Update docs

Author: github-actions[bot]

_sources/autoapi/tilelang/autotuner/param/index.rst.txt

@@ -73,7 +73,7 @@ Module Contents
 
    .. attribute:: execution_backend
 
-      Execution backend to use for kernel execution (default: "cython").
+      Execution backend to use for kernel execution (default: "auto").
 
    .. attribute:: target
 
@@ -103,8 +103,8 @@ Module Contents
 
 
    .. py:attribute:: execution_backend
-      :type:  Literal['dlpack', 'ctypes', 'cython']
-      :value: 'cython'
+      :type:  Literal['auto', 'tvm_ffi', 'ctypes', 'cython', 'nvrtc', 'torch']
+      :value: 'auto'
 
 
 

_sources/autoapi/tilelang/autotuner/tuner/index.rst.txt

@@ -132,7 +132,7 @@ Module Contents
 
 
 
-   .. py:method:: set_compile_args(out_idx = None, target = 'auto', execution_backend = 'cython', target_host = None, verbose = False, pass_configs = None)
+   .. py:method:: set_compile_args(out_idx = None, target = 'auto', execution_backend = 'auto', target_host = None, verbose = False, pass_configs = None)
 
       Set compilation arguments for the auto-tuner.
 
@@ -349,8 +349,9 @@ Module Contents
    :type target: Union[str, Target], optional
    :param target_host: Target host for cross-compilation. Defaults to None.
    :type target_host: Union[str, Target], optional
-   :param execution_backend: Backend for kernel execution and argument passing. Defaults to "cython".
-   :type execution_backend: Literal["dlpack", "ctypes", "cython"], optional
+   :param execution_backend: Backend for kernel execution and argument passing. Use "auto" to pick a sensible
+                             default per target (cuda->tvm_ffi, metal->torch, others->cython).
+   :type execution_backend: Literal["auto", "tvm_ffi", "ctypes", "cython", "nvrtc", "torch"], optional
    :param verbose: Enables verbose logging during compilation. Defaults to False.
    :type verbose: bool, optional
    :param pass_configs: Configurations for TVM's pass context. Defaults to None.

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

@@ -30,7 +30,7 @@ Functions
 Package Contents
 ----------------
 
-.. py:function:: cached(func = None, out_idx = None, *args, target = 'auto', target_host = None, execution_backend = 'cython', verbose = False, pass_configs = None, compile_flags = None)
+.. py:function:: cached(func = None, out_idx = None, *args, target = 'auto', target_host = None, execution_backend = 'auto', verbose = False, pass_configs = None, compile_flags = None)
 
    Caches and reuses compiled kernels (using KernelCache class).
 

_sources/autoapi/tilelang/cache/kernel_cache/index.rst.txt

@@ -14,8 +14,9 @@ Attributes
 
 .. autoapisummary::
 
-   tilelang.cache.kernel_cache.KERNEL_PATH
-   tilelang.cache.kernel_cache.WRAPPED_KERNEL_PATH
+   tilelang.cache.kernel_cache.DEVICE_KERNEL_PATH
+   tilelang.cache.kernel_cache.HOST_KERNEL_PATH
+   tilelang.cache.kernel_cache.EXECUTABLE_PATH
    tilelang.cache.kernel_cache.KERNEL_LIB_PATH
    tilelang.cache.kernel_cache.KERNEL_CUBIN_PATH
    tilelang.cache.kernel_cache.KERNEL_PY_PATH
@@ -33,12 +34,16 @@ Classes
 Module Contents
 ---------------
 
-.. py:data:: KERNEL_PATH
-   :value: 'kernel.cu'
+.. py:data:: DEVICE_KERNEL_PATH
+   :value: 'device_kernel.cu'
 
 
-.. py:data:: WRAPPED_KERNEL_PATH
-   :value: 'wrapped_kernel.cu'
+.. py:data:: HOST_KERNEL_PATH
+   :value: 'host_kernel.cu'
+
+
+.. py:data:: EXECUTABLE_PATH
+   :value: 'executable.so'
 
 
 .. py:data:: KERNEL_LIB_PATH
@@ -68,12 +73,12 @@ Module Contents
 
 
    .. py:attribute:: execution_backend
-      :type:  Literal['dlpack', 'ctypes', 'cython', 'nvrtc']
-      :value: 'cython'
+      :type:  Literal['tvm_ffi', 'ctypes', 'cython', 'nvrtc', 'torch']
+      :value: 'tvm_ffi'
 
 
 
-   .. py:method:: cached(func = None, out_idx = None, *args, target = 'auto', target_host = None, execution_backend = 'cython', verbose = False, pass_configs = None, compile_flags = None)
+   .. py:method:: cached(func = None, out_idx = None, *args, target = 'auto', target_host = None, execution_backend = 'auto', verbose = False, pass_configs = None, compile_flags = None)
 
       Caches and reuses compiled kernels to avoid redundant compilation.
 

_sources/autoapi/tilelang/contrib/dlpack/index.rst.txt

@@ -15,7 +15,6 @@ Functions
 .. autoapisummary::
 
    tilelang.contrib.dlpack.convert_func
-   tilelang.contrib.dlpack.to_pytorch_func
 
 
 Module Contents
@@ -34,14 +33,3 @@ Module Contents
    :type to_dlpack_func: Function
 
 
-.. py:function:: to_pytorch_func(tvm_func)
-
-   Convert a tvm function into one that accepts PyTorch tensors
-
-   :param tvm_func: Built tvm function operating on arrays
-   :type tvm_func: Function
-
-   :returns: **wrapped_func** -- Wrapped tvm function that operates on PyTorch tensors
-   :rtype: Function
-
-

_sources/autoapi/tilelang/jit/adapter/base/index.rst.txt

@@ -44,9 +44,34 @@ Module Contents
    .. py:attribute:: result_idx
 
 
+   .. py:method:: get_current_stream_functor()
+      :staticmethod:
+
+
+      Return a callable that reads Torch's current CUDA stream pointer.
+
+      The returned lambda yields the raw CUDA stream handle of the current
+      PyTorch stream on the active device. It's a thunk (evaluated at call
+      time) so that any upstream stream guards are respected. If CUDA is
+      unavailable, it returns a lambda that yields 0.
+
+
+
+   .. py:method:: get_current_device_functor()
+      :staticmethod:
+
+
+      Return a callable that yields Torch's current device.
+
+      Similar to the stream functor, we capture a callable that, when called,
+      fetches the current device according to PyTorch. On CPU or when CUDA is
+      unavailable, returns ``torch.device('cpu')``.
+
+
+
    .. py:method:: __call__(*args, **kwds)
 
 
-   .. py:method:: get_kernel_source()
+   .. py:method:: get_kernel_source(kernel_only = True)
 
 

_sources/autoapi/tilelang/jit/adapter/ctypes/adapter/index.rst.txt

@@ -20,7 +20,7 @@ Classes
 Module Contents
 ---------------
 
-.. py:class:: CtypesKernelAdapter(params, result_idx, target, func_or_mod, host_mod = None, device_mod = None, kernel_global_source = None, verbose = False, pass_configs = None, compile_flags = None)
+.. py:class:: CtypesKernelAdapter(params, result_idx, target, func_or_mod, host_mod = None, device_mod = None, host_kernel_source = None, device_kernel_source = None, verbose = False, pass_configs = None, compile_flags = None)
 
    Bases: :py:obj:`tilelang.jit.adapter.base.BaseKernelAdapter`
 
@@ -44,20 +44,20 @@ Module Contents
 
 
 
-   .. py:attribute:: kernel_global_source
+   .. py:attribute:: host_kernel_source
       :type:  str | None
       :value: None
 
 
 
-   .. py:attribute:: lib
-      :type:  ctypes.CDLL | None
+   .. py:attribute:: device_kernel_source
+      :type:  str | None
       :value: None
 
 
 
-   .. py:attribute:: wrapped_source
-      :type:  str | None
+   .. py:attribute:: lib
+      :type:  ctypes.CDLL | None
       :value: None
 
 
@@ -103,7 +103,10 @@ Module Contents
    .. py:attribute:: lib_generator
 
 
-   .. py:method:: from_database(params, result_idx, target, func_or_mod, kernel_global_source, kernel_lib_path, verbose = False, pass_configs = None, compile_flags = None)
+   .. py:attribute:: wrapped_source
+
+
+   .. py:method:: from_database(params, result_idx, target, func_or_mod, host_kernel_source, device_kernel_source, kernel_lib_path, verbose = False, pass_configs = None, compile_flags = None)
       :classmethod:
 
 

_sources/autoapi/tilelang/jit/adapter/cython/adapter/index.rst.txt

@@ -44,7 +44,7 @@ Module Contents
    A symbolic expression can be a simple tvm.Var, or an tvm.PrimExpr containing tvm.Var.
 
 
-.. py:class:: CythonKernelAdapter(params, result_idx, target, func_or_mod, host_mod = None, device_mod = None, kernel_global_source = None, verbose = False, pass_configs = None, compile_flags = None)
+.. py:class:: CythonKernelAdapter(params, result_idx, target, func_or_mod, host_mod = None, device_mod = None, device_kernel_source = None, verbose = False, pass_configs = None, compile_flags = None)
 
    Bases: :py:obj:`tilelang.jit.adapter.base.BaseKernelAdapter`
 
@@ -69,20 +69,20 @@ Module Contents
 
 
 
-   .. py:attribute:: kernel_global_source
+   .. py:attribute:: host_kernel_source
       :type:  str | None
       :value: None
 
 
 
-   .. py:attribute:: lib
-      :type:  ctypes.CDLL | None
+   .. py:attribute:: device_kernel_source
+      :type:  str | None
       :value: None
 
 
 
-   .. py:attribute:: wrapped_source
-      :type:  str | None
+   .. py:attribute:: lib
+      :type:  ctypes.CDLL | None
       :value: None
 
 
@@ -155,7 +155,7 @@ Module Contents
    .. py:attribute:: cython_wrapper
 
 
-   .. py:method:: from_database(params, result_idx, target, func_or_mod, kernel_global_source, kernel_lib_path, verbose = False, pass_configs = None, compile_flags = None)
+   .. py:method:: from_database(params, result_idx, target, func_or_mod, host_kernel_source, device_kernel_source, kernel_lib_path, verbose = False, pass_configs = None, compile_flags = None)
       :classmethod:
 
 

_sources/autoapi/tilelang/jit/adapter/dlpack/index.rst.txt

@@ -13,10 +13,10 @@ Submodules
    /autoapi/tilelang/jit/adapter/base/index
    /autoapi/tilelang/jit/adapter/ctypes/index
    /autoapi/tilelang/jit/adapter/cython/index
-   /autoapi/tilelang/jit/adapter/dlpack/index
    /autoapi/tilelang/jit/adapter/libgen/index
    /autoapi/tilelang/jit/adapter/nvrtc/index
    /autoapi/tilelang/jit/adapter/torch/index
+   /autoapi/tilelang/jit/adapter/tvm_ffi/index
    /autoapi/tilelang/jit/adapter/utils/index
    /autoapi/tilelang/jit/adapter/wrapper/index