mirror of
https://github.com/ml-explore/mlx.git
synced 2025-09-22 05:08:08 +08:00
rebase
This commit is contained in:
162
docs/build/html/attn_2mma_8h_source.html
vendored
162
docs/build/html/attn_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 1.13.1"/>
|
||||
<meta name="generator" content="Doxygen 1.13.2"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1"/>
|
||||
<title>MLX: mlx/backend/metal/kernels/steel/attn/mma.h Source File</title>
|
||||
<link href="tabs.css" rel="stylesheet" type="text/css"/>
|
||||
@@ -52,7 +52,7 @@
|
||||
</table>
|
||||
</div>
|
||||
<!-- end header part -->
|
||||
<!-- Generated by Doxygen 1.13.1 -->
|
||||
<!-- Generated by Doxygen 1.13.2 -->
|
||||
<script type="text/javascript">
|
||||
/* @license magnet:?xt=urn:btih:d3d9a9a6595521f9666a5e94cc830dab83b65699&dn=expat.txt MIT */
|
||||
var searchBox = new SearchBox("searchBox", "search/",'.html');
|
||||
@@ -143,7 +143,8 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
</div>
|
||||
<div class="line"><a id="l00035" name="l00035"></a><span class="lineno"> 35</span> </div>
|
||||
<div class="line"><a id="l00036" name="l00036"></a><span class="lineno"> 36</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="line"><a id="l00037" name="l00037"></a><span class="lineno"> 37</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> {</div>
|
||||
<div class="foldopen" id="foldopen00037" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00037" name="l00037"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html"> 37</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> {</div>
|
||||
<div class="line"><a id="l00038" name="l00038"></a><span class="lineno"> 38</span> <span class="keyword">static_assert</span>(</div>
|
||||
<div class="line"><a id="l00039" name="l00039"></a><span class="lineno"> 39</span> kFragRows_ == 8,</div>
|
||||
<div class="line"><a id="l00040" name="l00040"></a><span class="lineno"> 40</span> <span class="stringliteral">"Only 8 x 8 fragment matrices are currently supported"</span>);</div>
|
||||
@@ -151,28 +152,30 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00042" name="l00042"></a><span class="lineno"> 42</span> kFragCols_ == 8,</div>
|
||||
<div class="line"><a id="l00043" name="l00043"></a><span class="lineno"> 43</span> <span class="stringliteral">"Only 8 x 8 fragment matrices are currently supported"</span>);</div>
|
||||
<div class="line"><a id="l00044" name="l00044"></a><span class="lineno"> 44</span>};</div>
|
||||
</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">template</span> <<span class="keyword">typename</span> T></div>
|
||||
<div class="line"><a id="l00047" name="l00047"></a><span class="lineno"> 47</span><span class="keyword">struct </span>BaseMMAFrag<T, 8, 8> {</div>
|
||||
<div class="line"><a id="l00048" name="l00048"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a2fe53db449c692226f23f6b99fb2c0d4"> 48</a></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="l00049" name="l00049"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a211102315e2afbcfcd2e2c201b638e9f"> 49</a></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="foldopen" id="foldopen00047" data-start="{" data-end="};">
|
||||
<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_3_01_t_00_018_00_018_01_4.html"> 47</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="l00048" name="l00048"></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#a2fe53db449c692226f23f6b99fb2c0d4"> 48</a></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_3_01_t_00_018_00_018_01_4.html#a2fe53db449c692226f23f6b99fb2c0d4">kFragRows</a> = 8;</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_3_01_t_00_018_00_018_01_4.html#a211102315e2afbcfcd2e2c201b638e9f"> 49</a></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_3_01_t_00_018_00_018_01_4.html#a211102315e2afbcfcd2e2c201b638e9f">kFragCols</a> = 8;</div>
|
||||
<div class="line"><a id="l00050" name="l00050"></a><span class="lineno"> 50</span> </div>
|
||||
<div class="line"><a id="l00051" name="l00051"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a3c34dfdc944db110f4735f1b25307cf0"> 51</a></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="l00051" name="l00051"></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#a3c34dfdc944db110f4735f1b25307cf0"> 51</a></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_3_01_t_00_018_00_018_01_4.html#a3c34dfdc944db110f4735f1b25307cf0">kElemsPerFrag</a> = (<a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a2fe53db449c692226f23f6b99fb2c0d4">kFragRows</a> * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a211102315e2afbcfcd2e2c201b638e9f">kFragCols</a>) / 32;</div>
|
||||
<div class="line"><a id="l00052" name="l00052"></a><span class="lineno"> 52</span> </div>
|
||||
<div class="line"><a id="l00053" name="l00053"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a76aa5aa690dbcc954e957d767fad661f"> 53</a></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="l00054" name="l00054"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd"> 54</a></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="l00053" name="l00053"></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#a76aa5aa690dbcc954e957d767fad661f"> 53</a></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_3_01_t_00_018_00_018_01_4.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a> = 1;</div>
|
||||
<div class="line"><a id="l00054" name="l00054"></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#a7c212200d86b4e93f274d99addf668bd"> 54</a></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_3_01_t_00_018_00_018_01_4.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> = 2;</div>
|
||||
<div class="line"><a id="l00055" name="l00055"></a><span class="lineno"> 55</span> </div>
|
||||
<div class="line"><a id="l00056" name="l00056"></a><span class="lineno"> 56</span> <span class="keyword">static_assert</span>(</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_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="l00057" name="l00057"></a><span class="lineno"> 57</span> <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a> * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> == <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a3c34dfdc944db110f4735f1b25307cf0">kElemsPerFrag</a>,</div>
|
||||
<div class="line"><a id="l00058" name="l00058"></a><span class="lineno"> 58</span> <span class="stringliteral">"MMAFrag shape is not consistent with MMAFrag size"</span>);</div>
|
||||
<div class="line"><a id="l00059" name="l00059"></a><span class="lineno"> 59</span> </div>
|
||||
<div class="line"><a id="l00060" name="l00060"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a25675ae18947a97c6e04157b540103a9"> 60</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="l00061" name="l00061"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b"> 61</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="l00062" name="l00062"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a3dcd4301390937f89ed1dde6d28e341f"> 62</a></span> <span class="keyword">typedef</span> metal::vec<T, kElemRows> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a3dcd4301390937f89ed1dde6d28e341f">row_frag_type</a>;</div>
|
||||
<div class="line"><a id="l00063" name="l00063"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#adbb262a3c872e26533b68a39db16459e"> 63</a></span> <span class="keyword">typedef</span> metal::vec<T, kElemCols> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#adbb262a3c872e26533b68a39db16459e">col_frag_type</a>;</div>
|
||||
<div class="line"><a id="l00060" name="l00060"></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#a25675ae18947a97c6e04157b540103a9"> 60</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_3_01_t_00_018_00_018_01_4.html#a25675ae18947a97c6e04157b540103a9">mat_type</a>;</div>
|
||||
<div class="line"><a id="l00061" name="l00061"></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#a8536bfaa108031c2ea3e9ccdc766ee5b"> 61</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_3_01_t_00_018_00_018_01_4.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>;</div>
|
||||
<div class="line"><a id="l00062" name="l00062"></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#a3dcd4301390937f89ed1dde6d28e341f"> 62</a></span> <span class="keyword">typedef</span> metal::vec<T, kElemRows> <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#a3dcd4301390937f89ed1dde6d28e341f">row_frag_type</a>;</div>
|
||||
<div class="line"><a id="l00063" name="l00063"></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#adbb262a3c872e26533b68a39db16459e"> 63</a></span> <span class="keyword">typedef</span> metal::vec<T, kElemCols> <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#adbb262a3c872e26533b68a39db16459e">col_frag_type</a>;</div>
|
||||
<div class="line"><a id="l00064" name="l00064"></a><span class="lineno"> 64</span> </div>
|
||||
<div class="foldopen" id="foldopen00065" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00065" name="l00065"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7331fff1d12f2f8b72b0006a3ad0dd83"> 65</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="l00065" name="l00065"></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"> 65</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="l00066" name="l00066"></a><span class="lineno"> 66</span> [[thread_index_in_simdgroup]]) {</div>
|
||||
<div class="line"><a id="l00067" name="l00067"></a><span class="lineno"> 67</span> <span class="keyword">const</span> <span class="keywordtype">short</span> qid = simd_lane_id / 4;</div>
|
||||
<div class="line"><a id="l00068" name="l00068"></a><span class="lineno"> 68</span> <span class="keyword">const</span> <span class="keywordtype">short</span> fm = (qid & 4) + ((simd_lane_id / 2) % 4);</div>
|
||||
@@ -184,12 +187,12 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00073" name="l00073"></a><span class="lineno"> 73</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="l00074" name="l00074"></a><span class="lineno"> 74</span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span></div>
|
||||
<div class="foldopen" id="foldopen00075" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00075" name="l00075"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#ac73006b36fc710feda3a7c796e21415c"> 75</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="l00075" name="l00075"></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"> 75</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& dst, SrcPtrType src, StrX str_x, StrY str_y) {</div>
|
||||
<div class="line"><a id="l00076" name="l00076"></a><span class="lineno"> 76</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00077" name="l00077"></a><span class="lineno"> 77</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="l00077" name="l00077"></a><span class="lineno"> 77</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_3_01_t_00_018_00_018_01_4.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a>; i++) {</div>
|
||||
<div class="line"><a id="l00078" name="l00078"></a><span class="lineno"> 78</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00079" name="l00079"></a><span class="lineno"> 79</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="l00080" name="l00080"></a><span class="lineno"> 80</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="l00079" name="l00079"></a><span class="lineno"> 79</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_3_01_t_00_018_00_018_01_4.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a>; j++) {</div>
|
||||
<div class="line"><a id="l00080" name="l00080"></a><span class="lineno"> 80</span> dst[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.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="l00081" name="l00081"></a><span class="lineno"> 81</span> }</div>
|
||||
<div class="line"><a id="l00082" name="l00082"></a><span class="lineno"> 82</span> }</div>
|
||||
<div class="line"><a id="l00083" name="l00083"></a><span class="lineno"> 83</span> }</div>
|
||||
@@ -204,8 +207,8 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00091" name="l00091"></a><span class="lineno"> 91</span> <span class="keyword">typename</span> OffX,</div>
|
||||
<div class="line"><a id="l00092" name="l00092"></a><span class="lineno"> 92</span> <span class="keyword">typename</span> OffY></div>
|
||||
<div class="foldopen" id="foldopen00093" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00093" name="l00093"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#ad22aaee4a2938cbdd315b39eda84e07d"> 93</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="l00094" name="l00094"></a><span class="lineno"> 94</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="l00093" name="l00093"></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"> 93</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="l00094" name="l00094"></a><span class="lineno"> 94</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& dst,</div>
|
||||
<div class="line"><a id="l00095" name="l00095"></a><span class="lineno"> 95</span> SrcPtrType src,</div>
|
||||
<div class="line"><a id="l00096" name="l00096"></a><span class="lineno"> 96</span> StrX str_x,</div>
|
||||
<div class="line"><a id="l00097" name="l00097"></a><span class="lineno"> 97</span> StrY str_y,</div>
|
||||
@@ -231,14 +234,14 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00116" name="l00116"></a><span class="lineno"> 116</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="l00117" name="l00117"></a><span class="lineno"> 117</span> METAL_FUNC <span class="keyword">static</span> <span class="keyword">constexpr</span> <span class="keywordtype">void</span></div>
|
||||
<div class="foldopen" id="foldopen00118" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00118" name="l00118"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#aa8f50ea8961ec5b35c1b81366d64f2cb"> 118</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="l00118" name="l00118"></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"> 118</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& src, DstPtrType dst, StrX str_x, StrY str_y) {</div>
|
||||
<div class="line"><a id="l00119" name="l00119"></a><span class="lineno"> 119</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="l00120" name="l00120"></a><span class="lineno"> 120</span> </div>
|
||||
<div class="line"><a id="l00121" name="l00121"></a><span class="lineno"> 121</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00122" name="l00122"></a><span class="lineno"> 122</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="l00122" name="l00122"></a><span class="lineno"> 122</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_3_01_t_00_018_00_018_01_4.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a>; i++) {</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_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a>; j++) {</div>
|
||||
<div class="line"><a id="l00125" name="l00125"></a><span class="lineno"> 125</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="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_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a>; j++) {</div>
|
||||
<div class="line"><a id="l00125" name="l00125"></a><span class="lineno"> 125</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_3_01_t_00_018_00_018_01_4.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> + 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>
|
||||
@@ -253,8 +256,8 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00136" name="l00136"></a><span class="lineno"> 136</span> <span class="keyword">typename</span> OffX,</div>
|
||||
<div class="line"><a id="l00137" name="l00137"></a><span class="lineno"> 137</span> <span class="keyword">typename</span> OffY></div>
|
||||
<div class="foldopen" id="foldopen00138" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00138" name="l00138"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a1f0b00daad8eba2f855bb306e70d2328"> 138</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="l00139" name="l00139"></a><span class="lineno"> 139</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="l00138" name="l00138"></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"> 138</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="l00139" name="l00139"></a><span class="lineno"> 139</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& src,</div>
|
||||
<div class="line"><a id="l00140" name="l00140"></a><span class="lineno"> 140</span> DstPtrType dst,</div>
|
||||
<div class="line"><a id="l00141" name="l00141"></a><span class="lineno"> 141</span> StrX str_x,</div>
|
||||
<div class="line"><a id="l00142" name="l00142"></a><span class="lineno"> 142</span> StrY str_y,</div>
|
||||
@@ -278,40 +281,40 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
</div>
|
||||
<div class="line"><a id="l00160" name="l00160"></a><span class="lineno"> 160</span> </div>
|
||||
<div class="foldopen" id="foldopen00161" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00161" name="l00161"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8028512f5a3d2b6acaf966be529627a3"> 161</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="l00162" name="l00162"></a><span class="lineno"> 162</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="l00163" name="l00163"></a><span class="lineno"> 163</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="l00164" name="l00164"></a><span class="lineno"> 164</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="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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& C) {</div>
|
||||
<div class="line"><a id="l00166" name="l00166"></a><span class="lineno"> 166</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="l00167" name="l00167"></a><span class="lineno"> 167</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="l00168" name="l00168"></a><span class="lineno"> 168</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="l00169" name="l00169"></a><span class="lineno"> 169</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="l00161" name="l00161"></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"> 161</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="l00162" name="l00162"></a><span class="lineno"> 162</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& D,</div>
|
||||
<div class="line"><a id="l00163" name="l00163"></a><span class="lineno"> 163</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& A,</div>
|
||||
<div class="line"><a id="l00164" name="l00164"></a><span class="lineno"> 164</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& B,</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& C) {</div>
|
||||
<div class="line"><a id="l00166" name="l00166"></a><span class="lineno"> 166</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#a25675ae18947a97c6e04157b540103a9">mat_type</a> D_mat;</div>
|
||||
<div class="line"><a id="l00167" name="l00167"></a><span class="lineno"> 167</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#a25675ae18947a97c6e04157b540103a9">mat_type</a> A_mat;</div>
|
||||
<div class="line"><a id="l00168" name="l00168"></a><span class="lineno"> 168</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#a25675ae18947a97c6e04157b540103a9">mat_type</a> B_mat;</div>
|
||||
<div class="line"><a id="l00169" name="l00169"></a><span class="lineno"> 169</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#a25675ae18947a97c6e04157b540103a9">mat_type</a> C_mat;</div>
|
||||
<div class="line"><a id="l00170" name="l00170"></a><span class="lineno"> 170</span> </div>
|
||||
<div class="line"><a id="l00171" name="l00171"></a><span class="lineno"> 171</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="l00172" name="l00172"></a><span class="lineno"> 172</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="l00173" name="l00173"></a><span class="lineno"> 173</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="l00171" name="l00171"></a><span class="lineno"> 171</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>&<span class="keyword">></span>(A_mat.thread_elements()) = A;</div>
|
||||
<div class="line"><a id="l00172" name="l00172"></a><span class="lineno"> 172</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>&<span class="keyword">></span>(B_mat.thread_elements()) = B;</div>
|
||||
<div class="line"><a id="l00173" name="l00173"></a><span class="lineno"> 173</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>&<span class="keyword">></span>(C_mat.thread_elements()) = C;</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> <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="l00175" name="l00175"></a><span class="lineno"> 175</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>(D_mat, A_mat, B_mat, C_mat);</div>
|
||||
<div class="line"><a id="l00176" name="l00176"></a><span class="lineno"> 176</span> </div>
|
||||
<div class="line"><a id="l00177" name="l00177"></a><span class="lineno"> 177</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="l00177" name="l00177"></a><span class="lineno"> 177</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>&<span class="keyword">></span>(D_mat.thread_elements());</div>
|
||||
<div class="line"><a id="l00178" name="l00178"></a><span class="lineno"> 178</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00179" name="l00179"></a><span class="lineno"> 179</span> </div>
|
||||
<div class="foldopen" id="foldopen00180" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00180" name="l00180"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a1868f57d57c8adedab2c58492ec76946"> 180</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="l00181" name="l00181"></a><span class="lineno"> 181</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="l00182" name="l00182"></a><span class="lineno"> 182</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="l00183" name="l00183"></a><span class="lineno"> 183</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="l00184" name="l00184"></a><span class="lineno"> 184</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="l00180" name="l00180"></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"> 180</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="l00181" name="l00181"></a><span class="lineno"> 181</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#a25675ae18947a97c6e04157b540103a9">mat_type</a>& D,</div>
|
||||
<div class="line"><a id="l00182" name="l00182"></a><span class="lineno"> 182</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#a25675ae18947a97c6e04157b540103a9">mat_type</a>& A,</div>
|
||||
<div class="line"><a id="l00183" name="l00183"></a><span class="lineno"> 183</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#a25675ae18947a97c6e04157b540103a9">mat_type</a>& B,</div>
|
||||
<div class="line"><a id="l00184" name="l00184"></a><span class="lineno"> 184</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#a25675ae18947a97c6e04157b540103a9">mat_type</a>& C) {</div>
|
||||
<div class="line"><a id="l00185" name="l00185"></a><span class="lineno"> 185</span> simdgroup_multiply_accumulate(D, A, B, C);</div>
|
||||
<div class="line"><a id="l00186" name="l00186"></a><span class="lineno"> 186</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00187" name="l00187"></a><span class="lineno"> 187</span> </div>
|
||||
<div class="line"><a id="l00188" name="l00188"></a><span class="lineno"> 188</span> <span class="keyword">template</span> <<span class="keyword">typename</span> Op></div>
|
||||
<div class="foldopen" id="foldopen00189" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00189" name="l00189"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a51d662e4cff88b5ad17d7c44bb6b6970"> 189</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#a51d662e4cff88b5ad17d7c44bb6b6970">row_reduce</a>(</div>
|
||||
<div class="line"><a id="l00190" name="l00190"></a><span class="lineno"> 190</span> thread <span class="keyword">const</span> <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& inp_vals,</div>
|
||||
<div class="line"><a id="l00189" name="l00189"></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#a51d662e4cff88b5ad17d7c44bb6b6970"> 189</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#a51d662e4cff88b5ad17d7c44bb6b6970">row_reduce</a>(</div>
|
||||
<div class="line"><a id="l00190" name="l00190"></a><span class="lineno"> 190</span> thread <span class="keyword">const</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& inp_vals,</div>
|
||||
<div class="line"><a id="l00191" name="l00191"></a><span class="lineno"> 191</span> thread T* reduced_vals) {</div>
|
||||
<div class="line"><a id="l00192" name="l00192"></a><span class="lineno"> 192</span> T thr_reduce = Op::apply(inp_vals.x, inp_vals.y);</div>
|
||||
<div class="line"><a id="l00193" name="l00193"></a><span class="lineno"> 193</span> </div>
|
||||
@@ -327,27 +330,29 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00202" name="l00202"></a><span class="lineno"> 202</span> </div>
|
||||
<div class="line"><a id="l00203" name="l00203"></a><span class="lineno"> 203</span> <span class="keyword">template</span> <<span class="keyword">typename</span> Op></div>
|
||||
<div class="foldopen" id="foldopen00204" data-start="{" data-end="}">
|
||||
<div class="line"><a id="l00204" name="l00204"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a318c4279bdc7b39b7919f108b1cd8010"> 204</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#a318c4279bdc7b39b7919f108b1cd8010">row_bin_op</a>(</div>
|
||||
<div class="line"><a id="l00205" name="l00205"></a><span class="lineno"> 205</span> thread <a class="code hl_typedef" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& inp_vals,</div>
|
||||
<div class="line"><a id="l00204" name="l00204"></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#a318c4279bdc7b39b7919f108b1cd8010"> 204</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#a318c4279bdc7b39b7919f108b1cd8010">row_bin_op</a>(</div>
|
||||
<div class="line"><a id="l00205" name="l00205"></a><span class="lineno"> 205</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#a8536bfaa108031c2ea3e9ccdc766ee5b">frag_type</a>& inp_vals,</div>
|
||||
<div class="line"><a id="l00206" name="l00206"></a><span class="lineno"> 206</span> thread T* row_vals) {</div>
|
||||
<div class="line"><a id="l00207" name="l00207"></a><span class="lineno"> 207</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00208" name="l00208"></a><span class="lineno"> 208</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="l00208" name="l00208"></a><span class="lineno"> 208</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_3_01_t_00_018_00_018_01_4.html#a76aa5aa690dbcc954e957d767fad661f">kElemRows</a>; i++) {</div>
|
||||
<div class="line"><a id="l00209" name="l00209"></a><span class="lineno"> 209</span> <a class="code hl_define" href="steel_2defines_8h.html#a5a5c3095b132a7589bc19cd5cb80e2c6">STEEL_PRAGMA_UNROLL</a></div>
|
||||
<div class="line"><a id="l00210" name="l00210"></a><span class="lineno"> 210</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="l00211" name="l00211"></a><span class="lineno"> 211</span> inp_vals[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="l00212" name="l00212"></a><span class="lineno"> 212</span> Op::apply(inp_vals[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> + j], row_vals[i]);</div>
|
||||
<div class="line"><a id="l00210" name="l00210"></a><span class="lineno"> 210</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_3_01_t_00_018_00_018_01_4.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a>; j++) {</div>
|
||||
<div class="line"><a id="l00211" name="l00211"></a><span class="lineno"> 211</span> inp_vals[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> + j] =</div>
|
||||
<div class="line"><a id="l00212" name="l00212"></a><span class="lineno"> 212</span> Op::apply(inp_vals[i * <a class="code hl_variable" href="structmlx_1_1steel_1_1_base_m_m_a_frag_3_01_t_00_018_00_018_01_4.html#a7c212200d86b4e93f274d99addf668bd">kElemCols</a> + j], row_vals[i]);</div>
|
||||
<div class="line"><a id="l00213" name="l00213"></a><span class="lineno"> 213</span> }</div>
|
||||
<div class="line"><a id="l00214" name="l00214"></a><span class="lineno"> 214</span> }</div>
|
||||
<div class="line"><a id="l00215" name="l00215"></a><span class="lineno"> 215</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00216" name="l00216"></a><span class="lineno"> 216</span>};</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00217" name="l00217"></a><span class="lineno"> 217</span> </div>
|
||||
<div class="line"><a id="l00218" name="l00218"></a><span class="lineno"> 218</span><span class="keyword">template</span> <</div>
|
||||
<div class="line"><a id="l00219" name="l00219"></a><span class="lineno"> 219</span> <span class="keyword">typename</span> T,</div>
|
||||
<div class="line"><a id="l00220" name="l00220"></a><span class="lineno"> 220</span> <span class="keywordtype">int</span> kTileRows_,</div>
|
||||
<div class="line"><a id="l00221" name="l00221"></a><span class="lineno"> 221</span> <span class="keywordtype">int</span> kTileCols_,</div>
|
||||
<div class="line"><a id="l00222" name="l00222"></a><span class="lineno"> 222</span> <span class="keyword">class </span>MMAFrag_ = BaseMMAFrag<T, 8, 8>></div>
|
||||
<div class="line"><a id="l00223" name="l00223"></a><span class="lineno"> 223</span><span class="keyword">struct </span>MMATile {</div>
|
||||
<div class="foldopen" id="foldopen00223" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00223" name="l00223"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html"> 223</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="l00224" name="l00224"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#abe33de70e34300745bad9aa822fd0382"> 224</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="l00225" name="l00225"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a80078f0dfa4c225e79d9b460202d5e2c"> 225</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="l00226" name="l00226"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_m_m_a_tile.html#a594142f957ffb99296a243f7af7b59e7"> 226</a></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>
|
||||
@@ -563,6 +568,7 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00409" name="l00409"></a><span class="lineno"> 409</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00410" name="l00410"></a><span class="lineno"> 410</span>};</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00411" name="l00411"></a><span class="lineno"> 411</span> </div>
|
||||
<div class="line"><a id="l00412" name="l00412"></a><span class="lineno"> 412</span><span class="keyword">template</span> <<span class="keyword">typename</span> T, <span class="keyword">typename</span> U, <span class="keywordtype">int</span> M, <span class="keywordtype">int</span> N, <span class="keywordtype">int</span> K></div>
|
||||
<div class="foldopen" id="foldopen00413" data-start="{" data-end="}">
|
||||
@@ -603,7 +609,8 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00446" name="l00446"></a><span class="lineno"> 446</span> <span class="keywordtype">short</span> ldb_tgp,</div>
|
||||
<div class="line"><a id="l00447" name="l00447"></a><span class="lineno"> 447</span> <span class="keyword">typename</span> AccumType = float,</div>
|
||||
<div class="line"><a id="l00448" name="l00448"></a><span class="lineno"> 448</span> <span class="keyword">typename</span> Epilogue = TransformNone<U, AccumType>></div>
|
||||
<div class="line"><a id="l00449" name="l00449"></a><span class="lineno"> 449</span><span class="keyword">struct </span>BlockMMA {</div>
|
||||
<div class="foldopen" id="foldopen00449" data-start="{" data-end="};">
|
||||
<div class="line"><a id="l00449" name="l00449"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_m_m_a.html"> 449</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="l00450" name="l00450"></a><span class="lineno"> 450</span> <span class="comment">// MMAFrag size</span></div>
|
||||
<div class="line"><a id="l00451" name="l00451"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_m_m_a.html#aee8caec45c1f9e4428586effbfe6137d"> 451</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_m_m_a.html#aee8caec45c1f9e4428586effbfe6137d">kFragSize</a> = 8;</div>
|
||||
<div class="line"><a id="l00452" name="l00452"></a><span class="lineno"><a class="line" href="structmlx_1_1steel_1_1_block_m_m_a.html#ae2c42cb6d0dde785859164c195f4d13c"> 452</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>
|
||||
@@ -896,6 +903,7 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="line"><a id="l00722" name="l00722"></a><span class="lineno"> 722</span> }</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00723" name="l00723"></a><span class="lineno"> 723</span>};</div>
|
||||
</div>
|
||||
<div class="line"><a id="l00724" name="l00724"></a><span class="lineno"> 724</span> </div>
|
||||
<div class="line"><a id="l00725" name="l00725"></a><span class="lineno"> 725</span>} <span class="comment">// namespace steel</span></div>
|
||||
<div class="line"><a id="l00726" name="l00726"></a><span class="lineno"> 726</span>} <span class="comment">// namespace mlx</span></div>
|
||||
@@ -911,25 +919,25 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<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_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_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:180</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:138</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_a318c4279bdc7b39b7919f108b1cd8010"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a318c4279bdc7b39b7919f108b1cd8010">mlx::steel::BaseMMAFrag< T, 8, 8 >::row_bin_op</a></div><div class="ttdeci">static METAL_FUNC constexpr void row_bin_op(thread frag_type &inp_vals, thread T *row_vals)</div><div class="ttdef"><b>Definition</b> mma.h:204</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_a3dcd4301390937f89ed1dde6d28e341f"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a3dcd4301390937f89ed1dde6d28e341f">mlx::steel::BaseMMAFrag< T, 8, 8 >::row_frag_type</a></div><div class="ttdeci">metal::vec< T, kElemRows > row_frag_type</div><div class="ttdef"><b>Definition</b> mma.h:62</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_a51d662e4cff88b5ad17d7c44bb6b6970"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#a51d662e4cff88b5ad17d7c44bb6b6970">mlx::steel::BaseMMAFrag< T, 8, 8 >::row_reduce</a></div><div class="ttdeci">static METAL_FUNC constexpr void row_reduce(thread const frag_type &inp_vals, thread T *reduced_vals)</div><div class="ttdef"><b>Definition</b> mma.h:189</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:65</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:161</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:118</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:75</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:93</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_base_m_m_a_frag_html_adbb262a3c872e26533b68a39db16459e"><div class="ttname"><a href="structmlx_1_1steel_1_1_base_m_m_a_frag.html#adbb262a3c872e26533b68a39db16459e">mlx::steel::BaseMMAFrag< T, 8, 8 >::col_frag_type</a></div><div class="ttdeci">metal::vec< T, kElemCols > col_frag_type</div><div class="ttdef"><b>Definition</b> mma.h:63</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:180</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:138</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_a211102315e2afbcfcd2e2c201b638e9f"><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#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_3_01_t_00_018_00_018_01_4_html_a25675ae18947a97c6e04157b540103a9"><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#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_3_01_t_00_018_00_018_01_4_html_a2fe53db449c692226f23f6b99fb2c0d4"><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#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_3_01_t_00_018_00_018_01_4_html_a318c4279bdc7b39b7919f108b1cd8010"><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#a318c4279bdc7b39b7919f108b1cd8010">mlx::steel::BaseMMAFrag< T, 8, 8 >::row_bin_op</a></div><div class="ttdeci">static METAL_FUNC constexpr void row_bin_op(thread frag_type &inp_vals, thread T *row_vals)</div><div class="ttdef"><b>Definition</b> mma.h:204</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_a3c34dfdc944db110f4735f1b25307cf0"><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#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_3_01_t_00_018_00_018_01_4_html_a3dcd4301390937f89ed1dde6d28e341f"><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#a3dcd4301390937f89ed1dde6d28e341f">mlx::steel::BaseMMAFrag< T, 8, 8 >::row_frag_type</a></div><div class="ttdeci">metal::vec< T, kElemRows > row_frag_type</div><div class="ttdef"><b>Definition</b> mma.h:62</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_a51d662e4cff88b5ad17d7c44bb6b6970"><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#a51d662e4cff88b5ad17d7c44bb6b6970">mlx::steel::BaseMMAFrag< T, 8, 8 >::row_reduce</a></div><div class="ttdeci">static METAL_FUNC constexpr void row_reduce(thread const frag_type &inp_vals, thread T *reduced_vals)</div><div class="ttdef"><b>Definition</b> mma.h:189</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:65</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_a76aa5aa690dbcc954e957d767fad661f"><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#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_3_01_t_00_018_00_018_01_4_html_a7c212200d86b4e93f274d99addf668bd"><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#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_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:161</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_a8536bfaa108031c2ea3e9ccdc766ee5b"><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#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_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:118</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:75</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:93</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_adbb262a3c872e26533b68a39db16459e"><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#adbb262a3c872e26533b68a39db16459e">mlx::steel::BaseMMAFrag< T, 8, 8 >::col_frag_type</a></div><div class="ttdeci">metal::vec< T, kElemCols > col_frag_type</div><div class="ttdef"><b>Definition</b> mma.h:63</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:37</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:536</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:550</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>
|
||||
@@ -961,7 +969,7 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_layout2_d_html"><div class="ttname"><a href="structmlx_1_1steel_1_1_layout2_d.html">mlx::steel::Layout2D</a></div><div class="ttdef"><b>Definition</b> mma.h:31</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_layout2_d_html_a23183747ab1ddbdd3f1fcac6d0faa2cd"><div class="ttname"><a href="structmlx_1_1steel_1_1_layout2_d.html#a23183747ab1ddbdd3f1fcac6d0faa2cd">mlx::steel::Layout2D::shape</a></div><div class="ttdeci">Shape shape</div><div class="ttdef"><b>Definition</b> mma.h:32</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_layout2_d_html_a6beedf1677ee1b192fb48c83a29ac8a1"><div class="ttname"><a href="structmlx_1_1steel_1_1_layout2_d.html#a6beedf1677ee1b192fb48c83a29ac8a1">mlx::steel::Layout2D::layout</a></div><div class="ttdeci">Layout layout</div><div class="ttdef"><b>Definition</b> mma.h:33</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>
|
||||
<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:223</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a1a6b1446e8c8da46885bbaa8e8fdc7e4"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1a6b1446e8c8da46885bbaa8e8fdc7e4">mlx::steel::MMATile::frag_at</a></div><div class="ttdeci">METAL_FUNC constexpr thread frag_type & frag_at(const short i, const short j)</div><div class="ttdef"><b>Definition</b> mma.h:256</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a1d126b14910385ab644e224ac1d0307a"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1d126b14910385ab644e224ac1d0307a">mlx::steel::MMATile::kTileRows</a></div><div class="ttdeci">STEEL_CONST int kTileRows</div><div class="ttdef"><b>Definition</b> mma.h:230</div></div>
|
||||
<div class="ttc" id="astructmlx_1_1steel_1_1_m_m_a_tile_html_a1ea49efd92696b15302ee4b52ecd548c"><div class="ttname"><a href="structmlx_1_1steel_1_1_m_m_a_tile.html#a1ea49efd92696b15302ee4b52ecd548c">mlx::steel::MMATile::kColsPerThread</a></div><div class="ttdeci">STEEL_CONST int kColsPerThread</div><div class="ttdef"><b>Definition</b> mma.h:240</div></div>
|
||||
@@ -1002,7 +1010,7 @@ $(function(){initNavTree('attn_2mma_8h_source.html',''); initResizable(true); })
|
||||
<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_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>
|
||||
<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> 1.13.2 </li>
|
||||
</ul>
|
||||
</div>
|
||||
</body>
|
||||
|
Reference in New Issue
Block a user