rebase

2025-09-19 02:38:09 +08:00 · 2025-01-09 21:56:20 +00:00
parent 04b749a588
commit d8d647015b
2642 changed files with 137687 additions and 70861 deletions
--- a/docs/build/html/attn_8h_source.html
+++ b/docs/build/html/attn_8h_source.html
@@ -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/attn.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&amp;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()">&#160;</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&amp;dn=expat.txt MIT */
 var searchBox = new SearchBox("searchBox", "search/",'.html');
@@ -44,22 +63,23 @@ var searchBox = new SearchBox("searchBox", "search/",'.html');
 $(function() { codefold.init(0); });
 /* @license-end */
 </script>
-<script type="text/javascript" src="menudata.js"></script>
-<script type="text/javascript" src="menu.js"></script>
+</div><!-- top -->
+<div id="side-nav" class="ui-resizable side-nav-resizable">
+  <div id="nav-tree">
+    <div id="nav-tree-contents">
+      <div id="nav-sync" class="sync"></div>
+    </div>
+  </div>
+  <div id="splitbar" style="-moz-user-select:none;" 
+       class="ui-resizable-handle">
+  </div>
+</div>
 <script type="text/javascript">
 /* @license magnet:?xt=urn:btih:d3d9a9a6595521f9666a5e94cc830dab83b65699&amp;dn=expat.txt MIT */
-$(function() {
-  initMenu('',true,false,'search.php','Search',false);
-  $(function() { init_search(); });
-});
-/* @license-end */
-</script>
-<div id="main-nav"></div>
-<script type="text/javascript">
-/* @license magnet:?xt=urn:btih:d3d9a9a6595521f9666a5e94cc830dab83b65699&amp;dn=expat.txt MIT */
-$(function(){ initResizable(false); });
+$(function(){initNavTree('attn_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">attn.h</div></div>
 </div><!--header-->
@@ -103,12 +117,12 @@ $(function(){ initResizable(false); });
 <div class="line"><a id="l00010" name="l00010"></a><span class="lineno">   10</span><span class="preprocessor">#include &quot;<a class="code" href="backend_2metal_2kernels_2steel_2utils_8h.html">mlx/backend/metal/kernels/steel/utils.h</a>&quot;</span></div>
 <div class="line"><a id="l00011" name="l00011"></a><span class="lineno">   11</span> </div>
 <div class="line"><a id="l00012" name="l00012"></a><span class="lineno">   12</span><span class="keyword">using namespace </span><a class="code hl_namespace" href="namespacemetal.html">metal</a>;</div>
-<div class="line"><a id="l00013" name="l00013"></a><span class="lineno">   13</span> </div>
+<div class="line"><a id="l00013" name="l00013"></a><span class="lineno">   13</span></div>
 <div class="line"><a id="l00015" name="l00015"></a><span class="lineno">   15</span><span class="comment">// GEMM kernel class</span></div>
 <div class="line"><a id="l00017" name="l00017"></a><span class="lineno">   17</span> </div>
 <div class="line"><a id="l00018" name="l00018"></a><span class="lineno">   18</span><span class="keyword">namespace </span><a class="code hl_namespace" href="namespacemlx.html">mlx</a> {</div>
 <div class="foldopen" id="foldopen00019" data-start="{" data-end="}">
-<div class="line"><a id="l00019" name="l00019"></a><span class="lineno"><a class="line" href="namespacemlx_1_1steel.html">   19</a></span><span class="keyword">namespace </span>steel {</div>
+<div class="line"><a id="l00019" name="l00019"></a><span class="lineno"><a class="line" href="namespacemlx_1_1steel.html">   19</a></span><span class="keyword">namespace </span><a class="code hl_namespace" href="namespacemlx_1_1steel.html">steel</a> {</div>
 <div class="line"><a id="l00020" name="l00020"></a><span class="lineno">   20</span> </div>
 <div class="line"><a id="l00021" name="l00021"></a><span class="lineno">   21</span><span class="keyword">template</span> &lt;<span class="keywordtype">bool</span> M_aligned, <span class="keywordtype">bool</span> N_aligned, <span class="keywordtype">bool</span> K_aligned&gt;</div>
 <div class="line"><a id="l00022" name="l00022"></a><span class="lineno">   22</span><span class="keyword">struct </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_loop_alignment.html">LoopAlignment</a> {};</div>
@@ -138,21 +152,21 @@ $(function(){ initResizable(false); });
 <div class="line"><a id="l00046" name="l00046"></a><span class="lineno">   46</span> </div>
 <div class="line"><a id="l00047" name="l00047"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#a9058ddb73e30e83fb9c548ba22817d64">   47</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_g_e_m_m_kernel.html#a9058ddb73e30e83fb9c548ba22817d64">tgp_size</a> = WM * WN * 32;</div>
 <div class="line"><a id="l00048" name="l00048"></a><span class="lineno">   48</span> </div>
-<div class="line"><a id="l00049" name="l00049"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#a98b6ec692580510081e2aa887a61944b">   49</a></span>  <span class="keyword">using </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">loader_a_t</a> = <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">BlockLoader</a>&lt;</div>
+<div class="line"><a id="l00049" name="l00049"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa8a04ed74d2259f99b337d4662c64d83">   49</a></span>  <span class="keyword">using </span><a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa8a04ed74d2259f99b337d4662c64d83">loader_a_t</a> = <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">BlockLoader</a>&lt;</div>
 <div class="line"><a id="l00050" name="l00050"></a><span class="lineno">   50</span>      T,</div>
 <div class="line"><a id="l00051" name="l00051"></a><span class="lineno">   51</span>      transpose_a ? BK : BM,</div>
 <div class="line"><a id="l00052" name="l00052"></a><span class="lineno">   52</span>      transpose_a ? BM : BK,</div>
 <div class="line"><a id="l00053" name="l00053"></a><span class="lineno">   53</span>      transpose_a ? BM + <a class="code hl_variable" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#ad547704ccbff6c2076abeffa6628c5a0">tgp_padding_a</a> : BK + <a class="code hl_variable" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#ad547704ccbff6c2076abeffa6628c5a0">tgp_padding_a</a>,</div>
 <div class="line"><a id="l00054" name="l00054"></a><span class="lineno">   54</span>      !transpose_a,</div>
 <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_g_e_m_m_kernel.html#a9058ddb73e30e83fb9c548ba22817d64">tgp_size</a>&gt;;</div>
-<div class="line"><a id="l00056" name="l00056"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#a1a115d5af0fb6e260165adba2e377635">   56</a></span>  <span class="keyword">using </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">loader_b_t</a> = <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">BlockLoader</a>&lt;</div>
+<div class="line"><a id="l00056" name="l00056"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa98f32278b5fd98c93ae5483c3596395">   56</a></span>  <span class="keyword">using </span><a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa98f32278b5fd98c93ae5483c3596395">loader_b_t</a> = <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">BlockLoader</a>&lt;</div>
 <div class="line"><a id="l00057" name="l00057"></a><span class="lineno">   57</span>      T,</div>
 <div class="line"><a id="l00058" name="l00058"></a><span class="lineno">   58</span>      transpose_b ? BN : BK,</div>
 <div class="line"><a id="l00059" name="l00059"></a><span class="lineno">   59</span>      transpose_b ? BK : BN,</div>
 <div class="line"><a id="l00060" name="l00060"></a><span class="lineno">   60</span>      transpose_b ? BK + <a class="code hl_variable" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#ad1b03941e869017558423c08b08bc094">tgp_padding_b</a> : BN + <a class="code hl_variable" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#ad1b03941e869017558423c08b08bc094">tgp_padding_b</a>,</div>
 <div class="line"><a id="l00061" name="l00061"></a><span class="lineno">   61</span>      transpose_b,</div>
 <div class="line"><a id="l00062" name="l00062"></a><span class="lineno">   62</span>      <a class="code hl_variable" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#a9058ddb73e30e83fb9c548ba22817d64">tgp_size</a>&gt;;</div>
-<div class="line"><a id="l00063" name="l00063"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#ae52eb09c9478cd4f199662346ac0c83e">   63</a></span>  <span class="keyword">using </span><a class="code hl_struct" href="structmlx_1_1steel_1_1_block_m_m_a.html">mma_t</a> = <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_m_m_a.html">BlockMMA</a>&lt;</div>
+<div class="line"><a id="l00063" name="l00063"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#add8c6a31011a4895667c2a94a5af3782">   63</a></span>  <span class="keyword">using </span><a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#add8c6a31011a4895667c2a94a5af3782">mma_t</a> = <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_m_m_a.html">BlockMMA</a>&lt;</div>
 <div class="line"><a id="l00064" name="l00064"></a><span class="lineno">   64</span>      T,</div>
 <div class="line"><a id="l00065" name="l00065"></a><span class="lineno">   65</span>      U,</div>
 <div class="line"><a id="l00066" name="l00066"></a><span class="lineno">   66</span>      BM,</div>
@@ -174,9 +188,9 @@ $(function(){ initResizable(false); });
 <div class="line"><a id="l00081" name="l00081"></a><span class="lineno">   81</span>      threadgroup T* As [[threadgroup(0)]],</div>
 <div class="line"><a id="l00082" name="l00082"></a><span class="lineno">   82</span>      threadgroup T* Bs [[threadgroup(1)]],</div>
 <div class="line"><a id="l00083" name="l00083"></a><span class="lineno">   83</span>      <span class="keyword">const</span> <span class="keywordtype">int</span> gemm_k_iterations,</div>
-<div class="line"><a id="l00084" name="l00084"></a><span class="lineno">   84</span>      thread <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">loader_a_t</a>&amp; loader_a,</div>
-<div class="line"><a id="l00085" name="l00085"></a><span class="lineno">   85</span>      thread <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">loader_b_t</a>&amp; loader_b,</div>
-<div class="line"><a id="l00086" name="l00086"></a><span class="lineno">   86</span>      thread <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_m_m_a.html">mma_t</a>&amp; mma_op,</div>
+<div class="line"><a id="l00084" name="l00084"></a><span class="lineno">   84</span>      thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa8a04ed74d2259f99b337d4662c64d83">loader_a_t</a>&amp; loader_a,</div>
+<div class="line"><a id="l00085" name="l00085"></a><span class="lineno">   85</span>      thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa98f32278b5fd98c93ae5483c3596395">loader_b_t</a>&amp; loader_b,</div>
+<div class="line"><a id="l00086" name="l00086"></a><span class="lineno">   86</span>      thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#add8c6a31011a4895667c2a94a5af3782">mma_t</a>&amp; mma_op,</div>
 <div class="line"><a id="l00087" name="l00087"></a><span class="lineno">   87</span>      thread <span class="keyword">const</span> <span class="keywordtype">short</span>&amp; tgp_bm,</div>
 <div class="line"><a id="l00088" name="l00088"></a><span class="lineno">   88</span>      thread <span class="keyword">const</span> <span class="keywordtype">short</span>&amp; tgp_bn,</div>
 <div class="line"><a id="l00089" name="l00089"></a><span class="lineno">   89</span>      thread <span class="keyword">const</span> <span class="keywordtype">short</span>&amp; lbk,</div>
@@ -268,14 +282,14 @@ $(function(){ initResizable(false); });
 <div class="line"><a id="l00173" name="l00173"></a><span class="lineno">  173</span>    D += c_row_long * params-&gt;ldd + c_col_long;</div>
 <div class="line"><a id="l00174" name="l00174"></a><span class="lineno">  174</span> </div>
 <div class="line"><a id="l00175" name="l00175"></a><span class="lineno">  175</span>    <span class="comment">// Prepare threadgroup loading operations</span></div>
-<div class="line"><a id="l00176" name="l00176"></a><span class="lineno">  176</span>    thread <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">loader_a_t</a> loader_a(A, params-&gt;lda, As, simd_group_id, simd_lane_id);</div>
-<div class="line"><a id="l00177" name="l00177"></a><span class="lineno">  177</span>    thread <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_loader.html">loader_b_t</a> loader_b(B, params-&gt;ldb, Bs, simd_group_id, simd_lane_id);</div>
+<div class="line"><a id="l00176" name="l00176"></a><span class="lineno">  176</span>    thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa8a04ed74d2259f99b337d4662c64d83">loader_a_t</a> loader_a(A, params-&gt;lda, As, simd_group_id, simd_lane_id);</div>
+<div class="line"><a id="l00177" name="l00177"></a><span class="lineno">  177</span>    thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa98f32278b5fd98c93ae5483c3596395">loader_b_t</a> loader_b(B, params-&gt;ldb, Bs, simd_group_id, simd_lane_id);</div>
 <div class="line"><a id="l00178" name="l00178"></a><span class="lineno">  178</span> </div>
 <div class="line"><a id="l00179" name="l00179"></a><span class="lineno">  179</span>    <span class="comment">// Prepare threadgroup mma operation</span></div>
-<div class="line"><a id="l00180" name="l00180"></a><span class="lineno">  180</span>    thread <a class="code hl_struct" href="structmlx_1_1steel_1_1_block_m_m_a.html">mma_t</a> mma_op(simd_group_id, simd_lane_id);</div>
+<div class="line"><a id="l00180" name="l00180"></a><span class="lineno">  180</span>    thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#add8c6a31011a4895667c2a94a5af3782">mma_t</a> mma_op(simd_group_id, simd_lane_id);</div>
 <div class="line"><a id="l00181" name="l00181"></a><span class="lineno">  181</span> </div>
 <div class="line"><a id="l00182" name="l00182"></a><span class="lineno">  182</span>    <span class="keywordtype">int</span> gemm_k_iterations = params-&gt;gemm_k_iterations_aligned;</div>
-<div class="line"><a id="l00183" name="l00183"></a><span class="lineno">  183</span> </div>
+<div class="line"><a id="l00183" name="l00183"></a><span class="lineno">  183</span></div>
 <div class="line"><a id="l00185" name="l00185"></a><span class="lineno">  185</span>    <span class="comment">// MNK aligned loop</span></div>
 <div class="line"><a id="l00186" name="l00186"></a><span class="lineno">  186</span>    <span class="keywordflow">if</span> (MN_aligned) {</div>
 <div class="line"><a id="l00187" name="l00187"></a><span class="lineno">  187</span>      <span class="keywordflow">for</span> (<span class="keywordtype">int</span> k = 0; k &lt; gemm_k_iterations; k++) {</div>
@@ -397,6 +411,7 @@ $(function(){ initResizable(false); });
 <div class="ttc" id="agemm_2params_8h_html"><div class="ttname"><a href="gemm_2params_8h.html">params.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_a6653b28c9473087141eddce39878d4d3"><div class="ttname"><a href="namespacemetal.html#a6653b28c9473087141eddce39878d4d3">metal::min</a></div><div class="ttdeci">METAL_FUNC bfloat16_t min(bfloat16_t x, bfloat16_t y)</div><div class="ttdef"><b>Definition</b> bf16_math.h:232</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_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_accum_helper_html_ae52abf69e7ba6af1a73d65d57182ed26"><div class="ttname"><a href="structmlx_1_1steel_1_1_accum_helper.html#ae52abf69e7ba6af1a73d65d57182ed26">mlx::steel::AccumHelper::accum_type</a></div><div class="ttdeci">float accum_type</div><div class="ttdef"><b>Definition</b> transforms.h:57</div></div>
@@ -408,17 +423,23 @@ $(function(){ initResizable(false); });
 <div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_a1ec583584e69dcbbb72106390a4fc5da"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#a1ec583584e69dcbbb72106390a4fc5da">mlx::steel::GEMMKernel::tgp_mem_size</a></div><div class="ttdeci">STEEL_CONST short tgp_mem_size</div><div class="ttdef"><b>Definition</b> attn.h:45</div></div>
 <div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_a756d7bbcc96e2919cd65eec4bc135780"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#a756d7bbcc96e2919cd65eec4bc135780">mlx::steel::GEMMKernel::gemm_loop</a></div><div class="ttdeci">static METAL_FUNC void gemm_loop(threadgroup T *As, threadgroup T *Bs, const int gemm_k_iterations, thread loader_a_t &amp;loader_a, thread loader_b_t &amp;loader_b, thread mma_t &amp;mma_op, thread const short &amp;tgp_bm, thread const short &amp;tgp_bn, thread const short &amp;lbk, LoopAlignment&lt; M_aligned, N_aligned, K_aligned_ &gt; l={})</div><div class="ttdef"><b>Definition</b> attn.h:80</div></div>
 <div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_a9058ddb73e30e83fb9c548ba22817d64"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#a9058ddb73e30e83fb9c548ba22817d64">mlx::steel::GEMMKernel::tgp_size</a></div><div class="ttdeci">STEEL_CONST short tgp_size</div><div class="ttdef"><b>Definition</b> attn.h:47</div></div>
+<div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_aa8a04ed74d2259f99b337d4662c64d83"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa8a04ed74d2259f99b337d4662c64d83">mlx::steel::GEMMKernel::loader_a_t</a></div><div class="ttdeci">BlockLoader&lt; T, transpose_a ? BK :BM, transpose_a ? BM :BK, transpose_a ? BM+tgp_padding_a :BK+tgp_padding_a, !transpose_a, tgp_size &gt; loader_a_t</div><div class="ttdef"><b>Definition</b> attn.h:49</div></div>
+<div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_aa98f32278b5fd98c93ae5483c3596395"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#aa98f32278b5fd98c93ae5483c3596395">mlx::steel::GEMMKernel::loader_b_t</a></div><div class="ttdeci">BlockLoader&lt; T, transpose_b ? BN :BK, transpose_b ? BK :BN, transpose_b ? BK+tgp_padding_b :BN+tgp_padding_b, transpose_b, tgp_size &gt; loader_b_t</div><div class="ttdef"><b>Definition</b> attn.h:56</div></div>
 <div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_ac00b149d76a903c2f91b0f477dc5037f"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#ac00b149d76a903c2f91b0f477dc5037f">mlx::steel::GEMMKernel::tgp_mem_size_a</a></div><div class="ttdeci">STEEL_CONST short tgp_mem_size_a</div><div class="ttdef"><b>Definition</b> attn.h:41</div></div>
 <div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_ad1b03941e869017558423c08b08bc094"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#ad1b03941e869017558423c08b08bc094">mlx::steel::GEMMKernel::tgp_padding_b</a></div><div class="ttdeci">STEEL_CONST short tgp_padding_b</div><div class="ttdef"><b>Definition</b> attn.h:40</div></div>
 <div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_ad547704ccbff6c2076abeffa6628c5a0"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#ad547704ccbff6c2076abeffa6628c5a0">mlx::steel::GEMMKernel::tgp_padding_a</a></div><div class="ttdeci">STEEL_CONST short tgp_padding_a</div><div class="ttdef"><b>Definition</b> attn.h:39</div></div>
+<div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_kernel_html_add8c6a31011a4895667c2a94a5af3782"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_kernel.html#add8c6a31011a4895667c2a94a5af3782">mlx::steel::GEMMKernel::mma_t</a></div><div class="ttdeci">BlockMMA&lt; 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 &gt; mma_t</div><div class="ttdef"><b>Definition</b> attn.h:63</div></div>
 <div class="ttc" id="astructmlx_1_1steel_1_1_g_e_m_m_params_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_g_e_m_m_params.html">mlx::steel::GEMMParams</a></div><div class="ttdef"><b>Definition</b> params.h:12</div></div>
 <div class="ttc" id="astructmlx_1_1steel_1_1_loop_alignment_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_loop_alignment.html">mlx::steel::LoopAlignment</a></div><div class="ttdef"><b>Definition</b> gemm.h:21</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&#160;<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_8h.html">attn.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>