> 442</span> values[i] = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>(values[i], prefix[i]);</div>
|
||||
<div class="line"><a id="l00441" name="l00441"></a><span class="lineno"> 441</span> values[i] = op.simd_scan(values[i]);</div>
|
||||
<div class="line"><a id="l00442" name="l00442"></a><span class="lineno"> 442</span> values[i] = op(values[i], prefix[i]);</div>
|
||||
<div class="line"><a id="l00443" name="l00443"></a><span class="lineno"> 443</span> prefix[i] = <a class="code hl_function" href="backend_2metal_2kernels_2utils_8h.html#a71986ecdd7d18f975dd22c3df7421ce2">simd_shuffle</a>(values[i], <a class="code hl_variable" href="backend_2metal_2kernels_2reduction_2ops_8h.html#a515b75d563a93d3c09ee677948dc83e3">simd_size</a> - 1);</div>
|
||||
<div class="line"><a id="l00444" name="l00444"></a><span class="lineno"> 444</span> }</div>
|
||||
<div class="line"><a id="l00445" name="l00445"></a><span class="lineno"> 445</span> </div>
|
||||
@@ -623,10 +637,9 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00487" name="l00487"></a><span class="lineno"> 487</span>}</div>
|
||||
</div>
|
||||
<div class="ttc" id="abackend_2metal_2kernels_2reduction_2ops_8h_html_a515b75d563a93d3c09ee677948dc83e3"><div class="ttname"><a href="backend_2metal_2kernels_2reduction_2ops_8h.html#a515b75d563a93d3c09ee677948dc83e3">simd_size</a></div><div class="ttdeci">static constant constexpr const uint8_t simd_size</div><div class="ttdef"><b>Definition</b> ops.h:22</div></div>
|
||||
<div class="ttc" id="abackend_2metal_2kernels_2utils_8h_html_a5862d5ea154c9b76cf56a630cf6385b4"><div class="ttname"><a href="backend_2metal_2kernels_2utils_8h.html#a5862d5ea154c9b76cf56a630cf6385b4">simd_shuffle_and_fill_up</a></div><div class="ttdeci">uint64_t simd_shuffle_and_fill_up(uint64_t data, uint64_t filling, uint16_t delta)</div><div class="ttdef"><b>Definition</b> utils.h:383</div></div>
|
||||
<div class="ttc" id="abackend_2metal_2kernels_2utils_8h_html_a71986ecdd7d18f975dd22c3df7421ce2"><div class="ttname"><a href="backend_2metal_2kernels_2utils_8h.html#a71986ecdd7d18f975dd22c3df7421ce2">simd_shuffle</a></div><div class="ttdeci">uint64_t simd_shuffle(uint64_t data, uint16_t lane)</div><div class="ttdef"><b>Definition</b> utils.h:408</div></div>
|
||||
<div class="ttc" id="abackend_2metal_2kernels_2utils_8h_html_a8e5a4b0fb5d018d7b078d147efe4f1e3"><div class="ttname"><a href="backend_2metal_2kernels_2utils_8h.html#a8e5a4b0fb5d018d7b078d147efe4f1e3">ceildiv</a></div><div class="ttdeci">T ceildiv(T N, U M)</div><div class="ttdoc">Compute ceil((float)N/(float)M)</div><div class="ttdef"><b>Definition</b> utils.h:313</div></div>
|
||||
<div class="ttc" id="acommon_2binary_8h_html_a70228731d29946574b238d21fb4b360c"><div class="ttname"><a href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a></div><div class="ttdeci">Op op</div><div class="ttdef"><b>Definition</b> binary.h:129</div></div>
|
||||
<div class="ttc" id="abackend_2metal_2kernels_2utils_8h_html_a5862d5ea154c9b76cf56a630cf6385b4"><div class="ttname"><a href="backend_2metal_2kernels_2utils_8h.html#a5862d5ea154c9b76cf56a630cf6385b4">simd_shuffle_and_fill_up</a></div><div class="ttdeci">uint64_t simd_shuffle_and_fill_up(uint64_t data, uint64_t filling, uint16_t delta)</div><div class="ttdef"><b>Definition</b> utils.h:372</div></div>
|
||||
<div class="ttc" id="abackend_2metal_2kernels_2utils_8h_html_a71986ecdd7d18f975dd22c3df7421ce2"><div class="ttname"><a href="backend_2metal_2kernels_2utils_8h.html#a71986ecdd7d18f975dd22c3df7421ce2">simd_shuffle</a></div><div class="ttdeci">uint64_t simd_shuffle(uint64_t data, uint16_t lane)</div><div class="ttdef"><b>Definition</b> utils.h:397</div></div>
|
||||
<div class="ttc" id="abackend_2metal_2kernels_2utils_8h_html_a8e5a4b0fb5d018d7b078d147efe4f1e3"><div class="ttname"><a href="backend_2metal_2kernels_2utils_8h.html#a8e5a4b0fb5d018d7b078d147efe4f1e3">ceildiv</a></div><div class="ttdeci">T ceildiv(T N, U M)</div><div class="ttdoc">Compute ceil((float)N/(float)M)</div><div class="ttdef"><b>Definition</b> utils.h:302</div></div>
|
||||
<div class="ttc" id="ascan_8h_html_a0d8d6a9b0f3a1263629380bda8eca7bc"><div class="ttname"><a href="scan_8h.html#a0d8d6a9b0f3a1263629380bda8eca7bc">DEFINE_SIMD_SCAN</a></div><div class="ttdeci">#define DEFINE_SIMD_SCAN()</div><div class="ttdef"><b>Definition</b> scan.h:5</div></div>
|
||||
<div class="ttc" id="ascan_8h_html_a185f66aac8c5317587e6abd43f3013fc"><div class="ttname"><a href="scan_8h.html#a185f66aac8c5317587e6abd43f3013fc">DEFINE_SIMD_EXCLUSIVE_SCAN</a></div><div class="ttdeci">#define DEFINE_SIMD_EXCLUSIVE_SCAN()</div><div class="ttdef"><b>Definition</b> scan.h:19</div></div>
|
||||
<div class="ttc" id="ascan_8h_html_a60d279b9add7d56639bb209408f09d79"><div class="ttname"><a href="scan_8h.html#a60d279b9add7d56639bb209408f09d79">contiguous_scan</a></div><div class="ttdeci">void contiguous_scan(const device T *in, device U *out, const constant size_t &axis_size, uint3 gid, uint3 gsize, uint3 lid, uint3 lsize, uint simd_lane_id, uint simd_group_id)</div><div class="ttdef"><b>Definition</b> scan.h:211</div></div>
|
||||
@@ -645,17 +658,22 @@ $(function(){ initResizable(false); });
|
||||
<div class="ttc" id="astruct_cum_min_html_a83e65017ff33018b585c043fb803773b"><div class="ttname"><a href="struct_cum_min.html#a83e65017ff33018b585c043fb803773b">CumMin::simd_exclusive_scan</a></div><div class="ttdeci">U simd_exclusive_scan(U x)</div><div class="ttdef"><b>Definition</b> scan.h:136</div></div>
|
||||
<div class="ttc" id="astruct_cum_min_html_a8b67f739c620d0cc194b533190990ab9"><div class="ttname"><a href="struct_cum_min.html#a8b67f739c620d0cc194b533190990ab9">CumMin::init</a></div><div class="ttdeci">static constexpr constant U init</div><div class="ttdef"><b>Definition</b> scan.h:121</div></div>
|
||||
<div class="ttc" id="astruct_cum_min_html_ae0b8c3761e04fa538d304ca842281a66"><div class="ttname"><a href="struct_cum_min.html#ae0b8c3761e04fa538d304ca842281a66">CumMin::operator()</a></div><div class="ttdeci">U operator()(U a, T b)</div><div class="ttdef"><b>Definition</b> scan.h:124</div></div>
|
||||
<div class="ttc" id="astruct_cum_prod_3_01bool_01_4_html_a1a86e9398bae24182b7be0a6577bf223"><div class="ttname"><a href="struct_cum_prod_3_01bool_01_4.html#a1a86e9398bae24182b7be0a6577bf223">CumProd< bool >::simd_exclusive_scan</a></div><div class="ttdeci">bool simd_exclusive_scan(bool x)</div><div class="ttdef"><b>Definition</b> scan.h:90</div></div>
|
||||
<div class="ttc" id="astruct_cum_prod_3_01bool_01_4_html_abeb5ec4237b330e7219f4e881cf10d7a"><div class="ttname"><a href="struct_cum_prod_3_01bool_01_4.html#abeb5ec4237b330e7219f4e881cf10d7a">CumProd< bool >::simd_scan</a></div><div class="ttdeci">bool simd_scan(bool x)</div><div class="ttdef"><b>Definition</b> scan.h:82</div></div>
|
||||
<div class="ttc" id="astruct_cum_prod_3_01bool_01_4_html_ad634be0b139d10ce6d21332eef0d936b"><div class="ttname"><a href="struct_cum_prod_3_01bool_01_4.html#ad634be0b139d10ce6d21332eef0d936b">CumProd< bool >::operator()</a></div><div class="ttdeci">bool operator()(bool a, T b)</div><div class="ttdef"><b>Definition</b> scan.h:78</div></div>
|
||||
<div class="ttc" id="astruct_cum_prod_html"><div class="ttname"><a href="struct_cum_prod.html">CumProd</a></div><div class="ttdef"><b>Definition</b> scan.h:53</div></div>
|
||||
<div class="ttc" id="astruct_cum_prod_html_a1a86e9398bae24182b7be0a6577bf223"><div class="ttname"><a href="struct_cum_prod.html#a1a86e9398bae24182b7be0a6577bf223">CumProd< bool >::simd_exclusive_scan</a></div><div class="ttdeci">bool simd_exclusive_scan(bool x)</div><div class="ttdef"><b>Definition</b> scan.h:90</div></div>
|
||||
<div class="ttc" id="astruct_cum_prod_html_abeb5ec4237b330e7219f4e881cf10d7a"><div class="ttname"><a href="struct_cum_prod.html#abeb5ec4237b330e7219f4e881cf10d7a">CumProd< bool >::simd_scan</a></div><div class="ttdeci">bool simd_scan(bool x)</div><div class="ttdef"><b>Definition</b> scan.h:82</div></div>
|
||||
<div class="ttc" id="astruct_cum_prod_html_ad634be0b139d10ce6d21332eef0d936b"><div class="ttname"><a href="struct_cum_prod.html#ad634be0b139d10ce6d21332eef0d936b">CumProd< bool >::operator()</a></div><div class="ttdeci">bool operator()(bool a, T b)</div><div class="ttdef"><b>Definition</b> scan.h:78</div></div>
|
||||
<div class="ttc" id="astruct_cum_prod_html_ae7a8b0ba9e6898356b87b18766e76d2c"><div class="ttname"><a href="struct_cum_prod.html#ae7a8b0ba9e6898356b87b18766e76d2c">CumProd< bool >::init</a></div><div class="ttdeci">static constexpr constant bool init</div><div class="ttdef"><b>Definition</b> scan.h:75</div></div>
|
||||
<div class="ttc" id="astruct_cum_sum_html"><div class="ttname"><a href="struct_cum_sum.html">CumSum</a></div><div class="ttdef"><b>Definition</b> scan.h:32</div></div>
|
||||
<div class="ttc" id="astruct_limits_html"><div class="ttname"><a href="struct_limits.html">Limits</a></div><div class="ttdef"><b>Definition</b> utils.h:23</div></div>
|
||||
<div class="ttc" id="astruct_limits_html_a2f0673b6f9da89ce1d64f9f3d74f50a8"><div class="ttname"><a href="struct_limits.html#a2f0673b6f9da89ce1d64f9f3d74f50a8">Limits::max</a></div><div class="ttdeci">static const constant U max</div><div class="ttdef"><b>Definition</b> utils.h:24</div></div>
|
||||
<div class="ttc" id="astruct_limits_html_a6e81584ba65a4dc6ff9366b458e3a20e"><div class="ttname"><a href="struct_limits.html#a6e81584ba65a4dc6ff9366b458e3a20e">Limits::min</a></div><div class="ttdeci">static const constant U min</div><div class="ttdef"><b>Definition</b> utils.h:25</div></div>
|
||||
</div><!-- fragment --></div><!-- contents -->
|
||||
<!-- start footer part -->
|
||||
<hr class="footer"/><address class="footer"><small>
|
||||
Generated by <a href="https://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.12.0
|
||||
</small></address>
|
||||
</div><!-- doc-content -->
|
||||
<!-- start footer part -->
|
||||
<div id="nav-path" class="navpath"><!-- id is needed for treeview function! -->
|
||||
<ul>
|
||||
<li class="navelem"><a class="el" href="dir_938ab0ecf10b8b860ff766c820f665fd.html">mlx</a></li><li class="navelem"><a class="el" href="dir_1d446c9bd3c99228254c9484e0bc5c06.html">backend</a></li><li class="navelem"><a class="el" href="dir_d0c977ea65824390717cdb7efc36c157.html">metal</a></li><li class="navelem"><a class="el" href="dir_70a37effa88bcbd6b791977fa1e64356.html">kernels</a></li><li class="navelem"><a class="el" href="scan_8h.html">scan.h</a></li>
|
||||
<li class="footer">Generated by <a href="https://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.13.1 </li>
|
||||
</ul>
|
||||
</div>
|
||||
</body>
|
||||
</html>
|
||||
|
Reference in New Issue
Block a user