Update docs

Author: github-actions[bot]

_sources/autoapi/tilelang/language/kernel/index.rst.txt

@@ -216,6 +216,31 @@ Module Contents
    :returns: **res** -- The result LaunchThreadFrame.
    :rtype: Tuple[frame.LaunchThreadFrame]
 
+   .. rubric:: Examples
+
+   Create a 1-D CUDA kernel launch and unpack the single block index:
+
+   .. code-block:: python
+
+       with T.Kernel(T.ceildiv(N, 128), threads=128) as bx:
+           # bx is the blockIdx.x binding (also iterable as (bx,))
+           ...
+
+   Launch a 2-D grid while requesting two thread dimensions:
+
+   .. code-block:: python
+
+       with T.Kernel(grid_x, grid_y, threads=(64, 2)) as (bx, by):
+           tx, ty = T.get_thread_bindings()
+           ...
+
+   Emit a CPU kernel where thread bindings are skipped:
+
+   .. code-block:: python
+
+       with T.Kernel(loop_extent, is_cpu=True) as (i,):
+           ...
+
 
 .. py:function:: get_thread_binding(dim = 0)
 

_sources/autoapi/tilelang/transform/index.rst.txt

@@ -49,6 +49,7 @@ Functions
    tilelang.transform.LegalizeSafeMemoryAccess
    tilelang.transform.MakePackedAPI
    tilelang.transform.AnnotateDeviceRegions
+   tilelang.transform.SplitHostDevice
    tilelang.transform.VectorizeLoop
    tilelang.transform.InjectPTXAsyncCopy
    tilelang.transform.LowerDeviceStorageAccessInfo
@@ -273,6 +274,14 @@ Package Contents
    :rtype: tvm.transform.Pass
 
 
+.. py:function:: SplitHostDevice()
+
+   Split host/device functions even for empty kernels.
+
+   :returns: **fpass** -- The result pass
+   :rtype: tvm.transform.Pass
+
+
 .. py:function:: VectorizeLoop(enable_vectorize = True)
 
    VectorizeLoop

autoapi/tilelang/language/kernel/index.html

@@ -790,6 +790,24 @@ <h2>Module Contents<a class="headerlink" href="#module-contents" title="Link to
 <dd class="field-odd"><p>Tuple[frame.LaunchThreadFrame]</p>
 </dd>
 </dl>
+<p class="rubric">Examples</p>
+<p>Create a 1-D CUDA kernel launch and unpack the single block index:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">Kernel</span><span class="p">(</span><span class="n">T</span><span class="o">.</span><span class="n">ceildiv</span><span class="p">(</span><span class="n">N</span><span class="p">,</span> <span class="mi">128</span><span class="p">),</span> <span class="n">threads</span><span class="o">=</span><span class="mi">128</span><span class="p">)</span> <span class="k">as</span> <span class="n">bx</span><span class="p">:</span>
+    <span class="c1"># bx is the blockIdx.x binding (also iterable as (bx,))</span>
+    <span class="o">...</span>
+</pre></div>
+</div>
+<p>Launch a 2-D grid while requesting two thread dimensions:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">Kernel</span><span class="p">(</span><span class="n">grid_x</span><span class="p">,</span> <span class="n">grid_y</span><span class="p">,</span> <span class="n">threads</span><span class="o">=</span><span class="p">(</span><span class="mi">64</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span> <span class="k">as</span> <span class="p">(</span><span class="n">bx</span><span class="p">,</span> <span class="n">by</span><span class="p">):</span>
+    <span class="n">tx</span><span class="p">,</span> <span class="n">ty</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">get_thread_bindings</span><span class="p">()</span>
+    <span class="o">...</span>
+</pre></div>
+</div>
+<p>Emit a CPU kernel where thread bindings are skipped:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">Kernel</span><span class="p">(</span><span class="n">loop_extent</span><span class="p">,</span> <span class="n">is_cpu</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span> <span class="k">as</span> <span class="p">(</span><span class="n">i</span><span class="p">,):</span>
+    <span class="o">...</span>
+</pre></div>
+</div>
 </dd></dl>
 
 <dl class="py function">

autoapi/tilelang/transform/index.html

@@ -564,58 +564,61 @@ <h2>Functions<a class="headerlink" href="#functions" title="Link to this heading
 <tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.AnnotateDeviceRegions" title="tilelang.transform.AnnotateDeviceRegions"><code class="xref py py-obj docutils literal notranslate"><span class="pre">AnnotateDeviceRegions</span></code></a>()</p></td>
 <td><p>AnnotateDeviceRegions</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.VectorizeLoop" title="tilelang.transform.VectorizeLoop"><code class="xref py py-obj docutils literal notranslate"><span class="pre">VectorizeLoop</span></code></a>([enable_vectorize])</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.SplitHostDevice" title="tilelang.transform.SplitHostDevice"><code class="xref py py-obj docutils literal notranslate"><span class="pre">SplitHostDevice</span></code></a>()</p></td>
+<td><p>Split host/device functions even for empty kernels.</p></td>
+</tr>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.VectorizeLoop" title="tilelang.transform.VectorizeLoop"><code class="xref py py-obj docutils literal notranslate"><span class="pre">VectorizeLoop</span></code></a>([enable_vectorize])</p></td>
 <td><p>VectorizeLoop</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.InjectPTXAsyncCopy" title="tilelang.transform.InjectPTXAsyncCopy"><code class="xref py py-obj docutils literal notranslate"><span class="pre">InjectPTXAsyncCopy</span></code></a>()</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.InjectPTXAsyncCopy" title="tilelang.transform.InjectPTXAsyncCopy"><code class="xref py py-obj docutils literal notranslate"><span class="pre">InjectPTXAsyncCopy</span></code></a>()</p></td>
 <td><p>Rewrite global to shared memory copy on CUDA with asynchronous copy.</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LowerDeviceStorageAccessInfo" title="tilelang.transform.LowerDeviceStorageAccessInfo"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerDeviceStorageAccessInfo</span></code></a>()</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LowerDeviceStorageAccessInfo" title="tilelang.transform.LowerDeviceStorageAccessInfo"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerDeviceStorageAccessInfo</span></code></a>()</p></td>
 <td><p>Lower attached storage access information on device.</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LoopVectorizeDynamic" title="tilelang.transform.LoopVectorizeDynamic"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LoopVectorizeDynamic</span></code></a>()</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LoopVectorizeDynamic" title="tilelang.transform.LoopVectorizeDynamic"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LoopVectorizeDynamic</span></code></a>()</p></td>
 <td><p>Try to vectorize loop with dynamic shape.</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.ConfigIndexBitwidth" title="tilelang.transform.ConfigIndexBitwidth"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConfigIndexBitwidth</span></code></a>()</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.ConfigIndexBitwidth" title="tilelang.transform.ConfigIndexBitwidth"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConfigIndexBitwidth</span></code></a>()</p></td>
 <td><p>Config index bitwidth.</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.FlattenBuffer" title="tilelang.transform.FlattenBuffer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">FlattenBuffer</span></code></a>()</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.FlattenBuffer" title="tilelang.transform.FlattenBuffer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">FlattenBuffer</span></code></a>()</p></td>
 <td><p>FlattenBuffer</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.EliminateStorageSyncForMBarrier" title="tilelang.transform.EliminateStorageSyncForMBarrier"><code class="xref py py-obj docutils literal notranslate"><span class="pre">EliminateStorageSyncForMBarrier</span></code></a>()</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.EliminateStorageSyncForMBarrier" title="tilelang.transform.EliminateStorageSyncForMBarrier"><code class="xref py py-obj docutils literal notranslate"><span class="pre">EliminateStorageSyncForMBarrier</span></code></a>()</p></td>
 <td><p>EliminateStorageSyncForMBarrier</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.MergeSharedMemoryAllocations" title="tilelang.transform.MergeSharedMemoryAllocations"><code class="xref py py-obj docutils literal notranslate"><span class="pre">MergeSharedMemoryAllocations</span></code></a>([...])</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.MergeSharedMemoryAllocations" title="tilelang.transform.MergeSharedMemoryAllocations"><code class="xref py py-obj docutils literal notranslate"><span class="pre">MergeSharedMemoryAllocations</span></code></a>([...])</p></td>
 <td><p>MergeSharedMemoryAllocations</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LowerL2Persistent" title="tilelang.transform.LowerL2Persistent"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerL2Persistent</span></code></a>()</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LowerL2Persistent" title="tilelang.transform.LowerL2Persistent"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerL2Persistent</span></code></a>()</p></td>
 <td><p>LowerL2Persistent</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.PersistThreadblock" title="tilelang.transform.PersistThreadblock"><code class="xref py py-obj docutils literal notranslate"><span class="pre">PersistThreadblock</span></code></a>()</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.PersistThreadblock" title="tilelang.transform.PersistThreadblock"><code class="xref py py-obj docutils literal notranslate"><span class="pre">PersistThreadblock</span></code></a>()</p></td>
 <td><p>PersistThreadblock</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.AlignDynamicSharedMemoryAllocations" title="tilelang.transform.AlignDynamicSharedMemoryAllocations"><code class="xref py py-obj docutils literal notranslate"><span class="pre">AlignDynamicSharedMemoryAllocations</span></code></a>([align_bytes])</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.AlignDynamicSharedMemoryAllocations" title="tilelang.transform.AlignDynamicSharedMemoryAllocations"><code class="xref py py-obj docutils literal notranslate"><span class="pre">AlignDynamicSharedMemoryAllocations</span></code></a>([align_bytes])</p></td>
 <td><p>AlignDynamicSharedMemoryAllocations</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LowerSharedBarrier" title="tilelang.transform.LowerSharedBarrier"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerSharedBarrier</span></code></a>()</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LowerSharedBarrier" title="tilelang.transform.LowerSharedBarrier"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerSharedBarrier</span></code></a>()</p></td>
 <td><p>LowerSharedBarrier</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.StorageRewrite" title="tilelang.transform.StorageRewrite"><code class="xref py py-obj docutils literal notranslate"><span class="pre">StorageRewrite</span></code></a>()</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.StorageRewrite" title="tilelang.transform.StorageRewrite"><code class="xref py py-obj docutils literal notranslate"><span class="pre">StorageRewrite</span></code></a>()</p></td>
 <td><p>StorageRewrite</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LowerOpaqueBlock" title="tilelang.transform.LowerOpaqueBlock"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerOpaqueBlock</span></code></a>()</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LowerOpaqueBlock" title="tilelang.transform.LowerOpaqueBlock"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerOpaqueBlock</span></code></a>()</p></td>
 <td><p>LowerOpaqueBlock</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LowerThreadAllreduce" title="tilelang.transform.LowerThreadAllreduce"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerThreadAllreduce</span></code></a>()</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LowerThreadAllreduce" title="tilelang.transform.LowerThreadAllreduce"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerThreadAllreduce</span></code></a>()</p></td>
 <td><p>LowerThreadAllreduce</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LowerDeviceKernelLaunch" title="tilelang.transform.LowerDeviceKernelLaunch"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerDeviceKernelLaunch</span></code></a>()</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LowerDeviceKernelLaunch" title="tilelang.transform.LowerDeviceKernelLaunch"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerDeviceKernelLaunch</span></code></a>()</p></td>
 <td><p>Create and return a transform pass that lowers device kernel launch constructs to target-specific IR.</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LowerSharedTmem" title="tilelang.transform.LowerSharedTmem"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerSharedTmem</span></code></a>()</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LowerSharedTmem" title="tilelang.transform.LowerSharedTmem"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LowerSharedTmem</span></code></a>()</p></td>
 <td><p>LowerSharedTmem</p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.transform.LayoutReducer" title="tilelang.transform.LayoutReducer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LayoutReducer</span></code></a>()</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.transform.LayoutReducer" title="tilelang.transform.LayoutReducer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LayoutReducer</span></code></a>()</p></td>
 <td><p>Return a TVM transform pass that performs layout reduction/normalization.</p></td>
 </tr>
 </tbody>
@@ -960,6 +963,20 @@ <h3>Returns:<a class="headerlink" href="#returns" title="Link to this heading">
 </dl>
 </dd></dl>
 
+<dl class="py function">
+<dt class="sig sig-object py" id="tilelang.transform.SplitHostDevice">
+<span class="sig-prename descclassname"><span class="pre">tilelang.transform.</span></span><span class="sig-name descname"><span class="pre">SplitHostDevice</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="headerlink" href="#tilelang.transform.SplitHostDevice" title="Link to this definition">¶</a></dt>
+<dd><p>Split host/device functions even for empty kernels.</p>
+<dl class="field-list simple">
+<dt class="field-odd">Returns<span class="colon">:</span></dt>
+<dd class="field-odd"><p><strong>fpass</strong> – The result pass</p>
+</dd>
+<dt class="field-even">Return type<span class="colon">:</span></dt>
+<dd class="field-even"><p>tvm.transform.Pass</p>
+</dd>
+</dl>
+</dd></dl>
+
 <dl class="py function">
 <dt class="sig sig-object py" id="tilelang.transform.VectorizeLoop">
 <span class="sig-prename descclassname"><span class="pre">tilelang.transform.</span></span><span class="sig-name descname"><span class="pre">VectorizeLoop</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">enable_vectorize</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#tilelang.transform.VectorizeLoop" title="Link to this definition">¶</a></dt>
@@ -1254,6 +1271,7 @@ <h3>Returns:<a class="headerlink" href="#returns" title="Link to this heading">
 <li><a class="reference internal" href="#tilelang.transform.LegalizeSafeMemoryAccess"><code class="docutils literal notranslate"><span class="pre">LegalizeSafeMemoryAccess()</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.transform.MakePackedAPI"><code class="docutils literal notranslate"><span class="pre">MakePackedAPI()</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.transform.AnnotateDeviceRegions"><code class="docutils literal notranslate"><span class="pre">AnnotateDeviceRegions()</span></code></a></li>
+<li><a class="reference internal" href="#tilelang.transform.SplitHostDevice"><code class="docutils literal notranslate"><span class="pre">SplitHostDevice()</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.transform.VectorizeLoop"><code class="docutils literal notranslate"><span class="pre">VectorizeLoop()</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.transform.InjectPTXAsyncCopy"><code class="docutils literal notranslate"><span class="pre">InjectPTXAsyncCopy()</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.transform.LowerDeviceStorageAccessInfo"><code class="docutils literal notranslate"><span class="pre">LowerDeviceStorageAccessInfo()</span></code></a></li>

genindex.html

@@ -4587,6 +4587,8 @@ <h2>S</h2>
         <li><a href="autoapi/tilelang/jit/param/index.html#tilelang.jit.param.Kernel.source_code">source_code (tilelang.jit.param.Kernel attribute)</a>
 </li>
         <li><a href="autoapi/tilelang/carver/roller/hint/index.html#tilelang.carver.roller.hint.Hint.split_k_factor">split_k_factor (tilelang.carver.roller.hint.Hint attribute)</a>
+</li>
+        <li><a href="autoapi/tilelang/transform/index.html#tilelang.transform.SplitHostDevice">SplitHostDevice() (in module tilelang.transform)</a>
 </li>
         <li><a href="autoapi/tilelang/language/tir/ir/index.html#tilelang.language.tir.ir.sqrt">sqrt (in module tilelang.language.tir.ir)</a>
 </li>