mirror of
https://github.com/ml-explore/mlx.git
synced 2025-09-19 10:48:09 +08:00
rebase
This commit is contained in:
270
docs/build/html/gemm_2mma_8h_source.html
vendored
270
docs/build/html/gemm_2mma_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/steel/gemm/mma.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('gemm_2mma_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><li class="navelem"><a class="el" href="dir_76215a6c54e2b67053e723fc2395583c.html">steel</a></li><li class="navelem"><a class="el" href="dir_6768c99e6145fb9510ccdb40db8ede25.html">gemm</a></li> </ul>
|
||||
</div>
|
||||
</div><!-- top -->
|
||||
<div id="doc-content">
|
||||
<div class="header">
|
||||
<div class="headertitle"><div class="title">mma.h</div></div>
|
||||
</div><!--header-->
|
||||
@@ -104,11 +118,11 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00011" name="l00011"></a><span class="lineno"> 11</span><span class="preprocessor">#include "<a class="code" href="integral__constant_8h.html">mlx/backend/metal/kernels/steel/utils/integral_constant.h</a>"</span></div>
|
||||
<div class="line"><a id="l00012" name="l00012"></a><span class="lineno"> 12</span> </div>
|
||||
<div class="line"><a id="l00013" name="l00013"></a><span class="lineno"> 13</span><span class="keyword">using namespace </span><a class="code hl_namespace" href="namespacemetal.html">metal</a>;</div>
|
||||
<div class="line"><a id="l00014" name="l00014"></a><span class="lineno"> 14</span> </div>
|
||||
<div class="line"><a id="l00014" name="l00014"></a><span class="lineno"> 14</span></div>
|
||||
<div class="line"><a id="l00016" name="l00016"></a><span class="lineno"> 16</span><span class="comment">// MMA helper</span></div>
|
||||
<div class="line"><a id="l00018" name="l00018"></a><span class="lineno"> 18</span> </div>
|
||||
<div class="line"><a id="l00019" name="l00019"></a><span class="lineno"> 19</span><span class="keyword">namespace </span><a class="code hl_namespace" href="namespacemlx.html">mlx</a> {</div>
|
||||
<div class="line"><a id="l00020" name="l00020"></a><span class="lineno"> 20</span><span class="keyword">namespace </span>steel {</div>
|
||||
<div class="line"><a id="l00020" name="l00020"></a><span class="lineno"> 20</span><span class="keyword">namespace </span><a class="code hl_namespace" href="namespacemlx_1_1steel.html">steel</a> {</div>
|
||||
<div class="line"><a id="l00021" name="l00021"></a><span class="lineno"> 21</span> </div>
|
||||
<div class="line"><a id="l00022" name="l00022"></a><span class="lineno"> 22</span><span class="keyword">template</span> <<span class="keyword">typename</span> T, <span class="keywordtype">int</span> kFragRows_, <span class="keywordtype">int</span> kFragCols_></div>
|
||||
<div class="foldopen" id="foldopen00023" data-start="{" data-end="};">
|
||||
@@ -124,24 +138,24 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00031" name="l00031"></a><span class="lineno"> 31</span> </div>
|
||||
<div class="line"><a id="l00032" name="l00032"></a><span class="lineno"> 32</span><span class="keyword">template</span> <<span class="keyword">typename</span> T></div>
|
||||
<div class="foldopen" id="foldopen00033" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00033" name="l00033"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html"> 33</a></span><span class="keyword">struct </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html">BaseMMAFrag</a><T, 8, 8> {</div>
|
||||
<div class="line"><a id="l00034" name="l00034"></a><span class="lineno"> 34</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> kFragRows = 8;</div>
|
||||
<div class="line"><a id="l00035" name="l00035"></a><span class="lineno"> 35</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> kFragCols = 8;</div>
|
||||
<div class="line"><a id="l00033" name="l00033"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html"> 33</a></span><span class="keyword">struct </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html">BaseMMAFrag</a><T, 8, 8> {</div>
|
||||
<div class="line"><a id="l00034" name="l00034"></a><span class="lineno"> 34</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a2fe53db449c692226f23f6b99fb2c0d4">kFragRows</a> = 8;</div>
|
||||
<div class="line"><a id="l00035" name="l00035"></a><span class="lineno"> 35</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a211102315e2afbcfcd2e2c201b638e9f">kFragCols</a> = 8;</div>
|
||||
<div class="line"><a id="l00036" name="l00036"></a><span class="lineno"> 36</span> </div>
|
||||
<div class="line"><a id="l00037" name="l00037"></a><span class="lineno"> 37</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> kElemsPerFrag = (kFragRows * kFragCols) / 32;</div>
|
||||
<div class="line"><a id="l00037" name="l00037"></a><span class="lineno"> 37</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a3c34dfdc944db110f4735f1b25307cf0">kElemsPerFrag</a> = (<a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a2fe53db449c692226f23f6b99fb2c0d4">kFragRows</a> * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a211102315e2afbcfcd2e2c201b638e9f">kFragCols</a>) / 32;</div>
|
||||
<div class="line"><a id="l00038" name="l00038"></a><span class="lineno"> 38</span> </div>
|
||||
<div class="line"><a id="l00039" name="l00039"></a><span class="lineno"> 39</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> kElemRows = 1;</div>
|
||||
<div class="line"><a id="l00040" name="l00040"></a><span class="lineno"> 40</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> kElemCols = 2;</div>
|
||||
<div class="line"><a id="l00039" name="l00039"></a><span class="lineno"> 39</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a> = 1;</div>
|
||||
<div class="line"><a id="l00040" name="l00040"></a><span class="lineno"> 40</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> = 2;</div>
|
||||
<div class="line"><a id="l00041" name="l00041"></a><span class="lineno"> 41</span> </div>
|
||||
<div class="line"><a id="l00042" name="l00042"></a><span class="lineno"> 42</span> <span class="keyword">static_assert</span>(</div>
|
||||
<div class="line"><a id="l00043" name="l00043"></a><span class="lineno"> 43</span> kElemRows * kElemCols == kElemsPerFrag,</div>
|
||||
<div class="line"><a id="l00043" name="l00043"></a><span class="lineno"> 43</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a> * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> == <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a3c34dfdc944db110f4735f1b25307cf0">kElemsPerFrag</a>,</div>
|
||||
<div class="line"><a id="l00044" name="l00044"></a><span class="lineno"> 44</span> <span class="stringliteral">"MMAFrag shape is not consistent with MMAFrag size"</span>);</div>
|
||||
<div class="line"><a id="l00045" name="l00045"></a><span class="lineno"> 45</span> </div>
|
||||
<div class="line"><a id="l00046" name="l00046"></a><span class="lineno"> 46</span> <span class="keyword">typedef</span> metal::simdgroup_matrix<T, kFragRows, kFragCols> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a>;</div>
|
||||
<div class="line"><a id="l00047" name="l00047"></a><span class="lineno"> 47</span> <span class="keyword">typedef</span> metal::vec<T, kElemsPerFrag> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>;</div>
|
||||
<div class="line"><a id="l00046" name="l00046"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9"> 46</a></span> <span class="keyword">typedef</span> metal::simdgroup_matrix<T, kFragRows, kFragCols> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a>;</div>
|
||||
<div class="line"><a id="l00047" name="l00047"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b"> 47</a></span> <span class="keyword">typedef</span> metal::vec<T, kElemsPerFrag> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>;</div>
|
||||
<div class="line"><a id="l00048" name="l00048"></a><span class="lineno"> 48</span> </div>
|
||||
<div class="foldopen" id="foldopen00049" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00049" name="l00049"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a7331fff1d12f2f8b72b0006a3ad0dd83"> 49</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> short2 <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a7331fff1d12f2f8b72b0006a3ad0dd83">get_coord</a>(ushort simd_lane_id</div>
|
||||
<div class="line"><a id="l00049" name="l00049"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7331fff1d12f2f8b72b0006a3ad0dd83"> 49</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> short2 <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7331fff1d12f2f8b72b0006a3ad0dd83">get_coord</a>(ushort simd_lane_id</div>
|
||||
<div class="line"><a id="l00050" name="l00050"></a><span class="lineno"> 50</span> [[thread_index_in_simdgroup]]) {</div>
|
||||
<div class="line"><a id="l00051" name="l00051"></a><span class="lineno"> 51</span> <span class="keyword">const</span> <span class="keywordtype">short</span> qid = simd_lane_id / 4;</div>
|
||||
<div class="line"><a id="l00052" name="l00052"></a><span class="lineno"> 52</span> <span class="keyword">const</span> <span class="keywordtype">short</span> fm = (qid & 4) + ((simd_lane_id / 2) % 4);</div>
|
||||
@@ -153,12 +167,12 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00057" name="l00057"></a><span class="lineno"> 57</span> <span class="keyword">template</span> <<span class="keyword">typename</span> SrcPtrType, <span class="keyword">typename</span> StrX, <span class="keyword">typename</span> StrY></div>
|
||||
<div class="line"><a id="l00058" name="l00058"></a><span class="lineno"> 58</span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span></div>
|
||||
<div class="foldopen" id="foldopen00059" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00059" name="l00059"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#ac73006b36fc710feda3a7c796e21415c"> 59</a></span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#ac73006b36fc710feda3a7c796e21415c">load</a>(thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>& dst, SrcPtrType src, StrX str_x, StrY str_y) {</div>
|
||||
<div class="line"><a id="l00059" name="l00059"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#ac73006b36fc710feda3a7c796e21415c"> 59</a></span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#ac73006b36fc710feda3a7c796e21415c">load</a>(thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& dst, SrcPtrType src, StrX str_x, StrY str_y) {</div>
|
||||
<div class="line"><a id="l00060" name="l00060"></a><span class="lineno"> 60</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00061" name="l00061"></a><span class="lineno"> 61</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < kElemRows; i++) {</div>
|
||||
<div class="line"><a id="l00061" name="l00061"></a><span class="lineno"> 61</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a>; i++) {</div>
|
||||
<div class="line"><a id="l00062" name="l00062"></a><span class="lineno"> 62</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00063" name="l00063"></a><span class="lineno"> 63</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> j = 0; j < kElemCols; j++) {</div>
|
||||
<div class="line"><a id="l00064" name="l00064"></a><span class="lineno"> 64</span> dst[i * kElemCols + j] = <span class="keyword">static_cast<</span>T<span class="keyword">></span>(src[i * str_x + j * str_y]);</div>
|
||||
<div class="line"><a id="l00063" name="l00063"></a><span class="lineno"> 63</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> j = 0; j < <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a>; j++) {</div>
|
||||
<div class="line"><a id="l00064" name="l00064"></a><span class="lineno"> 64</span> dst[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> + j] = <span class="keyword">static_cast<</span>T<span class="keyword">></span>(src[i * str_x + j * str_y]);</div>
|
||||
<div class="line"><a id="l00065" name="l00065"></a><span class="lineno"> 65</span> }</div>
|
||||
<div class="line"><a id="l00066" name="l00066"></a><span class="lineno"> 66</span> }</div>
|
||||
<div class="line"><a id="l00067" name="l00067"></a><span class="lineno"> 67</span> }</div>
|
||||
@@ -173,15 +187,15 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00075" name="l00075"></a><span class="lineno"> 75</span> <span class="keyword">typename</span> OffX,</div>
|
||||
<div class="line"><a id="l00076" name="l00076"></a><span class="lineno"> 76</span> <span class="keyword">typename</span> OffY></div>
|
||||
<div class="foldopen" id="foldopen00077" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00077" name="l00077"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#ad22aaee4a2938cbdd315b39eda84e07d"> 77</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#ad22aaee4a2938cbdd315b39eda84e07d">load_safe</a>(</div>
|
||||
<div class="line"><a id="l00078" name="l00078"></a><span class="lineno"> 78</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>& dst,</div>
|
||||
<div class="line"><a id="l00077" name="l00077"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#ad22aaee4a2938cbdd315b39eda84e07d"> 77</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#ad22aaee4a2938cbdd315b39eda84e07d">load_safe</a>(</div>
|
||||
<div class="line"><a id="l00078" name="l00078"></a><span class="lineno"> 78</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& dst,</div>
|
||||
<div class="line"><a id="l00079" name="l00079"></a><span class="lineno"> 79</span> SrcPtrType src,</div>
|
||||
<div class="line"><a id="l00080" name="l00080"></a><span class="lineno"> 80</span> StrX str_x,</div>
|
||||
<div class="line"><a id="l00081" name="l00081"></a><span class="lineno"> 81</span> StrY str_y,</div>
|
||||
<div class="line"><a id="l00082" name="l00082"></a><span class="lineno"> 82</span> LimX lim_x,</div>
|
||||
<div class="line"><a id="l00083" name="l00083"></a><span class="lineno"> 83</span> LimY lim_y,</div>
|
||||
<div class="line"><a id="l00084" name="l00084"></a><span class="lineno"> 84</span> OffX off_x = <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<0></a>{},</div>
|
||||
<div class="line"><a id="l00085" name="l00085"></a><span class="lineno"> 85</span> OffY off_y = <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<0></a>{}) {</div>
|
||||
<div class="line"><a id="l00084" name="l00084"></a><span class="lineno"> 84</span> OffX off_x = <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<0></a>{},</div>
|
||||
<div class="line"><a id="l00085" name="l00085"></a><span class="lineno"> 85</span> OffY off_y = <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<0></a>{}) {</div>
|
||||
<div class="line"><a id="l00086" name="l00086"></a><span class="lineno"> 86</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00087" name="l00087"></a><span class="lineno"> 87</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < kElemRows; i++) {</div>
|
||||
<div class="line"><a id="l00088" name="l00088"></a><span class="lineno"> 88</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
@@ -200,14 +214,14 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00100" name="l00100"></a><span class="lineno"> 100</span> <span class="keyword">template</span> <<span class="keyword">typename</span> DstPtrType, <span class="keyword">typename</span> StrX, <span class="keyword">typename</span> StrY></div>
|
||||
<div class="line"><a id="l00101" name="l00101"></a><span class="lineno"> 101</span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span></div>
|
||||
<div class="foldopen" id="foldopen00102" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00102" name="l00102"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#aa8f50ea8961ec5b35c1b81366d64f2cb"> 102</a></span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#aa8f50ea8961ec5b35c1b81366d64f2cb">store</a>(<span class="keyword">const</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>& src, DstPtrType dst, StrX str_x, StrY str_y) {</div>
|
||||
<div class="line"><a id="l00102" name="l00102"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#aa8f50ea8961ec5b35c1b81366d64f2cb"> 102</a></span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#aa8f50ea8961ec5b35c1b81366d64f2cb">store</a>(<span class="keyword">const</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& src, DstPtrType dst, StrX str_x, StrY str_y) {</div>
|
||||
<div class="line"><a id="l00103" name="l00103"></a><span class="lineno"> 103</span> <span class="keyword">using </span>U = <a class="code hl_typedef" href="namespacemetal.html#ac82ee6c3fbe9ec5c78c07329424aaec9">pointer_element_t<DstPtrType></a>;</div>
|
||||
<div class="line"><a id="l00104" name="l00104"></a><span class="lineno"> 104</span> </div>
|
||||
<div class="line"><a id="l00105" name="l00105"></a><span class="lineno"> 105</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00106" name="l00106"></a><span class="lineno"> 106</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < kElemRows; i++) {</div>
|
||||
<div class="line"><a id="l00106" name="l00106"></a><span class="lineno"> 106</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a>; i++) {</div>
|
||||
<div class="line"><a id="l00107" name="l00107"></a><span class="lineno"> 107</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00108" name="l00108"></a><span class="lineno"> 108</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> j = 0; j < kElemCols; j++) {</div>
|
||||
<div class="line"><a id="l00109" name="l00109"></a><span class="lineno"> 109</span> dst[i * str_x + j * str_y] = <span class="keyword">static_cast<</span>U<span class="keyword">></span>(src[i * kElemCols + j]);</div>
|
||||
<div class="line"><a id="l00108" name="l00108"></a><span class="lineno"> 108</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> j = 0; j < <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a>; j++) {</div>
|
||||
<div class="line"><a id="l00109" name="l00109"></a><span class="lineno"> 109</span> dst[i * str_x + j * str_y] = <span class="keyword">static_cast<</span>U<span class="keyword">></span>(src[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> + j]);</div>
|
||||
<div class="line"><a id="l00110" name="l00110"></a><span class="lineno"> 110</span> }</div>
|
||||
<div class="line"><a id="l00111" name="l00111"></a><span class="lineno"> 111</span> }</div>
|
||||
<div class="line"><a id="l00112" name="l00112"></a><span class="lineno"> 112</span> }</div>
|
||||
@@ -222,15 +236,15 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00120" name="l00120"></a><span class="lineno"> 120</span> <span class="keyword">typename</span> OffX,</div>
|
||||
<div class="line"><a id="l00121" name="l00121"></a><span class="lineno"> 121</span> <span class="keyword">typename</span> OffY></div>
|
||||
<div class="foldopen" id="foldopen00122" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00122" name="l00122"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a1f0b00daad8eba2f855bb306e70d2328"> 122</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a1f0b00daad8eba2f855bb306e70d2328">store_safe</a>(</div>
|
||||
<div class="line"><a id="l00123" name="l00123"></a><span class="lineno"> 123</span> <span class="keyword">const</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>& src,</div>
|
||||
<div class="line"><a id="l00122" name="l00122"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a1f0b00daad8eba2f855bb306e70d2328"> 122</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a1f0b00daad8eba2f855bb306e70d2328">store_safe</a>(</div>
|
||||
<div class="line"><a id="l00123" name="l00123"></a><span class="lineno"> 123</span> <span class="keyword">const</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& src,</div>
|
||||
<div class="line"><a id="l00124" name="l00124"></a><span class="lineno"> 124</span> DstPtrType dst,</div>
|
||||
<div class="line"><a id="l00125" name="l00125"></a><span class="lineno"> 125</span> StrX str_x,</div>
|
||||
<div class="line"><a id="l00126" name="l00126"></a><span class="lineno"> 126</span> StrY str_y,</div>
|
||||
<div class="line"><a id="l00127" name="l00127"></a><span class="lineno"> 127</span> LimX lim_x,</div>
|
||||
<div class="line"><a id="l00128" name="l00128"></a><span class="lineno"> 128</span> LimY lim_y,</div>
|
||||
<div class="line"><a id="l00129" name="l00129"></a><span class="lineno"> 129</span> OffX off_x = <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<0></a>{},</div>
|
||||
<div class="line"><a id="l00130" name="l00130"></a><span class="lineno"> 130</span> OffY off_y = <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<0></a>{}) {</div>
|
||||
<div class="line"><a id="l00129" name="l00129"></a><span class="lineno"> 129</span> OffX off_x = <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<0></a>{},</div>
|
||||
<div class="line"><a id="l00130" name="l00130"></a><span class="lineno"> 130</span> OffY off_y = <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<0></a>{}) {</div>
|
||||
<div class="line"><a id="l00131" name="l00131"></a><span class="lineno"> 131</span> <span class="keyword">using </span>U = <a class="code hl_typedef" href="namespacemetal.html#ac82ee6c3fbe9ec5c78c07329424aaec9">pointer_element_t<DstPtrType></a>;</div>
|
||||
<div class="line"><a id="l00132" name="l00132"></a><span class="lineno"> 132</span> </div>
|
||||
<div class="line"><a id="l00133" name="l00133"></a><span class="lineno"> 133</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
@@ -247,32 +261,32 @@ $(function(){ initResizable(false); });
|
||||
</div>
|
||||
<div class="line"><a id="l00144" name="l00144"></a><span class="lineno"> 144</span> </div>
|
||||
<div class="foldopen" id="foldopen00145" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00145" name="l00145"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a8028512f5a3d2b6acaf966be529627a3"> 145</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a8028512f5a3d2b6acaf966be529627a3">mma</a>(</div>
|
||||
<div class="line"><a id="l00146" name="l00146"></a><span class="lineno"> 146</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>& D,</div>
|
||||
<div class="line"><a id="l00147" name="l00147"></a><span class="lineno"> 147</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>& A,</div>
|
||||
<div class="line"><a id="l00148" name="l00148"></a><span class="lineno"> 148</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>& B,</div>
|
||||
<div class="line"><a id="l00149" name="l00149"></a><span class="lineno"> 149</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>& C) {</div>
|
||||
<div class="line"><a id="l00150" name="l00150"></a><span class="lineno"> 150</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a> D_mat;</div>
|
||||
<div class="line"><a id="l00151" name="l00151"></a><span class="lineno"> 151</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a> A_mat;</div>
|
||||
<div class="line"><a id="l00152" name="l00152"></a><span class="lineno"> 152</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a> B_mat;</div>
|
||||
<div class="line"><a id="l00153" name="l00153"></a><span class="lineno"> 153</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a> C_mat;</div>
|
||||
<div class="line"><a id="l00145" name="l00145"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8028512f5a3d2b6acaf966be529627a3"> 145</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8028512f5a3d2b6acaf966be529627a3">mma</a>(</div>
|
||||
<div class="line"><a id="l00146" name="l00146"></a><span class="lineno"> 146</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& D,</div>
|
||||
<div class="line"><a id="l00147" name="l00147"></a><span class="lineno"> 147</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& A,</div>
|
||||
<div class="line"><a id="l00148" name="l00148"></a><span class="lineno"> 148</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& B,</div>
|
||||
<div class="line"><a id="l00149" name="l00149"></a><span class="lineno"> 149</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& C) {</div>
|
||||
<div class="line"><a id="l00150" name="l00150"></a><span class="lineno"> 150</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a> D_mat;</div>
|
||||
<div class="line"><a id="l00151" name="l00151"></a><span class="lineno"> 151</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a> A_mat;</div>
|
||||
<div class="line"><a id="l00152" name="l00152"></a><span class="lineno"> 152</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a> B_mat;</div>
|
||||
<div class="line"><a id="l00153" name="l00153"></a><span class="lineno"> 153</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a> C_mat;</div>
|
||||
<div class="line"><a id="l00154" name="l00154"></a><span class="lineno"> 154</span> </div>
|
||||
<div class="line"><a id="l00155" name="l00155"></a><span class="lineno"> 155</span> <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>&<span class="keyword">></span>(A_mat.thread_elements()) = A;</div>
|
||||
<div class="line"><a id="l00156" name="l00156"></a><span class="lineno"> 156</span> <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>&<span class="keyword">></span>(B_mat.thread_elements()) = B;</div>
|
||||
<div class="line"><a id="l00157" name="l00157"></a><span class="lineno"> 157</span> <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>&<span class="keyword">></span>(C_mat.thread_elements()) = C;</div>
|
||||
<div class="line"><a id="l00155" name="l00155"></a><span class="lineno"> 155</span> <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>&<span class="keyword">></span>(A_mat.thread_elements()) = A;</div>
|
||||
<div class="line"><a id="l00156" name="l00156"></a><span class="lineno"> 156</span> <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>&<span class="keyword">></span>(B_mat.thread_elements()) = B;</div>
|
||||
<div class="line"><a id="l00157" name="l00157"></a><span class="lineno"> 157</span> <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>&<span class="keyword">></span>(C_mat.thread_elements()) = C;</div>
|
||||
<div class="line"><a id="l00158" name="l00158"></a><span class="lineno"> 158</span> </div>
|
||||
<div class="line"><a id="l00159" name="l00159"></a><span class="lineno"> 159</span> mma(D_mat, A_mat, B_mat, C_mat);</div>
|
||||
<div class="line"><a id="l00159" name="l00159"></a><span class="lineno"> 159</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8028512f5a3d2b6acaf966be529627a3">mma</a>(D_mat, A_mat, B_mat, C_mat);</div>
|
||||
<div class="line"><a id="l00160" name="l00160"></a><span class="lineno"> 160</span> </div>
|
||||
<div class="line"><a id="l00161" name="l00161"></a><span class="lineno"> 161</span> D = <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">frag_type</a>&<span class="keyword">></span>(D_mat.thread_elements());</div>
|
||||
<div class="line"><a id="l00161" name="l00161"></a><span class="lineno"> 161</span> D = <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>&<span class="keyword">></span>(D_mat.thread_elements());</div>
|
||||
<div class="line"><a id="l00162" name="l00162"></a><span class="lineno"> 162</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00163" name="l00163"></a><span class="lineno"> 163</span> </div>
|
||||
<div class="foldopen" id="foldopen00164" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00164" name="l00164"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a1868f57d57c8adedab2c58492ec76946"> 164</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a1868f57d57c8adedab2c58492ec76946">mma</a>(</div>
|
||||
<div class="line"><a id="l00165" name="l00165"></a><span class="lineno"> 165</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a>& D,</div>
|
||||
<div class="line"><a id="l00166" name="l00166"></a><span class="lineno"> 166</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a>& A,</div>
|
||||
<div class="line"><a id="l00167" name="l00167"></a><span class="lineno"> 167</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a>& B,</div>
|
||||
<div class="line"><a id="l00168" name="l00168"></a><span class="lineno"> 168</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mat_type</a>& C) {</div>
|
||||
<div class="line"><a id="l00164" name="l00164"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a1868f57d57c8adedab2c58492ec76946"> 164</a></span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a1868f57d57c8adedab2c58492ec76946">mma</a>(</div>
|
||||
<div class="line"><a id="l00165" name="l00165"></a><span class="lineno"> 165</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a>& D,</div>
|
||||
<div class="line"><a id="l00166" name="l00166"></a><span class="lineno"> 166</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a>& A,</div>
|
||||
<div class="line"><a id="l00167" name="l00167"></a><span class="lineno"> 167</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a>& B,</div>
|
||||
<div class="line"><a id="l00168" name="l00168"></a><span class="lineno"> 168</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mat_type</a>& C) {</div>
|
||||
<div class="line"><a id="l00169" name="l00169"></a><span class="lineno"> 169</span> simdgroup_multiply_accumulate(D, A, B, C);</div>
|
||||
<div class="line"><a id="l00170" name="l00170"></a><span class="lineno"> 170</span> }</div>
|
||||
</div>
|
||||
@@ -285,9 +299,9 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00176" name="l00176"></a><span class="lineno"> 176</span> <span class="keywordtype">int</span> kTileCols_,</div>
|
||||
<div class="line"><a id="l00177" name="l00177"></a><span class="lineno"> 177</span> <span class="keyword">class </span>MMAFrag_ = BaseMMAFrag<T, 8, 8>></div>
|
||||
<div class="foldopen" id="foldopen00178" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00178" name="l00178"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html"> 178</a></span><span class="keyword">struct </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_m_m_a_tile.html">MMATile</a> {</div>
|
||||
<div class="line"><a id="l00179" name="l00179"></a><span class="lineno"> 179</span> <span class="keyword">using </span><a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a6dadcd666afb3759a11094e754560dd4">MMAFrag_t</a> = MMAFrag_;</div>
|
||||
<div class="line"><a id="l00180" name="l00180"></a><span class="lineno"> 180</span> <span class="keyword">using </span><a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a824409bc107330805853f932e80a7628">elem_type</a> = T;</div>
|
||||
<div class="line"><a id="l00178" name="l00178"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html"> 178</a></span><span class="keyword">struct </span><a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#aa3fb310dd08ec23c334511f7b316d1b6">MMATile</a> {</div>
|
||||
<div class="line"><a id="l00179" name="l00179"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#abe33de70e34300745bad9aa822fd0382"> 179</a></span> <span class="keyword">using </span><a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#abe33de70e34300745bad9aa822fd0382">MMAFrag_t</a> = MMAFrag_;</div>
|
||||
<div class="line"><a id="l00180" name="l00180"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a80078f0dfa4c225e79d9b460202d5e2c"> 180</a></span> <span class="keyword">using </span><a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a80078f0dfa4c225e79d9b460202d5e2c">elem_type</a> = T;</div>
|
||||
<div class="line"><a id="l00181" name="l00181"></a><span class="lineno"> 181</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7">kFragRows</a> = MMAFrag_t::kFragRows;</div>
|
||||
<div class="line"><a id="l00182" name="l00182"></a><span class="lineno"> 182</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ad095371db98e7c335ec41ca77c10f906">kFragCols</a> = MMAFrag_t::kFragCols;</div>
|
||||
<div class="line"><a id="l00183" name="l00183"></a><span class="lineno"> 183</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#aef0ea2387e1ff5767bff8563b2d36bd6">kElemsPerFrag</a> = MMAFrag_t::kElemsPerFrag;</div>
|
||||
@@ -301,8 +315,8 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00191" name="l00191"></a><span class="lineno"> 191</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae326e7693eb77c22d5a6e3e9219019d3">kNumFrags</a> = <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1d126b14910385ab644e224ac1d0307a">kTileRows</a> * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a46324d40f8ad61cade08a1ebad6d9ad4">kTileCols</a>;</div>
|
||||
<div class="line"><a id="l00192" name="l00192"></a><span class="lineno"> 192</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">int</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a98357339ec98f804a1b12597937b318f">kElemsPerTile</a> = <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae326e7693eb77c22d5a6e3e9219019d3">kNumFrags</a> * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#aef0ea2387e1ff5767bff8563b2d36bd6">kElemsPerFrag</a>;</div>
|
||||
<div class="line"><a id="l00193" name="l00193"></a><span class="lineno"> 193</span> </div>
|
||||
<div class="line"><a id="l00194" name="l00194"></a><span class="lineno"> 194</span> <span class="keyword">typedef</span> <span class="keyword">typename</span> MMAFrag_t::mat_type <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1eeb197c9bdf4db42892a39cdb9bd73a">mat_type</a>;</div>
|
||||
<div class="line"><a id="l00195" name="l00195"></a><span class="lineno"> 195</span> <span class="keyword">typedef</span> <span class="keyword">typename</span> MMAFrag_t::frag_type <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#aac25cd0a9bdf24aa2af809c95f0bd171">frag_type</a>;</div>
|
||||
<div class="line"><a id="l00194" name="l00194"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1eeb197c9bdf4db42892a39cdb9bd73a"> 194</a></span> <span class="keyword">typedef</span> <span class="keyword">typename</span> MMAFrag_t::mat_type <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1eeb197c9bdf4db42892a39cdb9bd73a">mat_type</a>;</div>
|
||||
<div class="line"><a id="l00195" name="l00195"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#aac25cd0a9bdf24aa2af809c95f0bd171"> 195</a></span> <span class="keyword">typedef</span> <span class="keyword">typename</span> MMAFrag_t::frag_type <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#aac25cd0a9bdf24aa2af809c95f0bd171">frag_type</a>;</div>
|
||||
<div class="line"><a id="l00196" name="l00196"></a><span class="lineno"> 196</span> </div>
|
||||
<div class="line"><a id="l00197" name="l00197"></a><span class="lineno"> 197</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#aac25cd0a9bdf24aa2af809c95f0bd171">frag_type</a> <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a684e6c6d9f00f583994285b60aaa3b62">val_frags</a>[<a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae326e7693eb77c22d5a6e3e9219019d3">kNumFrags</a>] = {<a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#aac25cd0a9bdf24aa2af809c95f0bd171">frag_type</a>(0)};</div>
|
||||
<div class="line"><a id="l00198" name="l00198"></a><span class="lineno"> 198</span> </div>
|
||||
@@ -343,14 +357,14 @@ $(function(){ initResizable(false); });
|
||||
</div>
|
||||
<div class="line"><a id="l00226" name="l00226"></a><span class="lineno"> 226</span> </div>
|
||||
<div class="foldopen" id="foldopen00227" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00227" name="l00227"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a865ece5ad0b9a56937b6d77a18b5a1dc"> 227</a></span> METAL_FUNC thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a824409bc107330805853f932e80a7628">elem_type</a>* <a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a865ece5ad0b9a56937b6d77a18b5a1dc">elems</a>() {</div>
|
||||
<div class="line"><a id="l00228" name="l00228"></a><span class="lineno"> 228</span> <span class="keywordflow">return</span> <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a824409bc107330805853f932e80a7628">elem_type</a>*<span class="keyword">></span>(<a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a684e6c6d9f00f583994285b60aaa3b62">val_frags</a>);</div>
|
||||
<div class="line"><a id="l00227" name="l00227"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a865ece5ad0b9a56937b6d77a18b5a1dc"> 227</a></span> METAL_FUNC thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a80078f0dfa4c225e79d9b460202d5e2c">elem_type</a>* <a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a865ece5ad0b9a56937b6d77a18b5a1dc">elems</a>() {</div>
|
||||
<div class="line"><a id="l00228" name="l00228"></a><span class="lineno"> 228</span> <span class="keywordflow">return</span> <span class="keyword">reinterpret_cast<</span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a80078f0dfa4c225e79d9b460202d5e2c">elem_type</a>*<span class="keyword">></span>(<a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a684e6c6d9f00f583994285b60aaa3b62">val_frags</a>);</div>
|
||||
<div class="line"><a id="l00229" name="l00229"></a><span class="lineno"> 229</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00230" name="l00230"></a><span class="lineno"> 230</span> </div>
|
||||
<div class="foldopen" id="foldopen00231" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00231" name="l00231"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae21bb7cce701290de84c6015e064d8a1"> 231</a></span> METAL_FUNC <span class="keyword">const</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a824409bc107330805853f932e80a7628">elem_type</a>* <a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae21bb7cce701290de84c6015e064d8a1">elems</a>()<span class="keyword"> const </span>{</div>
|
||||
<div class="line"><a id="l00232" name="l00232"></a><span class="lineno"> 232</span> <span class="keywordflow">return</span> <span class="keyword">reinterpret_cast<</span><span class="keyword">const </span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a824409bc107330805853f932e80a7628">elem_type</a>*<span class="keyword">></span>(<a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a684e6c6d9f00f583994285b60aaa3b62">val_frags</a>);</div>
|
||||
<div class="line"><a id="l00231" name="l00231"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae21bb7cce701290de84c6015e064d8a1"> 231</a></span> METAL_FUNC <span class="keyword">const</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a80078f0dfa4c225e79d9b460202d5e2c">elem_type</a>* <a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae21bb7cce701290de84c6015e064d8a1">elems</a>()<span class="keyword"> const </span>{</div>
|
||||
<div class="line"><a id="l00232" name="l00232"></a><span class="lineno"> 232</span> <span class="keywordflow">return</span> <span class="keyword">reinterpret_cast<</span><span class="keyword">const </span>thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a80078f0dfa4c225e79d9b460202d5e2c">elem_type</a>*<span class="keyword">></span>(<a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a684e6c6d9f00f583994285b60aaa3b62">val_frags</a>);</div>
|
||||
<div class="line"><a id="l00233" name="l00233"></a><span class="lineno"> 233</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00234" name="l00234"></a><span class="lineno"> 234</span> </div>
|
||||
@@ -366,8 +380,8 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00243" name="l00243"></a><span class="lineno"> 243</span> &(</div>
|
||||
<div class="line"><a id="l00244" name="l00244"></a><span class="lineno"> 244</span> src[(i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7">kFragRows</a>) * w_x * str_x +</div>
|
||||
<div class="line"><a id="l00245" name="l00245"></a><span class="lineno"> 245</span> (j * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ad095371db98e7c335ec41ca77c10f906">kFragCols</a>) * w_y * str_y]),</div>
|
||||
<div class="line"><a id="l00246" name="l00246"></a><span class="lineno"> 246</span> <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<str_x></a>{},</div>
|
||||
<div class="line"><a id="l00247" name="l00247"></a><span class="lineno"> 247</span> <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<str_y></a>{});</div>
|
||||
<div class="line"><a id="l00246" name="l00246"></a><span class="lineno"> 246</span> <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<str_x></a>{},</div>
|
||||
<div class="line"><a id="l00247" name="l00247"></a><span class="lineno"> 247</span> <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<str_y></a>{});</div>
|
||||
<div class="line"><a id="l00248" name="l00248"></a><span class="lineno"> 248</span> }</div>
|
||||
<div class="line"><a id="l00249" name="l00249"></a><span class="lineno"> 249</span> }</div>
|
||||
<div class="line"><a id="l00250" name="l00250"></a><span class="lineno"> 250</span> }</div>
|
||||
@@ -385,8 +399,8 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00260" name="l00260"></a><span class="lineno"> 260</span> &(</div>
|
||||
<div class="line"><a id="l00261" name="l00261"></a><span class="lineno"> 261</span> dst[(i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7">kFragRows</a>) * w_x * str_x +</div>
|
||||
<div class="line"><a id="l00262" name="l00262"></a><span class="lineno"> 262</span> (j * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ad095371db98e7c335ec41ca77c10f906">kFragCols</a>) * w_y * str_y]),</div>
|
||||
<div class="line"><a id="l00263" name="l00263"></a><span class="lineno"> 263</span> <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<str_x></a>{},</div>
|
||||
<div class="line"><a id="l00264" name="l00264"></a><span class="lineno"> 264</span> <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<str_y></a>{});</div>
|
||||
<div class="line"><a id="l00263" name="l00263"></a><span class="lineno"> 263</span> <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<str_x></a>{},</div>
|
||||
<div class="line"><a id="l00264" name="l00264"></a><span class="lineno"> 264</span> <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<str_y></a>{});</div>
|
||||
<div class="line"><a id="l00265" name="l00265"></a><span class="lineno"> 265</span> }</div>
|
||||
<div class="line"><a id="l00266" name="l00266"></a><span class="lineno"> 266</span> }</div>
|
||||
<div class="line"><a id="l00267" name="l00267"></a><span class="lineno"> 267</span> }</div>
|
||||
@@ -403,7 +417,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00276" name="l00276"></a><span class="lineno"> 276</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1a6b1446e8c8da46885bbaa8e8fdc7e4">frag_at</a>(i, j),</div>
|
||||
<div class="line"><a id="l00277" name="l00277"></a><span class="lineno"> 277</span> &(src[(i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7">kFragRows</a>) * w_x * ld + (j * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ad095371db98e7c335ec41ca77c10f906">kFragCols</a>) * w_y]),</div>
|
||||
<div class="line"><a id="l00278" name="l00278"></a><span class="lineno"> 278</span> ld,</div>
|
||||
<div class="line"><a id="l00279" name="l00279"></a><span class="lineno"> 279</span> <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<1></a>{});</div>
|
||||
<div class="line"><a id="l00279" name="l00279"></a><span class="lineno"> 279</span> <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<1></a>{});</div>
|
||||
<div class="line"><a id="l00280" name="l00280"></a><span class="lineno"> 280</span> }</div>
|
||||
<div class="line"><a id="l00281" name="l00281"></a><span class="lineno"> 281</span> }</div>
|
||||
<div class="line"><a id="l00282" name="l00282"></a><span class="lineno"> 282</span> }</div>
|
||||
@@ -420,7 +434,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00291" name="l00291"></a><span class="lineno"> 291</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1a6b1446e8c8da46885bbaa8e8fdc7e4">frag_at</a>(i, j),</div>
|
||||
<div class="line"><a id="l00292" name="l00292"></a><span class="lineno"> 292</span> &(dst[(i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7">kFragRows</a>) * w_x * ld + (j * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#ad095371db98e7c335ec41ca77c10f906">kFragCols</a>) * w_y]),</div>
|
||||
<div class="line"><a id="l00293" name="l00293"></a><span class="lineno"> 293</span> ld,</div>
|
||||
<div class="line"><a id="l00294" name="l00294"></a><span class="lineno"> 294</span> <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<1></a>{});</div>
|
||||
<div class="line"><a id="l00294" name="l00294"></a><span class="lineno"> 294</span> <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<1></a>{});</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> }</div>
|
||||
<div class="line"><a id="l00297" name="l00297"></a><span class="lineno"> 297</span> }</div>
|
||||
@@ -438,7 +452,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00307" name="l00307"></a><span class="lineno"> 307</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1a6b1446e8c8da46885bbaa8e8fdc7e4">frag_at</a>(i, j),</div>
|
||||
<div class="line"><a id="l00308" name="l00308"></a><span class="lineno"> 308</span> src,</div>
|
||||
<div class="line"><a id="l00309" name="l00309"></a><span class="lineno"> 309</span> ld,</div>
|
||||
<div class="line"><a id="l00310" name="l00310"></a><span class="lineno"> 310</span> <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<1></a>{},</div>
|
||||
<div class="line"><a id="l00310" name="l00310"></a><span class="lineno"> 310</span> <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<1></a>{},</div>
|
||||
<div class="line"><a id="l00311" name="l00311"></a><span class="lineno"> 311</span> src_tile_dims.y,</div>
|
||||
<div class="line"><a id="l00312" name="l00312"></a><span class="lineno"> 312</span> src_tile_dims.x,</div>
|
||||
<div class="line"><a id="l00313" name="l00313"></a><span class="lineno"> 313</span> (i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7">kFragRows</a>) * w_x,</div>
|
||||
@@ -460,7 +474,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00327" name="l00327"></a><span class="lineno"> 327</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1a6b1446e8c8da46885bbaa8e8fdc7e4">frag_at</a>(i, j),</div>
|
||||
<div class="line"><a id="l00328" name="l00328"></a><span class="lineno"> 328</span> dst,</div>
|
||||
<div class="line"><a id="l00329" name="l00329"></a><span class="lineno"> 329</span> ld,</div>
|
||||
<div class="line"><a id="l00330" name="l00330"></a><span class="lineno"> 330</span> <a class="code hl_struct" href="structmlx_1_1steel_1_1integral__constant.html">Int<1></a>{},</div>
|
||||
<div class="line"><a id="l00330" name="l00330"></a><span class="lineno"> 330</span> <a class="code hl_typedef" href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">Int<1></a>{},</div>
|
||||
<div class="line"><a id="l00331" name="l00331"></a><span class="lineno"> 331</span> dst_tile_dims.y,</div>
|
||||
<div class="line"><a id="l00332" name="l00332"></a><span class="lineno"> 332</span> dst_tile_dims.x,</div>
|
||||
<div class="line"><a id="l00333" name="l00333"></a><span class="lineno"> 333</span> (i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7">kFragRows</a>) * w_x,</div>
|
||||
@@ -508,12 +522,12 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00373" name="l00373"></a><span class="lineno"> 373</span> <span class="keywordtype">short</span> lda_tgp,</div>
|
||||
<div class="line"><a id="l00374" name="l00374"></a><span class="lineno"> 374</span> <span class="keywordtype">short</span> ldb_tgp,</div>
|
||||
<div class="line"><a id="l00375" name="l00375"></a><span class="lineno"> 375</span> <span class="keyword">typename</span> AccumType = float,</div>
|
||||
<div class="line"><a id="l00376" name="l00376"></a><span class="lineno"> 376</span> <span class="keyword">typename</span> Epilogue = TransformNone<U, AccumType>></div>
|
||||
<div class="line"><a id="l00376" name="l00376"></a><span class="lineno"> 376</span> <span class="keyword">typename</span> Epilogue = <a class="code hl_struct" href="structmlx_1_1steel_1_1_transform_none.html">TransformNone<U, AccumType></a>></div>
|
||||
<div class="foldopen" id="foldopen00377" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00377" name="l00377"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_m_m_a.html"> 377</a></span><span class="keyword">struct </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_block_m_m_a.html">BlockMMA</a> {</div>
|
||||
<div class="line"><a id="l00377" name="l00377"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_m_m_a.html"> 377</a></span><span class="keyword">struct </span><a class="code hl_function" href="structmlx_1_1steel_1_1_block_m_m_a.html#aa14406b7298456ac45d23dd3c4642dd8">BlockMMA</a> {</div>
|
||||
<div class="line"><a id="l00378" name="l00378"></a><span class="lineno"> 378</span> <span class="comment">// MMAFrag size</span></div>
|
||||
<div class="line"><a id="l00379" name="l00379"></a><span class="lineno"> 379</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">short</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_m_m_a.html#aee8caec45c1f9e4428586effbfe6137d">kFragSize</a> = 8;</div>
|
||||
<div class="line"><a id="l00380" name="l00380"></a><span class="lineno"> 380</span> <span class="keyword">using </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html">MMAFrag_acc_t</a> = <a class="code hl_struct" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html">BaseMMAFrag<AccumType, kFragSize, kFragSize></a>;</div>
|
||||
<div class="line"><a id="l00380" name="l00380"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_m_m_a.html#ae2c42cb6d0dde785859164c195f4d13c"> 380</a></span> <span class="keyword">using </span><a class="code hl_typedef" href="structmlx_1_1steel_1_1_block_m_m_a.html#ae2c42cb6d0dde785859164c195f4d13c">MMAFrag_acc_t</a> = <a class="code hl_struct" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html">BaseMMAFrag<AccumType, kFragSize, kFragSize></a>;</div>
|
||||
<div class="line"><a id="l00381" name="l00381"></a><span class="lineno"> 381</span> </div>
|
||||
<div class="line"><a id="l00382" name="l00382"></a><span class="lineno"> 382</span> <span class="comment">// Warp tile simdgroup matrix strides along M</span></div>
|
||||
<div class="line"><a id="l00383" name="l00383"></a><span class="lineno"> 383</span> <a class="code hl_define" href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a> <span class="keywordtype">short</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_m_m_a.html#a5b0029866f493363942133b55bff7307">TM_stride</a> = <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_m_m_a.html#aee8caec45c1f9e4428586effbfe6137d">kFragSize</a> * WM;</div>
|
||||
@@ -811,47 +825,54 @@ $(function(){ initResizable(false); });
|
||||
<div class="ttc" id="aintegral__constant_8h_html"><div class="ttname"><a href="integral__constant_8h.html">integral_constant.h</a></div></div>
|
||||
<div class="ttc" id="anamespacemetal_html"><div class="ttname"><a href="namespacemetal.html">metal</a></div><div class="ttdef"><b>Definition</b> bf16_math.h:226</div></div>
|
||||
<div class="ttc" id="anamespacemetal_html_ac82ee6c3fbe9ec5c78c07329424aaec9"><div class="ttname"><a href="namespacemetal.html#ac82ee6c3fbe9ec5c78c07329424aaec9">metal::pointer_element_t</a></div><div class="ttdeci">typename pointer_element< remove_cv_t< T > >::type pointer_element_t</div><div class="ttdef"><b>Definition</b> type_traits.h:51</div></div>
|
||||
<div class="ttc" id="anamespacemlx_1_1steel_html"><div class="ttname"><a href="namespacemlx_1_1steel.html">mlx::steel</a></div><div class="ttdef"><b>Definition</b> attn.h:19</div></div>
|
||||
<div class="ttc" id="anamespacemlx_1_1steel_html_ad583e6038efc119542410f43b603d4ad"><div class="ttname"><a href="namespacemlx_1_1steel.html#ad583e6038efc119542410f43b603d4ad">mlx::steel::tile_matmad</a></div><div class="ttdeci">METAL_FUNC void tile_matmad(thread MMATile< T, M, N > &D, thread MMATile< U, M, K > &A, thread MMATile< U, K, N > &B, thread MMATile< T, M, N > &C)</div><div class="ttdef"><b>Definition</b> mma.h:413</div></div>
|
||||
<div class="ttc" id="anamespacemlx_1_1steel_html_afe36ddf6725498d273e5eef4f1579891"><div class="ttname"><a href="namespacemlx_1_1steel.html#afe36ddf6725498d273e5eef4f1579891">mlx::steel::Int</a></div><div class="ttdeci">integral_constant< int, val > Int</div><div class="ttdef"><b>Definition</b> integral_constant.h:48</div></div>
|
||||
<div class="ttc" id="anamespacemlx_html"><div class="ttname"><a href="namespacemlx.html">mlx</a></div><div class="ttdef"><b>Definition</b> allocator.h:7</div></div>
|
||||
<div class="ttc" id="asteel_2defines_8h_html"><div class="ttname"><a href="steel_2defines_8h.html">defines.h</a></div></div>
|
||||
<div class="ttc" id="asteel_2defines_8h_html_a5a5c3095b132a7589bc19cd5cb80e2c6"><div class="ttname"><a href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div><div class="ttdeci">#define STEEL_PRAGMA_UNROLL</div><div class="ttdef"><b>Definition</b> defines.h:4</div></div>
|
||||
<div class="ttc" id="asteel_2defines_8h_html_a90b91c866313ffa46eff6d9cc944ad2b"><div class="ttname"><a href="steel_2defines_8h.html#a90b91c866313ffa46eff6d9cc944ad2b">STEEL_CONST</a></div><div class="ttdeci">#define STEEL_CONST</div><div class="ttdef"><b>Definition</b> defines.h:3</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_a1868f57d57c8adedab2c58492ec76946"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a1868f57d57c8adedab2c58492ec76946">mlx::steel::BaseMMAFrag< T, 8, 8 >::mma</a></div><div class="ttdeci">static METAL_FUNC constexpr void mma(thread mat_type &D, thread mat_type &A, thread mat_type &B, thread mat_type &C)</div><div class="ttdef"><b>Definition</b> mma.h:164</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_a1f0b00daad8eba2f855bb306e70d2328"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a1f0b00daad8eba2f855bb306e70d2328">mlx::steel::BaseMMAFrag< T, 8, 8 >::store_safe</a></div><div class="ttdeci">static METAL_FUNC constexpr void store_safe(const thread frag_type &src, DstPtrType dst, StrX str_x, StrY str_y, LimX lim_x, LimY lim_y, OffX off_x=Int< 0 >{}, OffY off_y=Int< 0 >{})</div><div class="ttdef"><b>Definition</b> mma.h:122</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_a7331fff1d12f2f8b72b0006a3ad0dd83"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a7331fff1d12f2f8b72b0006a3ad0dd83">mlx::steel::BaseMMAFrag< T, 8, 8 >::get_coord</a></div><div class="ttdeci">static METAL_FUNC constexpr short2 get_coord(ushort simd_lane_id)</div><div class="ttdef"><b>Definition</b> mma.h:49</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_a8028512f5a3d2b6acaf966be529627a3"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a8028512f5a3d2b6acaf966be529627a3">mlx::steel::BaseMMAFrag< T, 8, 8 >::mma</a></div><div class="ttdeci">static METAL_FUNC constexpr void mma(thread frag_type &D, thread frag_type &A, thread frag_type &B, thread frag_type &C)</div><div class="ttdef"><b>Definition</b> mma.h:145</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_a958b6952cbd9462d7ae9f6e029631887"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a958b6952cbd9462d7ae9f6e029631887">mlx::steel::BaseMMAFrag< T, 8, 8 >::mat_type</a></div><div class="ttdeci">metal::simdgroup_matrix< T, kFragRows, kFragCols > mat_type</div><div class="ttdef"><b>Definition</b> mma.h:60</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_a9f53a5e9b046b4f217e782b733941b0c"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a9f53a5e9b046b4f217e782b733941b0c">mlx::steel::BaseMMAFrag< T, 8, 8 >::frag_type</a></div><div class="ttdeci">metal::vec< T, kElemsPerFrag > frag_type</div><div class="ttdef"><b>Definition</b> mma.h:61</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_aa8f50ea8961ec5b35c1b81366d64f2cb"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#aa8f50ea8961ec5b35c1b81366d64f2cb">mlx::steel::BaseMMAFrag< T, 8, 8 >::store</a></div><div class="ttdeci">static METAL_FUNC constexpr void store(const thread frag_type &src, DstPtrType dst, StrX str_x, StrY str_y)</div><div class="ttdef"><b>Definition</b> mma.h:102</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_ac73006b36fc710feda3a7c796e21415c"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#ac73006b36fc710feda3a7c796e21415c">mlx::steel::BaseMMAFrag< T, 8, 8 >::load</a></div><div class="ttdeci">static METAL_FUNC constexpr void load(thread frag_type &dst, SrcPtrType src, StrX str_x, StrY str_y)</div><div class="ttdef"><b>Definition</b> mma.h:59</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4_html_ad22aaee4a2938cbdd315b39eda84e07d"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#ad22aaee4a2938cbdd315b39eda84e07d">mlx::steel::BaseMMAFrag< T, 8, 8 >::load_safe</a></div><div class="ttdeci">static METAL_FUNC constexpr void load_safe(thread frag_type &dst, SrcPtrType src, StrX str_x, StrY str_y, LimX lim_x, LimY lim_y, OffX off_x=Int< 0 >{}, OffY off_y=Int< 0 >{})</div><div class="ttdef"><b>Definition</b> mma.h:77</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html">mlx::steel::BaseMMAFrag</a></div><div class="ttdef"><b>Definition</b> mma.h:23</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html">mlx::steel::BlockMMA</a></div><div class="ttdef"><b>Definition</b> mma.h:377</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a1868f57d57c8adedab2c58492ec76946"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a1868f57d57c8adedab2c58492ec76946">mlx::steel::BaseMMAFrag< T, 8, 8 >::mma</a></div><div class="ttdeci">static METAL_FUNC constexpr void mma(thread mat_type &D, thread mat_type &A, thread mat_type &B, thread mat_type &C)</div><div class="ttdef"><b>Definition</b> mma.h:164</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a1f0b00daad8eba2f855bb306e70d2328"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a1f0b00daad8eba2f855bb306e70d2328">mlx::steel::BaseMMAFrag< T, 8, 8 >::store_safe</a></div><div class="ttdeci">static METAL_FUNC constexpr void store_safe(const thread frag_type &src, DstPtrType dst, StrX str_x, StrY str_y, LimX lim_x, LimY lim_y, OffX off_x=Int< 0 >{}, OffY off_y=Int< 0 >{})</div><div class="ttdef"><b>Definition</b> mma.h:122</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a211102315e2afbcfcd2e2c201b638e9f"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a211102315e2afbcfcd2e2c201b638e9f">mlx::steel::BaseMMAFrag< T, 8, 8 >::kFragCols</a></div><div class="ttdeci">STEEL_CONST int kFragCols</div><div class="ttdef"><b>Definition</b> mma.h:49</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a25675ae18947a97c6e04157b540103a9"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9">mlx::steel::BaseMMAFrag< T, 8, 8 >::mat_type</a></div><div class="ttdeci">metal::simdgroup_matrix< T, kFragRows, kFragCols > mat_type</div><div class="ttdef"><b>Definition</b> mma.h:60</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a2fe53db449c692226f23f6b99fb2c0d4"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a2fe53db449c692226f23f6b99fb2c0d4">mlx::steel::BaseMMAFrag< T, 8, 8 >::kFragRows</a></div><div class="ttdeci">STEEL_CONST int kFragRows</div><div class="ttdef"><b>Definition</b> mma.h:48</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a3c34dfdc944db110f4735f1b25307cf0"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a3c34dfdc944db110f4735f1b25307cf0">mlx::steel::BaseMMAFrag< T, 8, 8 >::kElemsPerFrag</a></div><div class="ttdeci">STEEL_CONST int kElemsPerFrag</div><div class="ttdef"><b>Definition</b> mma.h:51</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a7331fff1d12f2f8b72b0006a3ad0dd83"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7331fff1d12f2f8b72b0006a3ad0dd83">mlx::steel::BaseMMAFrag< T, 8, 8 >::get_coord</a></div><div class="ttdeci">static METAL_FUNC constexpr short2 get_coord(ushort simd_lane_id)</div><div class="ttdef"><b>Definition</b> mma.h:49</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a76aa5aa690dbcc954e957d767fad661f"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a76aa5aa690dbcc954e957d767fad661f">mlx::steel::BaseMMAFrag< T, 8, 8 >::kElemRows</a></div><div class="ttdeci">STEEL_CONST int kElemRows</div><div class="ttdef"><b>Definition</b> mma.h:53</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a7c212200d86b4e93f274d99addf668bd"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">mlx::steel::BaseMMAFrag< T, 8, 8 >::kElemCols</a></div><div class="ttdeci">STEEL_CONST int kElemCols</div><div class="ttdef"><b>Definition</b> mma.h:54</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a8028512f5a3d2b6acaf966be529627a3"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8028512f5a3d2b6acaf966be529627a3">mlx::steel::BaseMMAFrag< T, 8, 8 >::mma</a></div><div class="ttdeci">static METAL_FUNC constexpr void mma(thread frag_type &D, thread frag_type &A, thread frag_type &B, thread frag_type &C)</div><div class="ttdef"><b>Definition</b> mma.h:145</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a8536bfaa108031c2ea3e9ccdc766ee5b"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">mlx::steel::BaseMMAFrag< T, 8, 8 >::frag_type</a></div><div class="ttdeci">metal::vec< T, kElemsPerFrag > frag_type</div><div class="ttdef"><b>Definition</b> mma.h:61</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_aa8f50ea8961ec5b35c1b81366d64f2cb"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#aa8f50ea8961ec5b35c1b81366d64f2cb">mlx::steel::BaseMMAFrag< T, 8, 8 >::store</a></div><div class="ttdeci">static METAL_FUNC constexpr void store(const thread frag_type &src, DstPtrType dst, StrX str_x, StrY str_y)</div><div class="ttdef"><b>Definition</b> mma.h:102</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_ac73006b36fc710feda3a7c796e21415c"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#ac73006b36fc710feda3a7c796e21415c">mlx::steel::BaseMMAFrag< T, 8, 8 >::load</a></div><div class="ttdeci">static METAL_FUNC constexpr void load(thread frag_type &dst, SrcPtrType src, StrX str_x, StrY str_y)</div><div class="ttdef"><b>Definition</b> mma.h:59</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_ad22aaee4a2938cbdd315b39eda84e07d"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#ad22aaee4a2938cbdd315b39eda84e07d">mlx::steel::BaseMMAFrag< T, 8, 8 >::load_safe</a></div><div class="ttdeci">static METAL_FUNC constexpr void load_safe(thread frag_type &dst, SrcPtrType src, StrX str_x, StrY str_y, LimX lim_x, LimY lim_y, OffX off_x=Int< 0 >{}, OffY off_y=Int< 0 >{})</div><div class="ttdef"><b>Definition</b> mma.h:77</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a0461451ffb5041b6a916ea17ed34288b"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a0461451ffb5041b6a916ea17ed34288b">mlx::steel::BlockMMA::store_result</a></div><div class="ttdeci">METAL_FUNC void store_result(device U *D, const int ldd)</div><div class="ttdef"><b>Definition</b> mma.h:464</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a081ba538d30d1d02498a7f341e6bd611"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a081ba538d30d1d02498a7f341e6bd611">mlx::steel::BlockMMA::store_result_safe</a></div><div class="ttdeci">METAL_FUNC void store_result_safe(device U *D, const int ldd, short2 dst_tile_dims)</div><div class="ttdef"><b>Definition</b> mma.h:478</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a138ed1bbad2ca88d3a3c7d162cd36562"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a138ed1bbad2ca88d3a3c7d162cd36562">mlx::steel::BlockMMA::As_offset</a></div><div class="ttdeci">short As_offset</div><div class="ttdef"><b>Definition</b> mma.h:485</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a21b0c40d16eced109bd3196186170bc6"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a21b0c40d16eced109bd3196186170bc6">mlx::steel::BlockMMA::Ctile</a></div><div class="ttdeci">MMATile< AccumType, TM, TN, MMAFrag_acc_t > Ctile</div><div class="ttdef"><b>Definition</b> mma.h:479</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a257287702dc849d0d8a078fced453142"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a257287702dc849d0d8a078fced453142">mlx::steel::BlockMMA::A_str_k</a></div><div class="ttdeci">STEEL_CONST short A_str_k</div><div class="ttdef"><b>Definition</b> mma.h:466</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a44fca27c821764317263047a780977b0"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a44fca27c821764317263047a780977b0">mlx::steel::BlockMMA::Btile</a></div><div class="ttdeci">MMATile< AccumType, 1, TN, MMAFrag_acc_t > Btile</div><div class="ttdef"><b>Definition</b> mma.h:478</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a47e614120c650f7479db79f23a0df586"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a47e614120c650f7479db79f23a0df586">mlx::steel::BlockMMA::Atile</a></div><div class="ttdeci">MMATile< AccumType, TM, 1, MMAFrag_acc_t > Atile</div><div class="ttdef"><b>Definition</b> mma.h:477</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a49538190209e522ddbef45fe95563d17"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a49538190209e522ddbef45fe95563d17">mlx::steel::BlockMMA::B_str_n</a></div><div class="ttdeci">STEEL_CONST short B_str_n</div><div class="ttdef"><b>Definition</b> mma.h:470</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a5b0029866f493363942133b55bff7307"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a5b0029866f493363942133b55bff7307">mlx::steel::BlockMMA::TM_stride</a></div><div class="ttdeci">STEEL_CONST short TM_stride</div><div class="ttdef"><b>Definition</b> mma.h:455</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a138ed1bbad2ca88d3a3c7d162cd36562"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a138ed1bbad2ca88d3a3c7d162cd36562">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::As_offset</a></div><div class="ttdeci">short As_offset</div><div class="ttdef"><b>Definition</b> mma.h:485</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a21b0c40d16eced109bd3196186170bc6"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a21b0c40d16eced109bd3196186170bc6">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::Ctile</a></div><div class="ttdeci">MMATile< AccumType, TM, TN, MMAFrag_acc_t > Ctile</div><div class="ttdef"><b>Definition</b> mma.h:479</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a257287702dc849d0d8a078fced453142"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a257287702dc849d0d8a078fced453142">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::A_str_k</a></div><div class="ttdeci">STEEL_CONST short A_str_k</div><div class="ttdef"><b>Definition</b> mma.h:466</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a44fca27c821764317263047a780977b0"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a44fca27c821764317263047a780977b0">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::Btile</a></div><div class="ttdeci">MMATile< AccumType, 1, TN, MMAFrag_acc_t > Btile</div><div class="ttdef"><b>Definition</b> mma.h:478</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a47e614120c650f7479db79f23a0df586"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a47e614120c650f7479db79f23a0df586">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::Atile</a></div><div class="ttdeci">MMATile< AccumType, TM, 1, MMAFrag_acc_t > Atile</div><div class="ttdef"><b>Definition</b> mma.h:477</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a49538190209e522ddbef45fe95563d17"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a49538190209e522ddbef45fe95563d17">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::B_str_n</a></div><div class="ttdeci">STEEL_CONST short B_str_n</div><div class="ttdef"><b>Definition</b> mma.h:470</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a5b0029866f493363942133b55bff7307"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a5b0029866f493363942133b55bff7307">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::TM_stride</a></div><div class="ttdeci">STEEL_CONST short TM_stride</div><div class="ttdef"><b>Definition</b> mma.h:455</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a6a2c2a6d5e767d52c41b42a9d36086b0"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a6a2c2a6d5e767d52c41b42a9d36086b0">mlx::steel::BlockMMA::mma</a></div><div class="ttdeci">METAL_FUNC void mma(const threadgroup T *As, const threadgroup T *Bs)</div><div class="ttdef"><b>Definition</b> mma.h:437</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a706ae779c1f8d2eb18f19c248567d424"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a706ae779c1f8d2eb18f19c248567d424">mlx::steel::BlockMMA::TN</a></div><div class="ttdeci">STEEL_CONST short TN</div><div class="ttdef"><b>Definition</b> mma.h:462</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a706ae779c1f8d2eb18f19c248567d424"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a706ae779c1f8d2eb18f19c248567d424">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::TN</a></div><div class="ttdeci">STEEL_CONST short TN</div><div class="ttdef"><b>Definition</b> mma.h:462</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a7b324c992750ed3aaa4c485f15b2f391"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a7b324c992750ed3aaa4c485f15b2f391">mlx::steel::BlockMMA::store_result_safe</a></div><div class="ttdeci">METAL_FUNC void store_result_safe(device U *D, const int ldd, const device U *C, const int ldc, const int fdc, short2 dst_tile_dims, thread const Epilogue &epilogue_op) const</div><div class="ttdef"><b>Definition</b> mma.h:611</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a7cf757e9785e23997b1417e024559ed3"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a7cf757e9785e23997b1417e024559ed3">mlx::steel::BlockMMA::store_result</a></div><div class="ttdeci">METAL_FUNC void store_result(device U *D, const int ldd, const device U *C, const int ldc, const int fdc, thread const Epilogue &epilogue_op) const</div><div class="ttdef"><b>Definition</b> mma.h:579</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a823c56cbd2086f10272df7284a5247ae"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a823c56cbd2086f10272df7284a5247ae">mlx::steel::BlockMMA::apply_epilogue</a></div><div class="ttdeci">METAL_FUNC void apply_epilogue(const device U *C, const int ldc, const int fdc, thread const BinaryEpilogue &epilogue_op)</div><div class="ttdef"><b>Definition</b> mma.h:507</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a8b3690b383afd26563efb38f9c375e50"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a8b3690b383afd26563efb38f9c375e50">mlx::steel::BlockMMA::TN_stride</a></div><div class="ttdeci">STEEL_CONST short TN_stride</div><div class="ttdef"><b>Definition</b> mma.h:457</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a8fddaa78913cdc8eea5e1cf7d2776330"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a8fddaa78913cdc8eea5e1cf7d2776330">mlx::steel::BlockMMA::tile_stride_a</a></div><div class="ttdeci">STEEL_CONST short tile_stride_a</div><div class="ttdef"><b>Definition</b> mma.h:473</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a92f6aeee432f53638447eac842f43eca"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a92f6aeee432f53638447eac842f43eca">mlx::steel::BlockMMA::Bs_offset</a></div><div class="ttdeci">short Bs_offset</div><div class="ttdef"><b>Definition</b> mma.h:486</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a8b3690b383afd26563efb38f9c375e50"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a8b3690b383afd26563efb38f9c375e50">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::TN_stride</a></div><div class="ttdeci">STEEL_CONST short TN_stride</div><div class="ttdef"><b>Definition</b> mma.h:457</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a8fddaa78913cdc8eea5e1cf7d2776330"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a8fddaa78913cdc8eea5e1cf7d2776330">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::tile_stride_a</a></div><div class="ttdeci">STEEL_CONST short tile_stride_a</div><div class="ttdef"><b>Definition</b> mma.h:473</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a92f6aeee432f53638447eac842f43eca"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a92f6aeee432f53638447eac842f43eca">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::Bs_offset</a></div><div class="ttdeci">short Bs_offset</div><div class="ttdef"><b>Definition</b> mma.h:486</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_a9e48f2d51099ec00171506724faab54a"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#a9e48f2d51099ec00171506724faab54a">mlx::steel::BlockMMA::apply_epilogue_safe</a></div><div class="ttdeci">METAL_FUNC void apply_epilogue_safe(const device U *C, const int ldc, const int fdc, short2 dst_tile_dims, thread const BinaryEpilogue &epilogue_op)</div><div class="ttdef"><b>Definition</b> mma.h:535</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aa14406b7298456ac45d23dd3c4642dd8"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aa14406b7298456ac45d23dd3c4642dd8">mlx::steel::BlockMMA::BlockMMA</a></div><div class="ttdeci">METAL_FUNC BlockMMA(ushort simd_group_id, ushort simd_lane_id)</div><div class="ttdef"><b>Definition</b> mma.h:417</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aa71400922babd388177f228c2c82b211"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aa71400922babd388177f228c2c82b211">mlx::steel::BlockMMA::B_str_k</a></div><div class="ttdeci">STEEL_CONST short B_str_k</div><div class="ttdef"><b>Definition</b> mma.h:469</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aa85451edf6900fd6af164d4d50889ae3"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aa85451edf6900fd6af164d4d50889ae3">mlx::steel::BlockMMA::sm</a></div><div class="ttdeci">short sm</div><div class="ttdef"><b>Definition</b> mma.h:482</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_ab9c7f5386594497f5f4df7e59670b877"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#ab9c7f5386594497f5f4df7e59670b877">mlx::steel::BlockMMA::A_str_m</a></div><div class="ttdeci">STEEL_CONST short A_str_m</div><div class="ttdef"><b>Definition</b> mma.h:465</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aba5f749fdf32d8bd9d9e29f2a9ae4591"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aba5f749fdf32d8bd9d9e29f2a9ae4591">mlx::steel::BlockMMA::TM</a></div><div class="ttdeci">STEEL_CONST short TM</div><div class="ttdef"><b>Definition</b> mma.h:460</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_ade420e8b811d597345783c324c23a34a"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#ade420e8b811d597345783c324c23a34a">mlx::steel::BlockMMA::sn</a></div><div class="ttdeci">short sn</div><div class="ttdef"><b>Definition</b> mma.h:483</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_ae3f35453b3afbaac9df64ad5966b34a4"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#ae3f35453b3afbaac9df64ad5966b34a4">mlx::steel::BlockMMA::tile_stride_b</a></div><div class="ttdeci">STEEL_CONST short tile_stride_b</div><div class="ttdef"><b>Definition</b> mma.h:474</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aa14406b7298456ac45d23dd3c4642dd8"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aa14406b7298456ac45d23dd3c4642dd8">mlx::steel::BlockMMA::BlockMMA</a></div><div class="ttdeci">METAL_FUNC BlockMMA(ushort simd_group_id, ushort simd_lane_id)</div><div class="ttdef"><b>Definition</b> mma.h:489</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aa71400922babd388177f228c2c82b211"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aa71400922babd388177f228c2c82b211">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::B_str_k</a></div><div class="ttdeci">STEEL_CONST short B_str_k</div><div class="ttdef"><b>Definition</b> mma.h:469</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aa85451edf6900fd6af164d4d50889ae3"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aa85451edf6900fd6af164d4d50889ae3">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::sm</a></div><div class="ttdeci">short sm</div><div class="ttdef"><b>Definition</b> mma.h:482</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_ab9c7f5386594497f5f4df7e59670b877"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#ab9c7f5386594497f5f4df7e59670b877">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::A_str_m</a></div><div class="ttdeci">STEEL_CONST short A_str_m</div><div class="ttdef"><b>Definition</b> mma.h:465</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aba5f749fdf32d8bd9d9e29f2a9ae4591"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aba5f749fdf32d8bd9d9e29f2a9ae4591">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::TM</a></div><div class="ttdeci">STEEL_CONST short TM</div><div class="ttdef"><b>Definition</b> mma.h:460</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_ade420e8b811d597345783c324c23a34a"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#ade420e8b811d597345783c324c23a34a">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::sn</a></div><div class="ttdeci">short sn</div><div class="ttdef"><b>Definition</b> mma.h:483</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_ae2c42cb6d0dde785859164c195f4d13c"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#ae2c42cb6d0dde785859164c195f4d13c">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::MMAFrag_acc_t</a></div><div class="ttdeci">BaseMMAFrag< AccumType, kFragSize, kFragSize > MMAFrag_acc_t</div><div class="ttdef"><b>Definition</b> mma.h:452</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_ae3f35453b3afbaac9df64ad5966b34a4"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#ae3f35453b3afbaac9df64ad5966b34a4">mlx::steel::BlockMMA< T, U, BM, BN, BK, WM, WN, transpose_a, transpose_b, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, AccumType, Epilogue >::tile_stride_b</a></div><div class="ttdeci">STEEL_CONST short tile_stride_b</div><div class="ttdef"><b>Definition</b> mma.h:474</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_aee8caec45c1f9e4428586effbfe6137d"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#aee8caec45c1f9e4428586effbfe6137d">mlx::steel::BlockMMA::kFragSize</a></div><div class="ttdeci">STEEL_CONST short kFragSize</div><div class="ttdef"><b>Definition</b> mma.h:451</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_m_m_a_html_af653c0808ba4fa9a25286f1febb7baff"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_m_m_a.html#af653c0808ba4fa9a25286f1febb7baff">mlx::steel::BlockMMA::apply_epilogue</a></div><div class="ttdeci">METAL_FUNC void apply_epilogue(thread const UnaryEpilogue &epilogue_op)</div><div class="ttdef"><b>Definition</b> mma.h:497</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html">mlx::steel::MMATile</a></div><div class="ttdef"><b>Definition</b> mma.h:178</div></div>
|
||||
@@ -865,29 +886,32 @@ $(function(){ initResizable(false); });
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a594142f957ffb99296a243f7af7b59e7"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7">mlx::steel::MMATile::kFragRows</a></div><div class="ttdeci">STEEL_CONST int kFragRows</div><div class="ttdef"><b>Definition</b> mma.h:226</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a60ea6b8ff2923b7fe6f598e74ac54323"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a60ea6b8ff2923b7fe6f598e74ac54323">mlx::steel::MMATile::kRows</a></div><div class="ttdeci">STEEL_CONST int kRows</div><div class="ttdef"><b>Definition</b> mma.h:233</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a684e6c6d9f00f583994285b60aaa3b62"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a684e6c6d9f00f583994285b60aaa3b62">mlx::steel::MMATile::val_frags</a></div><div class="ttdeci">frag_type val_frags[kNumFrags]</div><div class="ttdef"><b>Definition</b> mma.h:245</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a6dadcd666afb3759a11094e754560dd4"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a6dadcd666afb3759a11094e754560dd4">mlx::steel::MMATile::MMAFrag_t</a></div><div class="ttdeci">MMAFrag_ MMAFrag_t</div><div class="ttdef"><b>Definition</b> mma.h:224</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a752f708e4fe5ef37fdd902dae153179f"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a752f708e4fe5ef37fdd902dae153179f">mlx::steel::MMATile::store</a></div><div class="ttdeci">METAL_FUNC void store(device U *dst, const int ld) const</div><div class="ttdef"><b>Definition</b> mma.h:285</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a824409bc107330805853f932e80a7628"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a824409bc107330805853f932e80a7628">mlx::steel::MMATile::elem_type</a></div><div class="ttdeci">T elem_type</div><div class="ttdef"><b>Definition</b> mma.h:225</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a80078f0dfa4c225e79d9b460202d5e2c"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a80078f0dfa4c225e79d9b460202d5e2c">mlx::steel::MMATile::elem_type</a></div><div class="ttdeci">T elem_type</div><div class="ttdef"><b>Definition</b> mma.h:225</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a865ece5ad0b9a56937b6d77a18b5a1dc"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a865ece5ad0b9a56937b6d77a18b5a1dc">mlx::steel::MMATile::elems</a></div><div class="ttdeci">METAL_FUNC thread elem_type * elems()</div><div class="ttdef"><b>Definition</b> mma.h:227</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a948784652e93830887ee8ad506ec3257"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a948784652e93830887ee8ad506ec3257">mlx::steel::MMATile::kCols</a></div><div class="ttdeci">STEEL_CONST int kCols</div><div class="ttdef"><b>Definition</b> mma.h:234</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a98357339ec98f804a1b12597937b318f"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a98357339ec98f804a1b12597937b318f">mlx::steel::MMATile::kElemsPerTile</a></div><div class="ttdeci">STEEL_CONST int kElemsPerTile</div><div class="ttdef"><b>Definition</b> mma.h:237</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_aa3a4af67813908109da08ce7352f82da"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#aa3a4af67813908109da08ce7352f82da">mlx::steel::MMATile::load_safe</a></div><div class="ttdeci">METAL_FUNC void load_safe(const device U *src, const int ld, const short2 src_tile_dims)</div><div class="ttdef"><b>Definition</b> mma.h:301</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_aa3fb310dd08ec23c334511f7b316d1b6"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#aa3fb310dd08ec23c334511f7b316d1b6">mlx::steel::MMATile::MMATile</a></div><div class="ttdeci">METAL_FUNC MMATile() thread</div><div class="ttdef"><b>Definition</b> mma.h:199</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_aa3fb310dd08ec23c334511f7b316d1b6"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#aa3fb310dd08ec23c334511f7b316d1b6">mlx::steel::MMATile::MMATile</a></div><div class="ttdeci">METAL_FUNC MMATile() thread</div><div class="ttdef"><b>Definition</b> mma.h:247</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_aa5426c6beabfb3ee41b58f01b3392a96"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#aa5426c6beabfb3ee41b58f01b3392a96">mlx::steel::MMATile::load</a></div><div class="ttdeci">METAL_FUNC void load(const threadgroup U *src)</div><div class="ttdef"><b>Definition</b> mma.h:236</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_aa97a98e423827a889c13a92217626ec7"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#aa97a98e423827a889c13a92217626ec7">mlx::steel::MMATile::clear</a></div><div class="ttdeci">METAL_FUNC constexpr void clear()</div><div class="ttdef"><b>Definition</b> mma.h:201</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_aa9e484d8cae936503898d5b772c573f9"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#aa9e484d8cae936503898d5b772c573f9">mlx::steel::MMATile::load</a></div><div class="ttdeci">METAL_FUNC void load(const device U *src, const int ld)</div><div class="ttdef"><b>Definition</b> mma.h:270</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_aac25cd0a9bdf24aa2af809c95f0bd171"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#aac25cd0a9bdf24aa2af809c95f0bd171">mlx::steel::MMATile::frag_type</a></div><div class="ttdeci">MMAFrag_t::frag_type frag_type</div><div class="ttdef"><b>Definition</b> mma.h:243</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_abe33de70e34300745bad9aa822fd0382"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#abe33de70e34300745bad9aa822fd0382">mlx::steel::MMATile::MMAFrag_t</a></div><div class="ttdeci">MMAFrag_ MMAFrag_t</div><div class="ttdef"><b>Definition</b> mma.h:224</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_ad095371db98e7c335ec41ca77c10f906"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#ad095371db98e7c335ec41ca77c10f906">mlx::steel::MMATile::kFragCols</a></div><div class="ttdeci">STEEL_CONST int kFragCols</div><div class="ttdef"><b>Definition</b> mma.h:227</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_ad476e1d9a12178fb35c207312339e485"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#ad476e1d9a12178fb35c207312339e485">mlx::steel::MMATile::frag_at</a></div><div class="ttdeci">METAL_FUNC constexpr const thread frag_type & frag_at(const short i, const short j) const</div><div class="ttdef"><b>Definition</b> mma.h:212</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_ae21bb7cce701290de84c6015e064d8a1"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae21bb7cce701290de84c6015e064d8a1">mlx::steel::MMATile::elems</a></div><div class="ttdeci">METAL_FUNC const thread elem_type * elems() const</div><div class="ttdef"><b>Definition</b> mma.h:231</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_ae326e7693eb77c22d5a6e3e9219019d3"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#ae326e7693eb77c22d5a6e3e9219019d3">mlx::steel::MMATile::kNumFrags</a></div><div class="ttdeci">STEEL_CONST int kNumFrags</div><div class="ttdef"><b>Definition</b> mma.h:236</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_aef0ea2387e1ff5767bff8563b2d36bd6"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#aef0ea2387e1ff5767bff8563b2d36bd6">mlx::steel::MMATile::kElemsPerFrag</a></div><div class="ttdeci">STEEL_CONST int kElemsPerFrag</div><div class="ttdef"><b>Definition</b> mma.h:228</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1integral__constant_html"><div class="ttname"><a href="structmlx_1_1steel_1_1integral__constant.html">mlx::steel::integral_constant</a></div><div class="ttdef"><b>Definition</b> integral_constant.h:18</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_transform_none_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_transform_none.html">mlx::steel::TransformNone</a></div><div class="ttdef"><b>Definition</b> transforms.h:15</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="dir_76215a6c54e2b67053e723fc2395583c.html">steel</a></li><li class="navelem"><a class="el" href="dir_6768c99e6145fb9510ccdb40db8ede25.html">gemm</a></li><li class="navelem"><a class="el" href="gemm_2mma_8h.html">mma.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