MCPcopy Create free account

hub / github.com/tile-ai/tilelang / functions

Functions7,254 in github.com/tile-ai/tilelang

↓ 1,379 callersFunctionprint
A generic print function that handles both TIR buffers and primitive expressions. - If the input is a TIR buffer, it prints its values, but
tilelang/language/print_op.py:159
↓ 386 callersMethodget
(self)
tilelang/language/eager/builder.py:269
↓ 371 callersMethodload
(self)
tilelang/language/eager/builder.py:163
↓ 332 callersMethodcompile
(self, *args: _P.args, **kwargs: _P.kwargs)
tilelang/jit/__init__.py:389
↓ 282 callersMethodmain
(A: T.Buffer((16,), "float32"))
testing/python/transform/test_tilelang_transform_unroll_loop.py:11
↓ 217 callersMethodget_kernel_source
(self)
testing/python/autotune/test_tilelang_autotune_atomic_save.py:25
↓ 164 callersFunctionmin
Compute the minimum value of two expressions. Parameters ---------- a : PrimExpr The left hand operand b : PrimExpr
tilelang/language/ast/ir.py:1572
↓ 159 callersMethodstride
(self)
tilelang/carver/roller/hint.py:45
↓ 151 callersMethodsize
Returns the number of items in the stack.
tilelang/language/kernel.py:60
↓ 147 callersMethodget
(self)
tilelang/env.py:259
↓ 143 callersFunctiondo_bench
Benchmark the runtime of a PyTorch function with L2 cache management. This function provides accurate GPU kernel timing by: - Clearing L2 cac
tilelang/profiler/bench.py:66
↓ 138 callersMethodget_thread_binding
Returns the thread binding for the given dimension. dim=0 corresponds to threadIdx.x, dim=1 to threadIdx.y, and dim=2 to threadIdx.z.
tilelang/language/kernel.py:209
↓ 123 callersFunctionvar_dom
Get domains of iterators
testing/python/arith/test_arith_iter_affine_map.py:42
↓ 117 callersMethodget_profiler
Creates a profiler to benchmark the compiled runtime module. Parameters ---------- tensor_supply_type : TensorSupply
tilelang/jit/kernel.py:431
↓ 116 callersMethodas_torch
(self)
tilelang/language/dtypes.py:19
↓ 116 callersMethodstore
(self, value)
tilelang/language/eager/builder.py:160
↓ 112 callersFunctionmax
Compute the maximum value of two expressions. Parameters ---------- a : PrimExpr The left hand operand b : PrimExpr
tilelang/language/ast/ir.py:1591
↓ 104 callersMethoddo_bench
Benchmarks the execution time of a given function. Args: func: Function to benchmark (uses adapter if None) warmup: W
tilelang/profiler/__init__.py:220
↓ 102 callersFunction_op_wrapper
(func)
tilelang/language/tir/ir.py:162
↓ 101 callersFunction_op_wrapper
(func)
tilelang/language/ast/ir.py:1748
↓ 96 callersMethodreshape
Reshape the input shape of the layout. Parameters ---------- shape : list[PrimExpr] or list[int] The new
tilelang/layout/layout.py:224
↓ 90 callersFunctionkernel
( A: T.Tensor((n * 4,), T.int32), )
testing/python/language/test_tilelang_language_annot.py:15
↓ 87 callersMethodparse_args
Parse arguments and return cache key and tensor args.
tilelang/language/eager/builder.py:1209
↓ 78 callersFunctionfunc_gen
Generate a function for each PrimExpr dtype. Parameters ---------- name: str The ffi function name to call.
tilelang/language/ast/ir.py:1331
↓ 77 callersFunctionis_fragment
Check if the buffer is a fragment (e.g., for matrix multiplication operations). Args: buffer: The TVM buffer, BufferLoad, or BufferR
tilelang/utils/language.py:108
↓ 71 callersMethodassert_allclose
Validates kernel output against a reference implementation. Args: reference_program: Reference implementation to compare against
tilelang/profiler/__init__.py:104
↓ 69 callersMethodindex
Property to retrieve the forward index of the layout. Returns ------- PrimExpr or List[PrimExpr] The com
tilelang/layout/layout.py:46
↓ 57 callersFunctionkernel
( K: T.Tensor(K_shape, dtype=input_dtype), V: T.Tensor(V_shape, dtype=input_dtype), Be
examples/kda/wy_fast.py:85
↓ 53 callersMethodrepeat
Repeat a layout along a single input dimension. This is useful for building a larger layout by tiling an "atom" layout. Conc
tilelang/layout/layout.py:123
↓ 52 callersFunctionassert_iter_sum_pattern
(expect_dict, dom_map, predicate=True, check_level="surjective", simplify_trivial_iterators=True)
testing/python/arith/test_arith_iter_affine_map.py:51
↓ 52 callersFunctionis_shared
Check if the buffer is in the shared memory scope. Args: buffer: The TVM buffer, BufferLoad, or BufferRegion to check. Returns:
tilelang/utils/language.py:48
↓ 47 callersMethodlower
Lower the GEMM operation to TIR statements based on target architecture.
tilelang/tileop/gemm/__init__.py:127
↓ 45 callersFunctionany
Create a new expression of the union of all conditions in the arguments Parameters ---------- args : list List of symbolic boolea
tilelang/language/tir/op.py:2072
↓ 45 callersFunctiondetermine_target
Determine the appropriate target for compilation (CUDA, HIP, or manual selection). Args: target (Union[str, dict, Target, Literal["a
tilelang/utils/target.py:222
↓ 44 callersFunctionall
Create a new expression of the intersection of all conditions in the arguments Parameters ---------- args : list List of sy
tilelang/language/tir/op.py:2091
↓ 39 callersMethodget_kernel_source
(self)
tilelang/jit/param.py:27
↓ 39 callersMethodupdate
(self, task: Task, is_interesting: bool)
tilelang/autodd.py:603
↓ 38 callersMethoddevice
(self)
examples/bitnet-1.58b/eval_utils.py:97
↓ 37 callersFunctionexp2
Calculate 2**x Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr The result.
tilelang/language/tir/op.py:2241
↓ 37 callersMethodexpand
Expand (lift) this layout by prepending new leading input dimensions. The new leading dimensions are forwarded unchanged to the outp
tilelang/layout/layout.py:168
↓ 36 callersFunctionto_buffer_region
Normalize a buffer-like object to BufferRegion, or encode a tl.region call. - Buffer/BufferLoad/BufferRegion -> returns BufferRegion when ex
tilelang/utils/language.py:192
↓ 35 callersMethodpop
Remove and return the top item from the stack. Returns: The top frame object from the stack Raises: IndexErr
tilelang/language/frame.py:35
↓ 34 callersMethodis_equal
Check if the current layout is equal to another layout. Parameters ---------- other : Layout The layout
tilelang/layout/layout.py:239
↓ 33 callersFunctionmax
Type-aware max: uses arith.maxsi for integers, nvvm.fmax for floats. Falls back to integer path if float conversion fails (signless int types).
tilelang/contrib/cutedsl/reduce.py:123
↓ 32 callersFunction_Simplify
(stmt: PrimFunc | IRModule, inline_let: bool = False)
tilelang/transform/simplify.py:31
↓ 30 callersMethodbind
(self, name: str, value: Any, annot: Any = empty)
tilelang/language/eager/ast.py:202
↓ 30 callersMethodreplace
(self, repl: dict[str, ast.AST])
tilelang/autodd.py:311
↓ 30 callersFunctionsimplify_and_compare
Helper function to run simplify pass and compare results.
testing/python/transform/test_tilelang_transform_simplify.py:22
↓ 28 callersFunction_ir
Extract MLIR IR value from a CuTeDSL value.
tilelang/contrib/cutedsl/gemm_tcgen05.py:122
↓ 27 callersFunction_dtype_forward
(func)
tilelang/language/ast/ir.py:1867
↓ 27 callersFunctioncompare_tensors
(name, x, y, atol=1e-5, rtol=1e-5)
examples/kda/test_utils_kda.py:43
↓ 27 callersFunctionkernel
( A: T.Tensor((N,), "float32"), B: T.Tensor((N,), "float32"), )
testing/python/cuda/test_tma_dsmem.py:12
↓ 26 callersMethodcount
(self)
maint/scripts/pytest_cuda_scheduler.py:62
↓ 26 callersFunctionmin
Type-aware min: uses arith.minsi for integers, nvvm.fmin for floats. Falls back to integer path if float conversion fails (signless int types).
tilelang/contrib/cutedsl/reduce.py:111
↓ 26 callersMethodreplicate
Replicate the Fragment across a new thread dimension. Parameters ---------- replicate : int The replicat
tilelang/layout/fragment.py:137
↓ 26 callersMethodverify
(self, data, dmap, expected)
testing/python/arith/test_arith_intset.py:28
↓ 25 callersFunction_dtype_forward
(func)
tilelang/language/tir/ir.py:152
↓ 25 callersFunction_shfl_xor_sync
Butterfly (XOR) shuffle within full warp.
tilelang/contrib/cutedsl/warp.py:84
↓ 24 callersFunction_elements_to_bytes
(elem_count: int, elem_bits: int)
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:97
↓ 24 callersFunctionis_full_region
Check whether a BufferRegion covers the full buffer region. A full region means each dimension has start 0 and extent equal to the corre
tilelang/utils/language.py:529
↓ 23 callersFunctionkernel
(b: T.Tensor[(16,), T.float32])
testing/python/transform/test_tilelang_transform_decouple_type_cast.py:303
↓ 22 callersFunctionassert_iter_sum_failure
(iters, dom_map, predicate=True, check_level="surjective")
testing/python/arith/test_arith_iter_affine_map.py:106
↓ 22 callersFunctionfloormod
Compute the floormod of two expressions. Parameters ---------- a : PrimExpr The left hand operand b : PrimExpr The r
tilelang/language/tir/op.py:3416
↓ 22 callersMethodwith_frame
(self, frame: AbstractContextManager[Any] | None)
tilelang/language/eager/builder.py:288
↓ 21 callersFunctionfloordiv
Compute the floordiv of two expressions. Parameters ---------- a : PrimExpr The left hand operand b : PrimExpr The r
tilelang/language/tir/op.py:3394
↓ 20 callersFunction_check
Helper function to verify structural equality after transformations
testing/python/transform/test_tilelang_transform_legalize_negative_index.py:7
↓ 20 callersFunctionassert_similar
(x, y, eps=1e-8, name="tensor", data="", raise_assert=True)
examples/gdn/test_utils.py:18
↓ 20 callersFunctioncopy
Copy data between memory regions. Args: src (Union[tirx.Buffer, tirx.BufferLoad, tirx.BufferRegion]): Source memory region dst (U
tilelang/language/copy_op.py:54
↓ 20 callersFunctiondo_bench
Do benchmark for a function.
examples/kda/test_utils_kda.py:86
↓ 20 callersFunctionfn
()
examples/dsa_hisa/hisa.py:207
↓ 20 callersFunctionifuse
Fuse iterators
testing/python/arith/test_arith_iter_affine_map.py:23
↓ 19 callersFunction_check_has_intrinsic
Check if the module contains a specific intrinsic call.
testing/python/transform/test_tilelang_transform_lower_ldgstg.py:32
↓ 19 callersFunctionisplit
Split iterators
testing/python/arith/test_arith_iter_affine_map.py:32
↓ 19 callersFunctionmake_swizzled_layout
(buffer: BufferLikeType, k_major: bool = True, allow_pad: bool = True)
tilelang/layout/swizzle.py:66
↓ 18 callersFunction_find_after
(src, needle, start=0)
testing/python/transform/test_tilelang_transform_producer_consumer_ws.py:297
↓ 18 callersFunctioncreate_iter
(name, extent)
testing/python/arith/test_arith_iter_affine_map.py:498
↓ 18 callersFunctionexp
Take exponential of input x. Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr
tilelang/language/tir/op.py:2225
↓ 18 callersMethodgeneric_visit
(self, node: ast.AST)
tilelang/language/eager/ast.py:42
↓ 18 callersMethodget_tag
(self, k: str)
tilelang/carver/roller/node.py:117
↓ 17 callersFunctionbuffer_region_to_tile_region
Clamp extents and return a tl.region call.
tilelang/language/utils.py:18
↓ 17 callersMethodcompute_warp_partition
Compute the warp partition (m_warp, n_warp) based on the given policy. Args: M (int): The number of rows in the GEMM wor
tilelang/tileop/base.py:65
↓ 17 callersMethodcurrent
(cls)
tilelang/language/eager/builder.py:221
↓ 17 callersFunctionjit_kernel
()
testing/python/language/test_tilelang_language_var_init.py:10
↓ 17 callersFunctionkernel
( A: T.Tensor((N,), dtype), B: T.Tensor((N,), dtype), )
testing/python/issue/test_tilelang_issue_814.py:10
↓ 17 callersFunctionretrieve_shape
Retrieve shape-like extents for a buffer-like object. - Buffer -> its `shape` - BufferRegion -> list of each range's `extent` - Buff
tilelang/utils/language.py:247
↓ 16 callersMethodCurrent
Returns the topmost (current) KernelLaunchFrame from the stack if it exists, or None if the stack is empty.
tilelang/language/kernel.py:173
↓ 16 callersFunction_skip_if_threads_exceed_n
Skip test if num_threads=256 and N < 32.
maint/gemm/correctness_evaluation.py:457
↓ 16 callersFunctioncheck_region_bound
Helper to check region bound estimation. Parameters ---------- expect_region: dict The keys are of form (begin, end) or PrimExpr
testing/python/arith/test_arith_intset.py:139
↓ 16 callersMethodclean_body
(self, body: list[ast.stmt], keep_one=True)
tilelang/autodd.py:894
↓ 16 callersFunctionconvert_division
(divisions)
testing/python/arith/test_arith_iter_affine_map.py:483
↓ 16 callersMethodencode
(self, prompts: List[str])
examples/bitnet-1.58b/vllm_workspace/conftest.py:395
↓ 16 callersFunctionfunc
( expr: Union[ None, PrimExpr, Literal["inf", "-inf", "nan"],
tilelang/language/ast/ir.py:1340
↓ 16 callersMethodgeneric_visit
(self, node)
tilelang/autodd.py:385
↓ 16 callersFunctionquote
(expr: str, *, passes: list[Any] | None = None, span=None, **kws)
tilelang/language/eager/ast.py:58
↓ 16 callersMethodrepeat
Returns a new Fragment that repeats the iteration space a given number of times. Parameters ---------- repeats : int
tilelang/layout/fragment.py:117
↓ 15 callersMethod_cxx_expr
Convert TVM expression to C++ string for generated launcher code.
tilelang/jit/adapter/cutedsl/wrapper.py:828
↓ 15 callersMethod_safe_write_file
Atomically write one cache file through a temporary sibling file.
tilelang/autotuner/param.py:171
↓ 15 callersMethoddisable_cache
(self)
tilelang/env.py:361
↓ 15 callersFunctionget_buffer_region_from_load
Get the buffer region from a buffer load. May encounter buffer load like C[0:128, 0:32], ref to pull request for buffer wise op: https:/
tilelang/language/utils.py:100
↓ 15 callersFunctionplot_layout
Plot the layout mapping as a 2D grid visualization. Dispatches to Fragment-specific or Layout-specific plotting based on the type of the
tilelang/tools/plot_layout.py:6
next →1–100 of 7,254, ranked by callers