Update docs

Author: github-actions[bot]

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

@@ -14,10 +14,9 @@ Attributes
    tilelang.env.COMPOSABLE_KERNEL_NOT_FOUND_MESSAGE
    tilelang.env.TL_TEMPLATE_NOT_FOUND_MESSAGE
    tilelang.env.TVM_LIBRARY_NOT_FOUND_MESSAGE
-   tilelang.env.SITE_PACKAGES
+   tilelang.env.TL_ROOT
    tilelang.env.TL_LIBS
    tilelang.env.TL_LIBS
-   tilelang.env.TL_ROOT
    tilelang.env.DEV
    tilelang.env.THIRD_PARTY_ROOT
    tilelang.env.DEV
@@ -81,16 +80,12 @@ Module Contents
    :value: 'TVM is not installed or found in the expected path'
 
 
-.. py:data:: SITE_PACKAGES
-   :value: []
-
+.. py:data:: TL_ROOT
 
 .. py:data:: TL_LIBS
 
 .. py:data:: TL_LIBS
 
-.. py:data:: TL_ROOT
-
 .. py:data:: DEV
    :value: False
 

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

@@ -34,6 +34,11 @@ Functions
    tilelang.language.builtin.warpgroup_arrive
    tilelang.language.builtin.warpgroup_commit_batch
    tilelang.language.builtin.warpgroup_wait
+   tilelang.language.builtin.get_lane_idx
+   tilelang.language.builtin.get_warp_idx_sync
+   tilelang.language.builtin.get_warp_idx
+   tilelang.language.builtin.get_warp_group_idx
+   tilelang.language.builtin.shuffle_elect
    tilelang.language.builtin.wait_wgmma
    tilelang.language.builtin.barrier_wait
    tilelang.language.builtin.barrier_arrive
@@ -270,6 +275,103 @@ Module Contents
    :rtype: tir.Call
 
 
+.. py:function:: get_lane_idx(warp_size = None)
+
+   Return the logical lane index of the calling thread within a warp.
+
+   :param warp_size: Logical warp (or wavefront) size. Defaults to 32 on NVIDIA and 64 on AMD.
+   :type warp_size: Optional[int, PrimExpr]
+
+   .. rubric:: Example
+
+   >>> lane = T.get_lane_idx()
+   >>> custom_lane = T.get_lane_idx(64)  # override warp size explicitly
+
+   Implementation Notes
+   --------------------
+   Lowers to the CUDA helper `tl::get_lane_idx(warp_size)` defined in
+   `src/tl_templates/cuda/intrin.h`, which computes the lane index from the
+   linear thread id using the provided `warp_size`.
+
+
+.. py:function:: get_warp_idx_sync(warp_size = None)
+
+   Return the canonical warp index, assuming the warp's threads are converged.
+
+   :param warp_size: Logical warp size used for the index calculation.
+   :type warp_size: Optional[int, PrimExpr]
+
+   .. rubric:: Example
+
+   >>> warp = T.get_warp_idx_sync()
+   >>> custom_warp = T.get_warp_idx_sync(64)
+
+   Implementation Notes
+   --------------------
+   Emits `tl::get_warp_idx_sync(warp_size)` which divides the block-linear
+   thread id by `warp_size`, matching the semantics of CUTLASS' canonical helpers.
+
+
+.. py:function:: get_warp_idx(warp_size = None)
+
+   Return the canonical warp index without synchronizing the warp.
+
+   :param warp_size: Logical warp size used for the index calculation.
+   :type warp_size: Optional[int, PrimExpr]
+
+   .. rubric:: Example
+
+   >>> warp = T.get_warp_idx()
+   >>> custom_warp = T.get_warp_idx(64)
+
+   Implementation Notes
+   --------------------
+   Lowers to `tl::get_warp_idx(warp_size)` which divides the block-linear
+   thread id by the provided `warp_size` without requiring warp convergence.
+
+
+.. py:function:: get_warp_group_idx(warp_size = None, warps_per_group = None)
+
+   Return the canonical warp group index for the calling thread.
+
+   :param warp_size: Logical warp size to use (defaults to 32 on NVIDIA / 64 on AMD).
+   :type warp_size: Optional[int, PrimExpr]
+   :param warps_per_group: Number of warps per warp-group. Defaults to 4 on NVIDIA architectures.
+   :type warps_per_group: Optional[int, PrimExpr]
+
+   .. rubric:: Example
+
+   >>> group = T.get_warp_group_idx()
+   >>> custom_group = T.get_warp_group_idx(32, 6)  # treat 6 warps as a group
+
+   Implementation Notes
+   --------------------
+   Generates `tl::get_warp_group_idx(warp_size, warps_per_group)` which
+   divides the block-linear thread id by `warp_size * warps_per_group`,
+   matching the canonical ordering while allowing architecture-specific overrides.
+
+
+.. py:function:: shuffle_elect(thread_extent)
+
+   Elect exactly one lane within a logical thread group.
+
+   :param thread_extent: Size (in threads) of the group in which a single lane should be elected.
+                         Passing 0 elects a single lane in the entire thread block.
+   :type thread_extent: int
+
+   .. rubric:: Example
+
+   >>> is_leader = T.shuffle_elect(64)
+   >>> T.if_then_else(is_leader, do_leader_work(), T.evaluate(0))
+
+   Implementation Notes
+   --------------------
+   Lowered to the CUDA helper `tl::tl_shuffle_elect<thread_extent>()` defined in
+   `src/tl_templates/cuda/intrin.h`, which relies on
+   `cutlass::canonical_warp_idx_sync()` and `cute::elect_one_sync()` (or
+   `__shfl_sync`) to pick one lane per group.
+
+
 .. py:function:: wait_wgmma(id)
 
    Wait for WGMMA (Warp Group Matrix Multiply-Accumulate) operations to complete.

autoapi/tilelang/env/index.html

@@ -498,7 +498,7 @@ <h2>Attributes<a class="headerlink" href="#attributes" title="Link to this headi
 <tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.TVM_LIBRARY_NOT_FOUND_MESSAGE" title="tilelang.env.TVM_LIBRARY_NOT_FOUND_MESSAGE"><code class="xref py py-obj docutils literal notranslate"><span class="pre">TVM_LIBRARY_NOT_FOUND_MESSAGE</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.SITE_PACKAGES" title="tilelang.env.SITE_PACKAGES"><code class="xref py py-obj docutils literal notranslate"><span class="pre">SITE_PACKAGES</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.TL_ROOT" title="tilelang.env.TL_ROOT"><code class="xref py py-obj docutils literal notranslate"><span class="pre">TL_ROOT</span></code></a></p></td>
 <td><p></p></td>
 </tr>
 <tr class="row-odd"><td><p><a class="reference internal" href="#id0" title="tilelang.env.TL_LIBS"><code class="xref py py-obj docutils literal notranslate"><span class="pre">TL_LIBS</span></code></a></p></td>
@@ -507,55 +507,52 @@ <h2>Attributes<a class="headerlink" href="#attributes" title="Link to this headi
 <tr class="row-even"><td><p><a class="reference internal" href="#id0" title="tilelang.env.TL_LIBS"><code class="xref py py-obj docutils literal notranslate"><span class="pre">TL_LIBS</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.TL_ROOT" title="tilelang.env.TL_ROOT"><code class="xref py py-obj docutils literal notranslate"><span class="pre">TL_ROOT</span></code></a></p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#id1" title="tilelang.env.DEV"><code class="xref py py-obj docutils literal notranslate"><span class="pre">DEV</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#id1" title="tilelang.env.DEV"><code class="xref py py-obj docutils literal notranslate"><span class="pre">DEV</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.THIRD_PARTY_ROOT" title="tilelang.env.THIRD_PARTY_ROOT"><code class="xref py py-obj docutils literal notranslate"><span class="pre">THIRD_PARTY_ROOT</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.THIRD_PARTY_ROOT" title="tilelang.env.THIRD_PARTY_ROOT"><code class="xref py py-obj docutils literal notranslate"><span class="pre">THIRD_PARTY_ROOT</span></code></a></p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#id1" title="tilelang.env.DEV"><code class="xref py py-obj docutils literal notranslate"><span class="pre">DEV</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#id1" title="tilelang.env.DEV"><code class="xref py py-obj docutils literal notranslate"><span class="pre">DEV</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.enable_cache" title="tilelang.env.enable_cache"><code class="xref py py-obj docutils literal notranslate"><span class="pre">enable_cache</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.enable_cache" title="tilelang.env.enable_cache"><code class="xref py py-obj docutils literal notranslate"><span class="pre">enable_cache</span></code></a></p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.disable_cache" title="tilelang.env.disable_cache"><code class="xref py py-obj docutils literal notranslate"><span class="pre">disable_cache</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.disable_cache" title="tilelang.env.disable_cache"><code class="xref py py-obj docutils literal notranslate"><span class="pre">disable_cache</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.is_cache_enabled" title="tilelang.env.is_cache_enabled"><code class="xref py py-obj docutils literal notranslate"><span class="pre">is_cache_enabled</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.is_cache_enabled" title="tilelang.env.is_cache_enabled"><code class="xref py py-obj docutils literal notranslate"><span class="pre">is_cache_enabled</span></code></a></p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.env" title="tilelang.env.env"><code class="xref py py-obj docutils literal notranslate"><span class="pre">env</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.env" title="tilelang.env.env"><code class="xref py py-obj docutils literal notranslate"><span class="pre">env</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.CUDA_HOME" title="tilelang.env.CUDA_HOME"><code class="xref py py-obj docutils literal notranslate"><span class="pre">CUDA_HOME</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.CUDA_HOME" title="tilelang.env.CUDA_HOME"><code class="xref py py-obj docutils literal notranslate"><span class="pre">CUDA_HOME</span></code></a></p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.ROCM_HOME" title="tilelang.env.ROCM_HOME"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ROCM_HOME</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.ROCM_HOME" title="tilelang.env.ROCM_HOME"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ROCM_HOME</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.tvm_path" title="tilelang.env.tvm_path"><code class="xref py py-obj docutils literal notranslate"><span class="pre">tvm_path</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.tvm_path" title="tilelang.env.tvm_path"><code class="xref py py-obj docutils literal notranslate"><span class="pre">tvm_path</span></code></a></p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.cutlass_inc_path" title="tilelang.env.cutlass_inc_path"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cutlass_inc_path</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.cutlass_inc_path" title="tilelang.env.cutlass_inc_path"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cutlass_inc_path</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.ck_inc_path" title="tilelang.env.ck_inc_path"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ck_inc_path</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.ck_inc_path" title="tilelang.env.ck_inc_path"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ck_inc_path</span></code></a></p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.tl_template_path" title="tilelang.env.tl_template_path"><code class="xref py py-obj docutils literal notranslate"><span class="pre">tl_template_path</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.tl_template_path" title="tilelang.env.tl_template_path"><code class="xref py py-obj docutils literal notranslate"><span class="pre">tl_template_path</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.CUTLASS_INCLUDE_DIR" title="tilelang.env.CUTLASS_INCLUDE_DIR"><code class="xref py py-obj docutils literal notranslate"><span class="pre">CUTLASS_INCLUDE_DIR</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.CUTLASS_INCLUDE_DIR" title="tilelang.env.CUTLASS_INCLUDE_DIR"><code class="xref py py-obj docutils literal notranslate"><span class="pre">CUTLASS_INCLUDE_DIR</span></code></a></p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.COMPOSABLE_KERNEL_INCLUDE_DIR" title="tilelang.env.COMPOSABLE_KERNEL_INCLUDE_DIR"><code class="xref py py-obj docutils literal notranslate"><span class="pre">COMPOSABLE_KERNEL_INCLUDE_DIR</span></code></a></p></td>
 <td><p></p></td>
 </tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.COMPOSABLE_KERNEL_INCLUDE_DIR" title="tilelang.env.COMPOSABLE_KERNEL_INCLUDE_DIR"><code class="xref py py-obj docutils literal notranslate"><span class="pre">COMPOSABLE_KERNEL_INCLUDE_DIR</span></code></a></p></td>
-<td><p></p></td>
-</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tilelang.env.TILELANG_TEMPLATE_PATH" title="tilelang.env.TILELANG_TEMPLATE_PATH"><code class="xref py py-obj docutils literal notranslate"><span class="pre">TILELANG_TEMPLATE_PATH</span></code></a></p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tilelang.env.TILELANG_TEMPLATE_PATH" title="tilelang.env.TILELANG_TEMPLATE_PATH"><code class="xref py py-obj docutils literal notranslate"><span class="pre">TILELANG_TEMPLATE_PATH</span></code></a></p></td>
 <td><p></p></td>
 </tr>
 </tbody>
@@ -623,8 +620,8 @@ <h2>Module Contents<a class="headerlink" href="#module-contents" title="Link to
 <dd></dd></dl>
 
 <dl class="py data">
-<dt class="sig sig-object py" id="tilelang.env.SITE_PACKAGES">
-<span class="sig-prename descclassname"><span class="pre">tilelang.env.</span></span><span class="sig-name descname"><span class="pre">SITE_PACKAGES</span></span><em class="property"><span class="w"> </span><span class="p"><span class="pre">=</span></span><span class="w"> </span><span class="pre">[]</span></em><a class="headerlink" href="#tilelang.env.SITE_PACKAGES" title="Link to this definition">¶</a></dt>
+<dt class="sig sig-object py" id="tilelang.env.TL_ROOT">
+<span class="sig-prename descclassname"><span class="pre">tilelang.env.</span></span><span class="sig-name descname"><span class="pre">TL_ROOT</span></span><a class="headerlink" href="#tilelang.env.TL_ROOT" title="Link to this definition">¶</a></dt>
 <dd></dd></dl>
 
 <dl class="py data">
@@ -637,11 +634,6 @@ <h2>Module Contents<a class="headerlink" href="#module-contents" title="Link to
 <span class="sig-prename descclassname"><span class="pre">tilelang.env.</span></span><span class="sig-name descname"><span class="pre">TL_LIBS</span></span><a class="headerlink" href="#id0" title="Link to this definition">¶</a></dt>
 <dd></dd></dl>
 
-<dl class="py data">
-<dt class="sig sig-object py" id="tilelang.env.TL_ROOT">
-<span class="sig-prename descclassname"><span class="pre">tilelang.env.</span></span><span class="sig-name descname"><span class="pre">TL_ROOT</span></span><a class="headerlink" href="#tilelang.env.TL_ROOT" title="Link to this definition">¶</a></dt>
-<dd></dd></dl>
-
 <dl class="py data">
 <dt class="sig sig-object py" id="tilelang.env.DEV">
 <span class="sig-prename descclassname"><span class="pre">tilelang.env.</span></span><span class="sig-name descname"><span class="pre">DEV</span></span><em class="property"><span class="w"> </span><span class="p"><span class="pre">=</span></span><span class="w"> </span><span class="pre">False</span></em><a class="headerlink" href="#tilelang.env.DEV" title="Link to this definition">¶</a></dt>
@@ -1052,10 +1044,9 @@ <h3>Benefits<a class="headerlink" href="#benefits" title="Link to this heading">
 <li><a class="reference internal" href="#tilelang.env.COMPOSABLE_KERNEL_NOT_FOUND_MESSAGE"><code class="docutils literal notranslate"><span class="pre">COMPOSABLE_KERNEL_NOT_FOUND_MESSAGE</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.env.TL_TEMPLATE_NOT_FOUND_MESSAGE"><code class="docutils literal notranslate"><span class="pre">TL_TEMPLATE_NOT_FOUND_MESSAGE</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.env.TVM_LIBRARY_NOT_FOUND_MESSAGE"><code class="docutils literal notranslate"><span class="pre">TVM_LIBRARY_NOT_FOUND_MESSAGE</span></code></a></li>
-<li><a class="reference internal" href="#tilelang.env.SITE_PACKAGES"><code class="docutils literal notranslate"><span class="pre">SITE_PACKAGES</span></code></a></li>
+<li><a class="reference internal" href="#tilelang.env.TL_ROOT"><code class="docutils literal notranslate"><span class="pre">TL_ROOT</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.env.TL_LIBS"><code class="docutils literal notranslate"><span class="pre">TL_LIBS</span></code></a></li>
 <li><a class="reference internal" href="#id0"><code class="docutils literal notranslate"><span class="pre">TL_LIBS</span></code></a></li>
-<li><a class="reference internal" href="#tilelang.env.TL_ROOT"><code class="docutils literal notranslate"><span class="pre">TL_ROOT</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.env.DEV"><code class="docutils literal notranslate"><span class="pre">DEV</span></code></a></li>
 <li><a class="reference internal" href="#tilelang.env.THIRD_PARTY_ROOT"><code class="docutils literal notranslate"><span class="pre">THIRD_PARTY_ROOT</span></code></a></li>
 <li><a class="reference internal" href="#id1"><code class="docutils literal notranslate"><span class="pre">DEV</span></code></a></li>