MCPcopy Create free account

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

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

Methodcos_cached
(self)
examples/bitnet-1.58b/modeling_bitnet.py:136
Functioncosh
Take cosh of input x. Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr The re
tilelang/language/tir/op.py:2417
Functioncp_async_barrier_noinc
Perform a ptx async copy barrier using cp.async.mbarrier.arrive.noinc.
tilelang/language/builtin.py:1258
Functioncp_async_gs_conditional
(size, dst, src, cond)
tilelang/contrib/cutedsl/cpasync.py:72
Functioncp_async_only_num_stages_one
( x: T.Tensor((4 * bytes_per_copy,), T.uint8), y: T.Tensor((4 * bytes_per_copy,), T.uint8),
testing/python/issue/test_tilelang_issue_tma_no_ws.py:210
Functioncp_async_only_num_stages_zero
( x: T.Tensor((4 * bytes_per_copy,), T.uint8), y: T.Tensor((4 * bytes_per_copy,), T.uint8),
testing/python/issue/test_tilelang_issue_tma_no_ws.py:170
Functioncreate_barriers
TVM intrinsic to create N barriers Parameters ---------- barrier_count : int The number of barriers to create. Returns -
tilelang/language/tir/op.py:1983
Functioncreate_executable
Create executable binary. Parameters ---------- output : str The target executable. objects : List[str] List of obje
tilelang/contrib/cc.py:185
Methodcreate_matmul
(B, M, N, K)
tilelang/carver/template/flashattention.py:67
Methodcreate_node_from_function
(func, name)
tilelang/carver/template/flashattention.py:118
Functioncreate_reference_output
Create reference output - not actually used in verification
examples/autodd/tilelang_buggy.py:82
Functioncreate_shared
Create shared library. Parameters ---------- output : str The target shared library. objects : List[str] List of obj
tilelang/contrib/cc.py:104
Functioncreate_shared
(output: str, objects, options=None, cc=None, cwd=None, ccache_env=None)
tilelang/contrib/msvc.py:330
Functioncreate_staticlib
Create static library. Parameters ---------- output : str The target shared library. inputs : List[str] List of inpu
tilelang/contrib/cc.py:163
Functioncreate_tma_descriptor
Create a Tensor Memory Access (TMA) descriptor. This is an internal API used by copy lowering. The argument list depends on the tensor rank a
tilelang/language/builtin.py:294
Methodcreate_token_type_ids_from_sequences
Creates a mask from the two sequences passed to be used in a sequence-pair classification task. An ALBERT sequence pair mask has the
examples/bitnet-1.58b/tokenization_bitnet.py:379
Functioncross_compiler
Create a cross compiler function by specializing compile_func with options. This function can be used to construct compile functions that can
tilelang/contrib/cc.py:298
Methodctx_break
(self)
tilelang/language/eager/ast.py:195
Methodctx_break
(self)
tilelang/language/eager/builder.py:391
Methodctx_continue
(self)
tilelang/language/eager/ast.py:192
Methodctx_continue
(self)
tilelang/language/eager/builder.py:385
Methodctx_else
(self, val: _T)
tilelang/language/eager/ast.py:182
Methodctx_else
(self, val)
tilelang/language/eager/builder.py:313
Methodctx_for
(self, range: Iterable[Any])
tilelang/language/eager/ast.py:189
Methodctx_for
(self, it)
tilelang/language/eager/builder.py:345
Methodctx_if
(self, cond)
tilelang/language/eager/ast.py:175
Methodctx_if
(self, cond)
tilelang/language/eager/builder.py:296
Methodctx_then
(self, val: _T)
tilelang/language/eager/ast.py:178
Methodctx_then
(self, val)
tilelang/language/eager/builder.py:305
Methodctx_while
(self, cond: Callable[[], Any])
tilelang/language/eager/ast.py:198
Methodctx_while
(self, cond)
tilelang/language/eager/builder.py:397
Methodctx_with
(self, ctx: AbstractContextManager[Any])
tilelang/language/eager/ast.py:232
Methodctx_with
(self, ctx)
tilelang/language/eager/builder.py:682
Methodcuda_device_wrapper
(func: Callable[..., Any], device: int)
tilelang/autotuner/tuner.py:551
Functioncummax
( A: T.Tensor((M, N), dtype), B: T.Tensor((M, N), dtype), )
testing/python/language/test_tilelang_language_scan.py:317
Functioncummax
Compute the cumulative maximum of `src` along `dim`, writing results to `dst`. Negative `dim` indices are normalized (Python-style). If `dst
tilelang/language/scan_op.py:168
Functioncumsum
( A: T.Tensor((M, N), dtype), B: T.Tensor((M, N), dtype), )
testing/python/language/test_tilelang_language_scan.py:16
Functioncumsum
Compute the cumulative sum of `src` along `dim`, writing results to `dst`. Negative `dim` indices are normalized (Python-style). If `dst` is
tilelang/language/scan_op.py:84
Functioncumsum_region
( InputG_fragment: T.Tensor((N,), dtype), OutputG_fragment: T.Tensor((N,), dtype), )
testing/python/language/test_tilelang_language_scan.py:187
Functiondeallocate_tmem
Explicitly deallocate a TMEM buffer allocated by ``T.alloc_tmem``. By default, TileLang inserts a TMEM deallocation automatically at the end
tilelang/language/builtin.py:268
Functiondec_max_nreg
Decrement the maximum number of registers to use.
tilelang/language/builtin.py:403
Functiondecl_buffer
Create a buffer declaration node. Parameters ---------- shape : Union[List[PrimExpr], Tuple[PrimExpr], PrimExpr, Integral] The ty
tilelang/language/ast/ir.py:1133
Functiondecode_fp4_to_bf16_twiddling
Decode FP4 to BF16 using twiddling technique. Reference: triton/tensor_details/layout_details/hopper_value.py For each iteration:
tilelang/contrib/cutedsl/quantize.py:278
Functiondecode_i4s_to_f16
Decode signed INT4 to FP16. N must be even.
tilelang/contrib/cutedsl/quantize.py:128
Functiondecode_i4u_to_f16
Decode unsigned INT4 to FP16. Equivalent to CUDA template: decode_i4b_to_f16<T1, T2, false>(_i4u, B_local_decode, N); Args:
tilelang/contrib/cutedsl/quantize.py:98
Functiondecorator
(impl)
tilelang/autotuner/tuner.py:1405
Methoddefault_chat_template
LLaMA uses [INST] and [/INST] to indicate user messages, and <<SYS>> and <</SYS>> to indicate system messages. Assistant messages do
examples/bitnet-1.58b/tokenization_bitnet.py:411
Functiondemo
( A: T.Tensor[(1,), T.int32], B: T.Tensor[(1,), T.int32], )
testing/python/language/test_tilelang_language_augassign.py:6
Functiondeprecated
A decorator to indicate that a method is deprecated Parameters ---------- method_name : str The name of the method to deprecate
tilelang/utils/deprecated.py:13
Functiondescribe_supported_targets
Return a mapping of supported target names to usage descriptions.
tilelang/utils/target.py:111
Functiondetect_dominant_read
Detect the dominant read indices in the block.
tilelang/carver/analysis.py:232
Functiondevice_assert
Device-side assert emulation. Emits a device-side assert call on CUDA targets when CUDA is available. The assert is always enabled and ca
tilelang/language/print_op.py:141
Methoddevice_func
(self)
tilelang/jit/adapter/wrapper.py:638
Functiondisable_warp_group_reg_alloc
Disable the warp group reg alloc.
tilelang/language/builtin.py:431
Functiondivf
( x: Union[TensorSSA, Numeric], y: Union[TensorSSA, Numeric], fastmath: bool = False, )
tilelang/contrib/cutedsl/math.py:147
Methoddom
The iteration domain of the loop.
tilelang/carver/analysis.py:36
Functiondp4a
Perform a 4-element dot product with accumulation (DP4A). Args: A (Buffer): First input buffer B (Buffer): Second input buffer
tilelang/language/customize.py:11
Functionds_read_tr16_b64
LDS transpose read, 64-bit, 16-element transpose (gfx950 only). Reads 8 bytes from LDS (__shared__ memory) with a 16-element transpose. Used
tilelang/language/builtin.py:1463
Functionds_read_tr8_b64
LDS transpose read, 64-bit, 8-element transpose (gfx950 only). Reads 8 bytes from LDS (__shared__ memory) with an 8-element transpose. Used f
tilelang/language/builtin.py:1484
Functiondummy_kernel
( a: T.Tensor[(1,), T.float32], )
testing/python/language/test_tilelang_memory_leak.py:16
Methode_dtype
(self)
tilelang/tileop/gemm_sp/gemm_sp_base.py:58
Functionelem_add
(A: T.Tensor((M, N), in_dtype), B: T.Tensor((M, N), in_dtype), C: T.Tensor((M, N), out_dtype))
testing/python/language/test_tilelang_language_tma_1d.py:13
Methodemit
Emit a log record. Messages are written to tqdm to ensure output in progress bars isn't corrupted.
tilelang/__init__.py:80
Functionempty
(shape, dtype: DType = _dtypes.float32)
tilelang/language/allocate.py:339
Functionempty_kernel
()
testing/python/issue/test_tilelang_issue_830.py:12
Methodencode
(self, prompts: List[str])
examples/bitnet-1.58b/vllm_workspace/conftest.py:536
Functionend_profile_intrinsic
End profile intrinsic. Parameters ---------- id : int The intrinsic id. Returns ------- call : PrimExpr The ca
tilelang/language/tir/op.py:401
Functionenv_thread
Bind a var to thread env Parameters ---------- thread_tag : str The thread type tag. dtype : str The data type of th
tilelang/language/ast/ir.py:1239
Methodeot_token_id
(self)
examples/bitnet-1.58b/eval_utils.py:62
Functionerf
Take gauss error function of the input x. Parameters ---------- x : PrimExpr Input argument. Returns ------- y : Pri
tilelang/language/tir/op.py:2273
Methoderr_msg
()
testing/python/arith/test_arith_intset.py:31
Methodeval
(self, val: Any)
tilelang/language/eager/builder.py:321
Functionexample_long_prompts
()
examples/bitnet-1.58b/vllm_workspace/conftest.py:118
Functionexample_prompts
()
examples/bitnet-1.58b/vllm_workspace/conftest.py:110
Functionexp
(x: Union[TensorSSA, Numeric], fastmath: bool = False, **kwargs)
tilelang/contrib/cutedsl/math.py:82
Functionexp10
Compute 10^x using exp2(x * log2(10)).
tilelang/contrib/cutedsl/math.py:122
Functionexp10
Calculate 10**x Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr The result.
tilelang/language/tir/op.py:2257
Functionexpected
(A: T.Buffer((4, 4), "float32"), E: T.Buffer((4, 4), "float32"))
testing/python/transform/test_tilelang_transform_thread_sync.py:221
Functionexpected
( A: T.Tensor((128,), T.float32), B: T.Tensor((128,), T.float32), cond: T.Tensor((1,),
testing/python/transform/test_tilelang_transform_loop_unswitching.py:37
Functionexpected
(A: T.Buffer((4,), "float32"))
testing/python/transform/test_tilelang_transform_if_stmt_binding.py:21
Functionexpected
( A: T.Tensor((M, N), dtype=dtype), )
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:95
Functionexpected
(A: T.Tensor((128, 128), T.float32), B: T.Tensor((128, 128), T.float32))
testing/python/transform/test_tilelang_transform_let_inline.py:25
Functionexpected
(A: T.Buffer((1,), "float32"))
testing/python/transform/test_tilelang_transform_simplify.py:152
Functionexpected
( A: T.Tensor((M, N, vec_len), dtype=T.float32), )
testing/python/transform/test_tilelang_transform_legalize_vectorized_loop.py:23
Methodexport_ptx
Export compiled PTX to a file (CUDA only). Parameters ---------- path : str Destination file path to wri
tilelang/jit/kernel.py:740
Methodexport_sass
Export disassembled SASS to a file (CUDA only). Parameters ---------- path : str Destination file path t
tilelang/jit/kernel.py:802
Methodextract_param_names_ast
Extract parameter names using AST parsing.
tilelang/jit/adapter/cutedsl/wrapper.py:944
Methodextract_param_names_split
Fallback: extract parameter names using naive split-based parsing.
tilelang/jit/adapter/cutedsl/wrapper.py:978
Methodf
()
tilelang/language/ast/ir.py:1738
Functionf_convert
(nbit: int, val: tirx.PrimExpr, pos: tirx.PrimExpr, dtype: str)
tilelang/quantize/quantization.py:119
Functionfabsf
(x: Union[TensorSSA, Numeric], fastmath: bool = False)
tilelang/contrib/cutedsl/math.py:128
Functionfail_compile
(self, *args, **kwargs)
testing/python/cache/test_tilelang_kernel_cache_atomic_save.py:256
Functionfail_from_database
(cls, **kwargs)
testing/python/cache/test_tilelang_kernel_cache_atomic_save.py:276
Functionfail_source_load
(*args, **kwargs)
testing/python/cache/test_tilelang_kernel_cache_atomic_save.py:82
Functionfake_from_database
(cls, **kwargs)
testing/python/cache/test_tilelang_kernel_cache_atomic_save.py:85
Functionfake_load_frontend_cached
(frontend_key_data, **kwargs)
testing/python/cache/test_tilelang_kernel_cache_atomic_save.py:252
Functionfast_dequant_bf16_fp4_twiddling
(B_shared, B_dequantize_shared, Scale, k)
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_cdna4.py:133
Functionfast_dequant_bf16_fp4_twiddling
Fast dequantization kernel: convert packed 4-bit quantized values in B_shared to bfloat16 in B_dequantize_shared using an ext
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_hopper.py:197
Functionfast_dequant_bf16_fp4_twiddling
Fast dequantization kernel routine that converts packed FP4 values in shared memory to BF16 and writes the results back into a shared deq
examples/dequantize_gemm/example_dequant_gemm_bf16_fp4_hopper.py:148
Functionfence_barrier_init
()
tilelang/contrib/cutedsl/cpasync.py:338
← previousnext →4,301–4,400 of 7,254, ranked by callers