mirror of
https://github.com/ml-explore/mlx.git
synced 2025-11-05 19:48:15 +08:00
rebase
This commit is contained in:
126
docs/build/html/attn_2loader_8h_source.html
vendored
126
docs/build/html/attn_2loader_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/attn/loader.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('attn_2loader_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_e1756c7634b0c14aead026895ad71c6d.html">attn</a></li> </ul>
|
||||
</div>
|
||||
</div><!-- top -->
|
||||
<div id="doc-content">
|
||||
<div class="header">
|
||||
<div class="headertitle"><div class="title">loader.h</div></div>
|
||||
</div><!--header-->
|
||||
@@ -96,11 +110,11 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00003" name="l00003"></a><span class="lineno"> 3</span><span class="preprocessor">#pragma once</span></div>
|
||||
<div class="line"><a id="l00004" name="l00004"></a><span class="lineno"> 4</span> </div>
|
||||
<div class="line"><a id="l00005" name="l00005"></a><span class="lineno"> 5</span><span class="preprocessor">#include "<a class="code" href="steel_2defines_8h.html">mlx/backend/metal/kernels/steel/defines.h</a>"</span></div>
|
||||
<div class="line"><a id="l00006" name="l00006"></a><span class="lineno"> 6</span> </div>
|
||||
<div class="line"><a id="l00006" name="l00006"></a><span class="lineno"> 6</span></div>
|
||||
<div class="line"><a id="l00008" name="l00008"></a><span class="lineno"> 8</span><span class="comment">// Loading helper</span></div>
|
||||
<div class="line"><a id="l00010" name="l00010"></a><span class="lineno"> 10</span> </div>
|
||||
<div class="line"><a id="l00011" name="l00011"></a><span class="lineno"> 11</span><span class="keyword">namespace </span><a class="code hl_namespace" href="namespacemlx.html">mlx</a> {</div>
|
||||
<div class="line"><a id="l00012" name="l00012"></a><span class="lineno"> 12</span><span class="keyword">namespace </span>steel {</div>
|
||||
<div class="line"><a id="l00012" name="l00012"></a><span class="lineno"> 12</span><span class="keyword">namespace </span><a class="code hl_namespace" href="namespacemlx_1_1steel.html">steel</a> {</div>
|
||||
<div class="line"><a id="l00013" name="l00013"></a><span class="lineno"> 13</span> </div>
|
||||
<div class="line"><a id="l00014" name="l00014"></a><span class="lineno"> 14</span><span class="keyword">template</span> <</div>
|
||||
<div class="line"><a id="l00015" name="l00015"></a><span class="lineno"> 15</span> <span class="keyword">typename</span> T,</div>
|
||||
@@ -113,7 +127,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00022" name="l00022"></a><span class="lineno"> 22</span> <span class="keywordtype">short</span> n_reads = (BCOLS * BROWS) / (tgp_size),</div>
|
||||
<div class="line"><a id="l00023" name="l00023"></a><span class="lineno"> 23</span> <span class="keywordtype">short</span> TCOLS = BCOLS / n_reads,</div>
|
||||
<div class="line"><a id="l00024" name="l00024"></a><span class="lineno"> 24</span> <span class="keywordtype">short</span> TROWS = tgp_size / TCOLS></div>
|
||||
<div class="line"><a id="l00025" name="l00025"></a><span class="lineno"> 25</span><span class="keyword">struct </span>BlockLoader {</div>
|
||||
<div class="line"><a id="l00025" name="l00025"></a><span class="lineno"> 25</span><span class="keyword">struct </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">BlockLoader</a> {</div>
|
||||
<div class="line"><a id="l00026" name="l00026"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader.html#a973804e5b1d418c98c90861cda1a6fb5"> 26</a></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_loader.html#a973804e5b1d418c98c90861cda1a6fb5">n_rows</a> = (BROWS + TROWS - 1) / TROWS;</div>
|
||||
<div class="line"><a id="l00027" name="l00027"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader.html#a58bdf9b9c81962733e22ecdeae28c092"> 27</a></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_loader.html#a58bdf9b9c81962733e22ecdeae28c092">vec_size</a> = n_reads;</div>
|
||||
<div class="line"><a id="l00028" name="l00028"></a><span class="lineno"> 28</span> </div>
|
||||
@@ -130,7 +144,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00039" name="l00039"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2"> 39</a></span> threadgroup T* <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">dst</a>;</div>
|
||||
<div class="line"><a id="l00040" name="l00040"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader.html#ad1db14517568ae9eddfb6986ef31c7aa"> 40</a></span> <span class="keyword">const</span> device T* <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#ad1db14517568ae9eddfb6986ef31c7aa">src</a>;</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">struct </span><span class="keyword">alignas</span>(alignment * sizeof(T)) <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader_1_1_read_vector.html">ReadVector</a> {</div>
|
||||
<div class="line"><a id="l00042" name="l00042"></a><span class="lineno"> 42</span> <span class="keyword">struct </span><span class="keyword">alignas</span>(alignment * sizeof(T)) ReadVector {</div>
|
||||
<div class="line"><a id="l00043" name="l00043"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader_1_1_read_vector.html#a20963f7191251defca48bf8a843d019d"> 43</a></span> uint8_t <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_1_1_read_vector.html#a20963f7191251defca48bf8a843d019d">v</a>[<span class="keyword">sizeof</span>(T) * <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a58bdf9b9c81962733e22ecdeae28c092">vec_size</a>];</div>
|
||||
<div class="line"><a id="l00044" name="l00044"></a><span class="lineno"> 44</span> };</div>
|
||||
<div class="line"><a id="l00045" name="l00045"></a><span class="lineno"> 45</span> </div>
|
||||
@@ -147,19 +161,19 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00055" name="l00055"></a><span class="lineno"> 55</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a064e2cc77e0b1cf0f8027929e031775b">thread_idx</a>(simd_group_id * 32 + simd_lane_id),</div>
|
||||
<div class="line"><a id="l00056" name="l00056"></a><span class="lineno"> 56</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a9ef13742bcdf07532d8f09394928a8af">bi</a>(<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a064e2cc77e0b1cf0f8027929e031775b">thread_idx</a> / TCOLS),</div>
|
||||
<div class="line"><a id="l00057" name="l00057"></a><span class="lineno"> 57</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a78c326e75ee35a484685771143047cd4">bj</a>(<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a58bdf9b9c81962733e22ecdeae28c092">vec_size</a> * (<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a064e2cc77e0b1cf0f8027929e031775b">thread_idx</a> % TCOLS)),</div>
|
||||
<div class="line"><a id="l00058" name="l00058"></a><span class="lineno"> 58</span> dst(dst_ + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a9ef13742bcdf07532d8f09394928a8af">bi</a> * dst_ld + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a78c326e75ee35a484685771143047cd4">bj</a>),</div>
|
||||
<div class="line"><a id="l00058" name="l00058"></a><span class="lineno"> 58</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">dst</a>(dst_ + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a9ef13742bcdf07532d8f09394928a8af">bi</a> * dst_ld + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a78c326e75ee35a484685771143047cd4">bj</a>),</div>
|
||||
<div class="line"><a id="l00059" name="l00059"></a><span class="lineno"> 59</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#ad1db14517568ae9eddfb6986ef31c7aa">src</a>(src_ + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a9ef13742bcdf07532d8f09394928a8af">bi</a> * <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#aadafc50f7f06af434149d7469df4714d">src_ld</a> + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#a78c326e75ee35a484685771143047cd4">bj</a>) {}</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00060" name="l00060"></a><span class="lineno"> 60</span> </div>
|
||||
<div class="line"><a id="l00061" name="l00061"></a><span class="lineno"> 61</span> <span class="comment">/* Apply operation to threadgroup without bound checking */</span></div>
|
||||
<div class="line"><a id="l00062" name="l00062"></a><span class="lineno"> 62</span> <span class="keyword">template</span> <<span class="keyword">typename</span> UnaryOp></div>
|
||||
<div class="foldopen" id="foldopen00063" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00063" name="l00063"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader.html#adb4ca2cc193630a779de552fa8847ddf"> 63</a></span> METAL_FUNC <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_block_loader.html#adb4ca2cc193630a779de552fa8847ddf">apply_inplace_op</a>(thread <span class="keyword">const</span> UnaryOp& <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>)<span class="keyword"> const </span>{</div>
|
||||
<div class="line"><a id="l00063" name="l00063"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader.html#adb4ca2cc193630a779de552fa8847ddf"> 63</a></span> METAL_FUNC <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_block_loader.html#adb4ca2cc193630a779de552fa8847ddf">apply_inplace_op</a>(thread <span class="keyword">const</span> UnaryOp& op)<span class="keyword"> const </span>{</div>
|
||||
<div class="line"><a id="l00064" name="l00064"></a><span class="lineno"> 64</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00065" name="l00065"></a><span class="lineno"> 65</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < BROWS; i += TROWS) {</div>
|
||||
<div class="line"><a id="l00066" name="l00066"></a><span class="lineno"> 66</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00067" name="l00067"></a><span class="lineno"> 67</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_block_loader.html#a58bdf9b9c81962733e22ecdeae28c092">vec_size</a>; j++) {</div>
|
||||
<div class="line"><a id="l00068" name="l00068"></a><span class="lineno"> 68</span> dst[i * dst_ld + j] = <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>.apply(dst[i * dst_ld + j]);</div>
|
||||
<div class="line"><a id="l00068" name="l00068"></a><span class="lineno"> 68</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">dst</a>[i * dst_ld + j] = op.apply(<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">dst</a>[i * dst_ld + j]);</div>
|
||||
<div class="line"><a id="l00069" name="l00069"></a><span class="lineno"> 69</span> }</div>
|
||||
<div class="line"><a id="l00070" name="l00070"></a><span class="lineno"> 70</span> }</div>
|
||||
<div class="line"><a id="l00071" name="l00071"></a><span class="lineno"> 71</span> }</div>
|
||||
@@ -170,8 +184,8 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00074" name="l00074"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader.html#a6c9e27f11f48b34580ed2c7e9cad9a27"> 74</a></span> METAL_FUNC <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_block_loader.html#a6c9e27f11f48b34580ed2c7e9cad9a27">load_unsafe</a>()<span class="keyword"> const </span>{</div>
|
||||
<div class="line"><a id="l00075" name="l00075"></a><span class="lineno"> 75</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00076" name="l00076"></a><span class="lineno"> 76</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < BROWS; i += TROWS) {</div>
|
||||
<div class="line"><a id="l00077" name="l00077"></a><span class="lineno"> 77</span> *((threadgroup <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader_1_1_read_vector.html">ReadVector</a>*)(&dst[i * dst_ld])) =</div>
|
||||
<div class="line"><a id="l00078" name="l00078"></a><span class="lineno"> 78</span> *((<span class="keyword">const</span> device <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader_1_1_read_vector.html">ReadVector</a>*)(&<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#ad1db14517568ae9eddfb6986ef31c7aa">src</a>[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#aadafc50f7f06af434149d7469df4714d">src_ld</a>]));</div>
|
||||
<div class="line"><a id="l00077" name="l00077"></a><span class="lineno"> 77</span> *((threadgroup ReadVector*)(&<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">dst</a>[i * dst_ld])) =</div>
|
||||
<div class="line"><a id="l00078" name="l00078"></a><span class="lineno"> 78</span> *((<span class="keyword">const</span> device ReadVector*)(&<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#ad1db14517568ae9eddfb6986ef31c7aa">src</a>[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#aadafc50f7f06af434149d7469df4714d">src_ld</a>]));</div>
|
||||
<div class="line"><a id="l00079" name="l00079"></a><span class="lineno"> 79</span> }</div>
|
||||
<div class="line"><a id="l00080" name="l00080"></a><span class="lineno"> 80</span> }</div>
|
||||
</div>
|
||||
@@ -187,7 +201,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00089" name="l00089"></a><span class="lineno"> 89</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < BROWS; i += TROWS) {</div>
|
||||
<div class="line"><a id="l00090" name="l00090"></a><span class="lineno"> 90</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00091" name="l00091"></a><span class="lineno"> 91</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_block_loader.html#a58bdf9b9c81962733e22ecdeae28c092">vec_size</a>; j++) {</div>
|
||||
<div class="line"><a id="l00092" name="l00092"></a><span class="lineno"> 92</span> dst[i * dst_ld + j] = T(0);</div>
|
||||
<div class="line"><a id="l00092" name="l00092"></a><span class="lineno"> 92</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">dst</a>[i * dst_ld + j] = T(0);</div>
|
||||
<div class="line"><a id="l00093" name="l00093"></a><span class="lineno"> 93</span> }</div>
|
||||
<div class="line"><a id="l00094" name="l00094"></a><span class="lineno"> 94</span> }</div>
|
||||
<div class="line"><a id="l00095" name="l00095"></a><span class="lineno"> 95</span> <span class="keywordflow">return</span>;</div>
|
||||
@@ -220,7 +234,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00122" name="l00122"></a><span class="lineno"> 122</span> <span class="comment">// Copy values to threadgroup memory</span></div>
|
||||
<div class="line"><a id="l00123" name="l00123"></a><span class="lineno"> 123</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00124" name="l00124"></a><span class="lineno"> 124</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_block_loader.html#a58bdf9b9c81962733e22ecdeae28c092">vec_size</a>; j++) {</div>
|
||||
<div class="line"><a id="l00125" name="l00125"></a><span class="lineno"> 125</span> dst[i * dst_ld + j] = tmp_val[j];</div>
|
||||
<div class="line"><a id="l00125" name="l00125"></a><span class="lineno"> 125</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">dst</a>[i * dst_ld + j] = tmp_val[j];</div>
|
||||
<div class="line"><a id="l00126" name="l00126"></a><span class="lineno"> 126</span> }</div>
|
||||
<div class="line"><a id="l00127" name="l00127"></a><span class="lineno"> 127</span> }</div>
|
||||
<div class="line"><a id="l00128" name="l00128"></a><span class="lineno"> 128</span> }</div>
|
||||
@@ -254,7 +268,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00151" name="l00151"></a><span class="lineno"> 151</span> <span class="keywordtype">short</span> TCOLS = BCOLS / n_reads,</div>
|
||||
<div class="line"><a id="l00152" name="l00152"></a><span class="lineno"> 152</span> <span class="keywordtype">short</span> TROWS = tgp_size / TCOLS></div>
|
||||
<div class="foldopen" id="foldopen00153" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00153" name="l00153"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader_t.html"> 153</a></span><span class="keyword">struct </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader_t.html">BlockLoaderT</a> {</div>
|
||||
<div class="line"><a id="l00153" name="l00153"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader_t.html"> 153</a></span><span class="keyword">struct </span><a class="code hl_function" href="structmlx_1_1steel_1_1_block_loader_t.html#a076616a7c67ad1b847e0e6b046077ee2">BlockLoaderT</a> {</div>
|
||||
<div class="line"><a id="l00154" name="l00154"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader_t.html#a0ccc7caa93e6e709981a1a08159d41dc"> 154</a></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_loader_t.html#a0ccc7caa93e6e709981a1a08159d41dc">n_rows</a> = (BROWS + TROWS - 1) / TROWS;</div>
|
||||
<div class="line"><a id="l00155" name="l00155"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader_t.html#a9ac651d9e5097507c57b10dfeb40bfe5"> 155</a></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_loader_t.html#a9ac651d9e5097507c57b10dfeb40bfe5">vec_size</a> = n_reads;</div>
|
||||
<div class="line"><a id="l00156" name="l00156"></a><span class="lineno"> 156</span> </div>
|
||||
@@ -284,20 +298,20 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00179" name="l00179"></a><span class="lineno"> 179</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#af2838998a02866f22b525f9b6ae004da">thread_idx</a>(simd_group_id * 32 + simd_lane_id),</div>
|
||||
<div class="line"><a id="l00180" name="l00180"></a><span class="lineno"> 180</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6964273994b06d6cf8ef7e59fb10bb35">bi</a>(<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#af2838998a02866f22b525f9b6ae004da">thread_idx</a> / TCOLS),</div>
|
||||
<div class="line"><a id="l00181" name="l00181"></a><span class="lineno"> 181</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#aca83e49c31095badc8a46eb3c8e00957">bj</a>(<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a9ac651d9e5097507c57b10dfeb40bfe5">vec_size</a> * (<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#af2838998a02866f22b525f9b6ae004da">thread_idx</a> % TCOLS)),</div>
|
||||
<div class="line"><a id="l00182" name="l00182"></a><span class="lineno"> 182</span> dst(dst_ + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6964273994b06d6cf8ef7e59fb10bb35">bi</a> * kDstStrRow + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#aca83e49c31095badc8a46eb3c8e00957">bj</a> * kDstStrCol),</div>
|
||||
<div class="line"><a id="l00182" name="l00182"></a><span class="lineno"> 182</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6eb4e566b687395e27f290da288362db">dst</a>(dst_ + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6964273994b06d6cf8ef7e59fb10bb35">bi</a> * kDstStrRow + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#aca83e49c31095badc8a46eb3c8e00957">bj</a> * kDstStrCol),</div>
|
||||
<div class="line"><a id="l00183" name="l00183"></a><span class="lineno"> 183</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a7004a4efaa483cc79b8b79810a17c777">src</a>(src_ + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6964273994b06d6cf8ef7e59fb10bb35">bi</a> * <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#aeba87e81185da6b20a092c5d240d3321">src_ld</a> + <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#aca83e49c31095badc8a46eb3c8e00957">bj</a>) {}</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00184" name="l00184"></a><span class="lineno"> 184</span> </div>
|
||||
<div class="line"><a id="l00185" name="l00185"></a><span class="lineno"> 185</span> <span class="comment">/* Apply operation to threadgroup without bound checking */</span></div>
|
||||
<div class="line"><a id="l00186" name="l00186"></a><span class="lineno"> 186</span> <span class="keyword">template</span> <<span class="keyword">typename</span> UnaryOp></div>
|
||||
<div class="foldopen" id="foldopen00187" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00187" name="l00187"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader_t.html#a2b136fad00dc54300e68aa6b905eff97"> 187</a></span> METAL_FUNC <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_block_loader_t.html#a2b136fad00dc54300e68aa6b905eff97">apply_inplace_op</a>(thread <span class="keyword">const</span> UnaryOp& <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>)<span class="keyword"> const </span>{</div>
|
||||
<div class="line"><a id="l00187" name="l00187"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_loader_t.html#a2b136fad00dc54300e68aa6b905eff97"> 187</a></span> METAL_FUNC <span class="keywordtype">void</span> <a class="code hl_function" href="structmlx_1_1steel_1_1_block_loader_t.html#a2b136fad00dc54300e68aa6b905eff97">apply_inplace_op</a>(thread <span class="keyword">const</span> UnaryOp& op)<span class="keyword"> const </span>{</div>
|
||||
<div class="line"><a id="l00188" name="l00188"></a><span class="lineno"> 188</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00189" name="l00189"></a><span class="lineno"> 189</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < BROWS; i += TROWS) {</div>
|
||||
<div class="line"><a id="l00190" name="l00190"></a><span class="lineno"> 190</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00191" name="l00191"></a><span class="lineno"> 191</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_block_loader_t.html#a9ac651d9e5097507c57b10dfeb40bfe5">vec_size</a>; j++) {</div>
|
||||
<div class="line"><a id="l00192" name="l00192"></a><span class="lineno"> 192</span> dst[i * kDstStrRow + j * kDstStrCol] =</div>
|
||||
<div class="line"><a id="l00193" name="l00193"></a><span class="lineno"> 193</span> <a class="code hl_variable" href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a>.apply(dst[i * kDstStrRow + j * kDstStrCol]);</div>
|
||||
<div class="line"><a id="l00192" name="l00192"></a><span class="lineno"> 192</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6eb4e566b687395e27f290da288362db">dst</a>[i * kDstStrRow + j * kDstStrCol] =</div>
|
||||
<div class="line"><a id="l00193" name="l00193"></a><span class="lineno"> 193</span> op.apply(<a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6eb4e566b687395e27f290da288362db">dst</a>[i * kDstStrRow + j * kDstStrCol]);</div>
|
||||
<div class="line"><a id="l00194" name="l00194"></a><span class="lineno"> 194</span> }</div>
|
||||
<div class="line"><a id="l00195" name="l00195"></a><span class="lineno"> 195</span> }</div>
|
||||
<div class="line"><a id="l00196" name="l00196"></a><span class="lineno"> 196</span> }</div>
|
||||
@@ -310,7 +324,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00201" name="l00201"></a><span class="lineno"> 201</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < BROWS; i += TROWS) {</div>
|
||||
<div class="line"><a id="l00202" name="l00202"></a><span class="lineno"> 202</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00203" name="l00203"></a><span class="lineno"> 203</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_block_loader_t.html#a9ac651d9e5097507c57b10dfeb40bfe5">vec_size</a>; j++) {</div>
|
||||
<div class="line"><a id="l00204" name="l00204"></a><span class="lineno"> 204</span> dst[i * kDstStrRow + j * kDstStrCol] = <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a7004a4efaa483cc79b8b79810a17c777">src</a>[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#aeba87e81185da6b20a092c5d240d3321">src_ld</a> + j];</div>
|
||||
<div class="line"><a id="l00204" name="l00204"></a><span class="lineno"> 204</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6eb4e566b687395e27f290da288362db">dst</a>[i * kDstStrRow + j * kDstStrCol] = <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a7004a4efaa483cc79b8b79810a17c777">src</a>[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#aeba87e81185da6b20a092c5d240d3321">src_ld</a> + j];</div>
|
||||
<div class="line"><a id="l00205" name="l00205"></a><span class="lineno"> 205</span> }</div>
|
||||
<div class="line"><a id="l00206" name="l00206"></a><span class="lineno"> 206</span> }</div>
|
||||
<div class="line"><a id="l00207" name="l00207"></a><span class="lineno"> 207</span> }</div>
|
||||
@@ -327,7 +341,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00216" name="l00216"></a><span class="lineno"> 216</span> <span class="keywordflow">for</span> (<span class="keywordtype">short</span> i = 0; i < BROWS; i += TROWS) {</div>
|
||||
<div class="line"><a id="l00217" name="l00217"></a><span class="lineno"> 217</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00218" name="l00218"></a><span class="lineno"> 218</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_block_loader_t.html#a9ac651d9e5097507c57b10dfeb40bfe5">vec_size</a>; j++) {</div>
|
||||
<div class="line"><a id="l00219" name="l00219"></a><span class="lineno"> 219</span> dst[i * kDstStrRow + j * kDstStrCol] = T(0);</div>
|
||||
<div class="line"><a id="l00219" name="l00219"></a><span class="lineno"> 219</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6eb4e566b687395e27f290da288362db">dst</a>[i * kDstStrRow + j * kDstStrCol] = T(0);</div>
|
||||
<div class="line"><a id="l00220" name="l00220"></a><span class="lineno"> 220</span> }</div>
|
||||
<div class="line"><a id="l00221" name="l00221"></a><span class="lineno"> 221</span> }</div>
|
||||
<div class="line"><a id="l00222" name="l00222"></a><span class="lineno"> 222</span> <span class="keywordflow">return</span>;</div>
|
||||
@@ -360,7 +374,7 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00249" name="l00249"></a><span class="lineno"> 249</span> <span class="comment">// Copy values to threadgroup memory</span></div>
|
||||
<div class="line"><a id="l00250" name="l00250"></a><span class="lineno"> 250</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00251" name="l00251"></a><span class="lineno"> 251</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_block_loader_t.html#a9ac651d9e5097507c57b10dfeb40bfe5">vec_size</a>; j++) {</div>
|
||||
<div class="line"><a id="l00252" name="l00252"></a><span class="lineno"> 252</span> dst[i * kDstStrRow + j * kDstStrCol] = tmp_val[j];</div>
|
||||
<div class="line"><a id="l00252" name="l00252"></a><span class="lineno"> 252</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_block_loader_t.html#a6eb4e566b687395e27f290da288362db">dst</a>[i * kDstStrRow + j * kDstStrCol] = tmp_val[j];</div>
|
||||
<div class="line"><a id="l00253" name="l00253"></a><span class="lineno"> 253</span> }</div>
|
||||
<div class="line"><a id="l00254" name="l00254"></a><span class="lineno"> 254</span> }</div>
|
||||
<div class="line"><a id="l00255" name="l00255"></a><span class="lineno"> 255</span> }</div>
|
||||
@@ -377,28 +391,27 @@ $(function(){ initResizable(false); });
|
||||
<div class="line"><a id="l00262" name="l00262"></a><span class="lineno"> 262</span> </div>
|
||||
<div class="line"><a id="l00263" name="l00263"></a><span class="lineno"> 263</span>} <span class="comment">// namespace steel</span></div>
|
||||
<div class="line"><a id="l00264" name="l00264"></a><span class="lineno"> 264</span>} <span class="comment">// namespace mlx</span></div>
|
||||
<div class="ttc" id="acommon_2binary_8h_html_a70228731d29946574b238d21fb4b360c"><div class="ttname"><a href="common_2binary_8h.html#a70228731d29946574b238d21fb4b360c">op</a></div><div class="ttdeci">Op op</div><div class="ttdef"><b>Definition</b> binary.h:129</div></div>
|
||||
<div class="ttc" id="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_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_block_loader_1_1_read_vector_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader_1_1_read_vector.html">mlx::steel::BlockLoader::ReadVector</a></div><div class="ttdef"><b>Definition</b> loader.h:42</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_1_1_read_vector_html_a20963f7191251defca48bf8a843d019d"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader_1_1_read_vector.html#a20963f7191251defca48bf8a843d019d">mlx::steel::BlockLoader::ReadVector::v</a></div><div class="ttdeci">uint8_t v[sizeof(T) *vec_size]</div><div class="ttdef"><b>Definition</b> loader.h:43</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a064e2cc77e0b1cf0f8027929e031775b"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a064e2cc77e0b1cf0f8027929e031775b">mlx::steel::BlockLoader::thread_idx</a></div><div class="ttdeci">const short thread_idx</div><div class="ttdef"><b>Definition</b> loader.h:34</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html">mlx::steel::BlockLoader</a></div><div class="ttdef"><b>Definition</b> loader.h:25</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a064e2cc77e0b1cf0f8027929e031775b"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a064e2cc77e0b1cf0f8027929e031775b">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::thread_idx</a></div><div class="ttdeci">const short thread_idx</div><div class="ttdef"><b>Definition</b> loader.h:34</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a37aca066e63dff238865b5923a2d4335"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a37aca066e63dff238865b5923a2d4335">mlx::steel::BlockLoader::BlockLoader</a></div><div class="ttdeci">METAL_FUNC BlockLoader(const device T *src_, const int src_ld_, threadgroup T *dst_, ushort simd_group_id, ushort simd_lane_id)</div><div class="ttdef"><b>Definition</b> loader.h:47</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a58bdf9b9c81962733e22ecdeae28c092"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a58bdf9b9c81962733e22ecdeae28c092">mlx::steel::BlockLoader::vec_size</a></div><div class="ttdeci">STEEL_CONST short vec_size</div><div class="ttdef"><b>Definition</b> loader.h:27</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a58bdf9b9c81962733e22ecdeae28c092"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a58bdf9b9c81962733e22ecdeae28c092">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::vec_size</a></div><div class="ttdeci">STEEL_CONST short vec_size</div><div class="ttdef"><b>Definition</b> loader.h:27</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a6af21428f0e7c17b48ddedf4dd20a1e8"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a6af21428f0e7c17b48ddedf4dd20a1e8">mlx::steel::BlockLoader::next</a></div><div class="ttdeci">METAL_FUNC void next()</div><div class="ttdef"><b>Definition</b> loader.h:131</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a6c9e27f11f48b34580ed2c7e9cad9a27"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a6c9e27f11f48b34580ed2c7e9cad9a27">mlx::steel::BlockLoader::load_unsafe</a></div><div class="ttdeci">METAL_FUNC void load_unsafe() const</div><div class="ttdef"><b>Definition</b> loader.h:74</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a78c326e75ee35a484685771143047cd4"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a78c326e75ee35a484685771143047cd4">mlx::steel::BlockLoader::bj</a></div><div class="ttdeci">const short bj</div><div class="ttdef"><b>Definition</b> loader.h:36</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a973804e5b1d418c98c90861cda1a6fb5"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a973804e5b1d418c98c90861cda1a6fb5">mlx::steel::BlockLoader::n_rows</a></div><div class="ttdeci">STEEL_CONST short n_rows</div><div class="ttdef"><b>Definition</b> loader.h:26</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a9ef13742bcdf07532d8f09394928a8af"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a9ef13742bcdf07532d8f09394928a8af">mlx::steel::BlockLoader::bi</a></div><div class="ttdeci">const short bi</div><div class="ttdef"><b>Definition</b> loader.h:35</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_aadafc50f7f06af434149d7469df4714d"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#aadafc50f7f06af434149d7469df4714d">mlx::steel::BlockLoader::src_ld</a></div><div class="ttdeci">const int src_ld</div><div class="ttdef"><b>Definition</b> loader.h:30</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_ab87876699d55473620c7ea99f9da911d"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#ab87876699d55473620c7ea99f9da911d">mlx::steel::BlockLoader::tile_stride</a></div><div class="ttdeci">const int tile_stride</div><div class="ttdef"><b>Definition</b> loader.h:31</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a78c326e75ee35a484685771143047cd4"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a78c326e75ee35a484685771143047cd4">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::bj</a></div><div class="ttdeci">const short bj</div><div class="ttdef"><b>Definition</b> loader.h:36</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a973804e5b1d418c98c90861cda1a6fb5"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a973804e5b1d418c98c90861cda1a6fb5">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::n_rows</a></div><div class="ttdeci">STEEL_CONST short n_rows</div><div class="ttdef"><b>Definition</b> loader.h:26</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_a9ef13742bcdf07532d8f09394928a8af"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#a9ef13742bcdf07532d8f09394928a8af">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::bi</a></div><div class="ttdeci">const short bi</div><div class="ttdef"><b>Definition</b> loader.h:35</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_aadafc50f7f06af434149d7469df4714d"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#aadafc50f7f06af434149d7469df4714d">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::src_ld</a></div><div class="ttdeci">const int src_ld</div><div class="ttdef"><b>Definition</b> loader.h:30</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_ab87876699d55473620c7ea99f9da911d"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#ab87876699d55473620c7ea99f9da911d">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::tile_stride</a></div><div class="ttdeci">const int tile_stride</div><div class="ttdef"><b>Definition</b> loader.h:31</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_abb0f4f66ec8b123627beb8eb4fbb609d"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#abb0f4f66ec8b123627beb8eb4fbb609d">mlx::steel::BlockLoader::load_safe</a></div><div class="ttdeci">METAL_FUNC void load_safe(short2 src_tile_dim) const</div><div class="ttdef"><b>Definition</b> loader.h:83</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_ad1db14517568ae9eddfb6986ef31c7aa"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#ad1db14517568ae9eddfb6986ef31c7aa">mlx::steel::BlockLoader::src</a></div><div class="ttdeci">const device T * src</div><div class="ttdef"><b>Definition</b> loader.h:40</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_ad1db14517568ae9eddfb6986ef31c7aa"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#ad1db14517568ae9eddfb6986ef31c7aa">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::src</a></div><div class="ttdeci">const device T * src</div><div class="ttdef"><b>Definition</b> loader.h:40</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_adb4ca2cc193630a779de552fa8847ddf"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#adb4ca2cc193630a779de552fa8847ddf">mlx::steel::BlockLoader::apply_inplace_op</a></div><div class="ttdeci">METAL_FUNC void apply_inplace_op(thread const UnaryOp &op) const</div><div class="ttdef"><b>Definition</b> loader.h:63</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_af1c6c35a42e9da4408c1013ff1741bc2"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">mlx::steel::BlockLoader::dst</a></div><div class="ttdeci">threadgroup T * dst</div><div class="ttdef"><b>Definition</b> loader.h:39</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_t_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader_t.html">mlx::steel::BlockLoaderT</a></div><div class="ttdef"><b>Definition</b> loader.h:153</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_html_af1c6c35a42e9da4408c1013ff1741bc2"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader.html#af1c6c35a42e9da4408c1013ff1741bc2">mlx::steel::BlockLoader< T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size >::dst</a></div><div class="ttdeci">threadgroup T * dst</div><div class="ttdef"><b>Definition</b> loader.h:39</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_t_html_a076616a7c67ad1b847e0e6b046077ee2"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader_t.html#a076616a7c67ad1b847e0e6b046077ee2">mlx::steel::BlockLoaderT::BlockLoaderT</a></div><div class="ttdeci">METAL_FUNC BlockLoaderT(const device T *src_, const int src_ld_, threadgroup T *dst_, ushort simd_group_id, ushort simd_lane_id)</div><div class="ttdef"><b>Definition</b> loader.h:171</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_t_html_a0ccc7caa93e6e709981a1a08159d41dc"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader_t.html#a0ccc7caa93e6e709981a1a08159d41dc">mlx::steel::BlockLoaderT::n_rows</a></div><div class="ttdeci">STEEL_CONST short n_rows</div><div class="ttdef"><b>Definition</b> loader.h:154</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_block_loader_t_html_a2b136fad00dc54300e68aa6b905eff97"><div class="ttname"><a href="structmlx_1_1steel_1_1_block_loader_t.html#a2b136fad00dc54300e68aa6b905eff97">mlx::steel::BlockLoaderT::apply_inplace_op</a></div><div class="ttdeci">METAL_FUNC void apply_inplace_op(thread const UnaryOp &op) const</div><div class="ttdef"><b>Definition</b> loader.h:187</div></div>
|
||||
@@ -417,10 +430,13 @@ $(function(){ initResizable(false); });
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_c_shape_html_a01b09227356b6a682a0694523a8e6901"><div class="ttname"><a href="structmlx_1_1steel_1_1_c_shape.html#a01b09227356b6a682a0694523a8e6901">mlx::steel::CShape::kCols</a></div><div class="ttdeci">STEEL_CONST int kCols</div><div class="ttdef"><b>Definition</b> loader.h:139</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_c_shape_html_a5caf36cb9acf9f90ba59a9b0b4197993"><div class="ttname"><a href="structmlx_1_1steel_1_1_c_shape.html#a5caf36cb9acf9f90ba59a9b0b4197993">mlx::steel::CShape::kRows</a></div><div class="ttdeci">STEEL_CONST int kRows</div><div class="ttdef"><b>Definition</b> loader.h:138</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_e1756c7634b0c14aead026895ad71c6d.html">attn</a></li><li class="navelem"><a class="el" href="attn_2loader_8h.html">loader.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