MCPcopy Create free account

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

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

↓ 1 callersMethod_frontend_cache_key_data
(self, key: tuple)
tilelang/jit/__init__.py:430
↓ 1 callersFunction_fsqrt_f32
(a: Float32, *, rounding: str = "rn", loc=None, ip=None)
tilelang/contrib/cutedsl/ieee_math.py:116
↓ 1 callersFunction_fsub_f32
(a: Float32, b: Float32, *, rounding: str = "rn", loc=None, ip=None)
tilelang/contrib/cutedsl/ieee_math.py:48
↓ 1 callersFunction_fwd_kernel_inner
( acc, l_i, m_i, q, k_block_col_idx, block_mask_ptr, k_ptrs, v_ptrs, offs_
benchmark/blocksparse_attention/benchmark_triton_block_sparse_fmha.py:35
↓ 1 callersFunction_fwd_kernel_inner
( acc, l_i, m_i, q, k_block_col_idx, block_mask_ptr, k_ptrs, v_ptrs, offs_
examples/blocksparse_attention/block_sparse_attn_triton.py:34
↓ 1 callersFunction_fwd_kernel_inner
( acc, l_i, m_i, q, k_block_col_idx, block_mask_ptr, k_ptrs, v_ptrs, offs_
examples/seer_attention/block_sparse_attn_triton.py:35
↓ 1 callersMethod_generate_cpp_launcher
Generate complete C++ launcher code using templates. TMA descriptors are stored on HOST memory in stack-local tma_descs[] array. cuLa
tilelang/jit/adapter/cutedsl/wrapper.py:1226
↓ 1 callersMethod_generate_cubin_gen_code
Generate cubin generation code for Python wrapper using templates. Args: lib_code: The CuTeDSL kernel definitions (@cute.kernel d
tilelang/jit/adapter/cutedsl/wrapper.py:1286
↓ 1 callersMethod_generate_kernel_init
Generate kernel initialization code.
tilelang/jit/adapter/cutedsl/wrapper.py:1139
↓ 1 callersMethod_generate_kernel_launch
Generate single kernel launch code. For __grid_constant__ CUtensorMap params: - Pass CUtensorMap* directly (not CUtensorMap**)
tilelang/jit/adapter/cutedsl/wrapper.py:1147
↓ 1 callersMethod_generate_key
Generates a unique hash key for caching compiled kernels. Args: func (Callable): The function to be compiled.
tilelang/cache/kernel_cache.py:236
↓ 1 callersMethod_generate_python_wrapper
Generate Python wrapper code.
tilelang/jit/adapter/cutedsl/wrapper.py:1385
↓ 1 callersFunction_generate_semi_sparse
(m: int, k: int, dtype: torch.dtype, device: str = "cuda")
benchmark/matmul/benchmark_matmul_sp_compress.py:18
↓ 1 callersMethod_generate_tma_desc_init
Generate single TMA descriptor initialization code.
tilelang/jit/adapter/cutedsl/wrapper.py:1038
↓ 1 callersMethod_generate_tma_init_func
Generate TMA init function code (creates descriptors in caller-provided host array). TMA descriptors are stored in stack-local tma_descs[] ar
tilelang/jit/adapter/cutedsl/wrapper.py:1081
↓ 1 callersMethod_generate_tma_launch_init
Generate TMA initialization code for launch function (host memory mode). TMA descriptors stay on host. cuLaunchKernel copies them to param sp
tilelang/jit/adapter/cutedsl/wrapper.py:1115
↓ 1 callersMethod_get_cache_dir
(cls)
tilelang/autotuner/tuner.py:253
↓ 1 callersMethod_get_cache_namespace
()
tilelang/cache/kernel_cache.py:157
↓ 1 callersFunction_get_clc_query_codegen_source
()
testing/python/language/test_tilelang_language_cluster.py:62
↓ 1 callersMethod_get_compile_args
()
tilelang/cache/kernel_cache.py:60
↓ 1 callersMethod_get_complete_cache_files
(self, cache_path: str)
tilelang/cache/kernel_cache.py:755
↓ 1 callersMethod_get_complete_result_files
Return the full file set that marks an autotune result complete.
tilelang/autotuner/param.py:559
↓ 1 callersMethod_get_default
(self)
tilelang/env.py:256
↓ 1 callersFunction_get_float8_dtypes
Collect available float8 dtypes - some may not exist in older torch versions.
tilelang/utils/tensor.py:9
↓ 1 callersFunction_get_gemm_ops
()
tilelang/metal/transform/metal_fragment_to_simdgroup.py:20
↓ 1 callersFunction_get_isfinite_expr
(code: str)
testing/python/language/test_tilelang_language_isfinite_codegen.py:25
↓ 1 callersFunction_get_isfinite_source
()
testing/python/language/test_tilelang_language_isfinite_codegen.py:8
↓ 1 callersFunction_get_issue_2307_source
()
testing/python/issue/test_tilelang_issue_2307.py:7
↓ 1 callersFunction_get_l2_cache_size_bytes
(target: Target)
tilelang/carver/arch/cuda.py:60
↓ 1 callersFunction_get_l2_cache_size_bytes
(target: Target, device: tvm.runtime.Device)
tilelang/carver/arch/cdna.py:15
↓ 1 callersFunction_get_laneid_kernel
(num_threads: int = 128, warp_size: Optional[int] = None)
testing/python/language/test_tilelang_language_get_warp_info.py:25
↓ 1 callersMethod_get_ldmatrix_index_map_32
Index maps for 32x32xK MFMA instructions (M_DIM=N_DIM=32). For int8 with mfma_shape=(32,32,32): k_dim*k_pack=32, local_size=16 so the
tilelang/rocm/intrinsics/mfma_macro_generator.py:293
↓ 1 callersMethod_get_or_create_bitblas_operator
(self, config, enable_tuning)
examples/bitnet-1.58b/utils_quant.py:72
↓ 1 callersFunction_get_package_version
(pkg: str)
tilelang/env.py:92
↓ 1 callersFunction_get_pointer_type_annotation
(value: object)
tilelang/language/proxy.py:291
↓ 1 callersFunction_get_ptx_load_ordering
Get PTX memory ordering modifier for load. TileLang memory order: 0: relaxed -> .relaxed.gpu 1: consume -> .acquire.gpu (consume
tilelang/contrib/cutedsl/atomic.py:443
↓ 1 callersFunction_get_ptx_store_ordering
Get PTX memory ordering modifier for store. TileLang memory order: 0: relaxed -> .relaxed.gpu 1: consume -> .release.gpu (invalid
tilelang/contrib/cutedsl/atomic.py:542
↓ 1 callersMethod_get_required_files
(self, cache_path: str)
tilelang/cache/kernel_cache.py:750
↓ 1 callersFunction_get_target
Return TVM hip target for gfx950.
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_cdna4.py:44
↓ 1 callersFunction_get_target_attr
(target: Target, name: str)
tilelang/carver/arch/cuda.py:53
↓ 1 callersFunction_get_tensor_instructions_for_generation
(rdna_generation: int)
tilelang/carver/arch/rdna.py:50
↓ 1 callersMethod_get_tilelang_lib_stamp
Return a content-based build-stamp for TileLang native libraries. The kernel cache key historically only depended on `tilelang.__version__`
tilelang/cache/kernel_cache.py:75
↓ 1 callersFunction_get_tma_create_tiled
()
testing/python/runtime/test_tilelang_runtime_tma_validation.py:29
↓ 1 callersFunction_get_unpad_data
(attention_mask)
examples/bitnet-1.58b/modeling_bitnet.py:76
↓ 1 callersFunction_get_warp_group_idx_kernel
( num_threads: int = 128, warp_size: Optional[int] = None, warps_per_group: Optional[int] = None,
testing/python/language/test_tilelang_language_get_warp_info.py:58
↓ 1 callersFunction_get_warp_idx_kernel
(num_threads: int = 128, warp_size: Optional[int] = None)
testing/python/language/test_tilelang_language_get_warp_info.py:47
↓ 1 callersFunction_get_warp_idx_sync_kernel
(num_threads: int = 128, warp_size: Optional[int] = None)
testing/python/language/test_tilelang_language_get_warp_info.py:36
↓ 1 callersFunction_get_workspace_dir_name
()
tilelang/jit/env.py:32
↓ 1 callersFunction_grid_sync_barrier
Software grid barrier using inline PTX. Declares a module-level global counter via llvm.mlir.global, then uses inline PTX for the barrier pro
tilelang/contrib/cutedsl/grid_sync.py:88
↓ 1 callersMethod_handle_timeout
Handle cleanup logic when a timeout happens.
tilelang/autodd.py:843
↓ 1 callersFunction_has_cast
Check if a statement tree contains any Cast node outside of indices.
tilelang/transform/decouple_type_cast.py:117
↓ 1 callersFunction_has_freeze_import
Return True if *source* already contains ``from tilelang.autodd import __freeze__`` as an actual import statement (not inside a comment or string
tilelang/autodd.py:971
↓ 1 callersFunction_has_sm100
()
testing/python/language/test_tilelang_language_tma_gather_scatter.py:11
↓ 1 callersFunction_import_vsdevcmd_environment
(vsdevcmd: str)
tilelang/contrib/msvc.py:172
↓ 1 callersMethod_infer
(self, shape: dict[str, list[arith.ConstIntBound]], rstep: dict[str, int])
tilelang/carver/roller/shape_inference/common.py:22
↓ 1 callersFunction_init_logger
Initialize the logger specific for this module with custom settings and a Tqdm-based handler.
tilelang/__init__.py:66
↓ 1 callersFunction_init_logger_handlers
()
tilelang/autotuner/tuner.py:176
↓ 1 callersMethod_init_rope
(self)
examples/bitnet-1.58b/modeling_bitnet.py:344
↓ 1 callersMethod_init_rope
(self)
examples/bitnet-1.58b/modeling_bitnet.py:460
↓ 1 callersMethod_init_with_prim_func
(self, func: tvm.tirx.PrimFunc, name: str = "PrimFuncNode")
tilelang/carver/roller/policy/default.py:47
↓ 1 callersMethod_initialize_abbrev
(self, a_dtype, b_dtype, accum_dtype)
tilelang/rocm/intrinsics/mfma_macro_generator.py:161
↓ 1 callersMethod_initialize_abbrev
(self, a_dtype, b_dtype, accum_dtype)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:103
↓ 1 callersMethod_initialize_abbrev
(self, a_dtype, b_dtype, accum_dtype)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:135
↓ 1 callersMethod_initialize_abbrev
(self, a_dtype, b_dtype, accum_dtype)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:135
↓ 1 callersMethod_initialize_b_preshuffle
(self, b_preshuffle: bool | None = False)
tilelang/rocm/intrinsics/mfma_macro_generator.py:232
↓ 1 callersMethod_initialize_is_m_first
(self, is_m_first: bool | None = False)
tilelang/rocm/intrinsics/mfma_macro_generator.py:228
↓ 1 callersMethod_initialize_is_m_first
(self, is_m_first: bool | None = False)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:138
↓ 1 callersMethod_initialize_is_m_first
(self, is_m_first: bool | None = False)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:213
↓ 1 callersMethod_initialize_is_m_first
(self, is_m_first: bool | None = False)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:177
↓ 1 callersMethod_initialize_k_dim
(self, a_dtype=T.float16)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:95
↓ 1 callersMethod_initialize_k_dim
(self, a_dtype=T.float16)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:118
↓ 1 callersMethod_initialize_k_dim
(self, a_dtype=T.float16)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:118
↓ 1 callersMethod_initialize_k_pack
(self, k_pack: int | None = None)
tilelang/rocm/intrinsics/mfma_macro_generator.py:213
↓ 1 callersMethod_initialize_local_size
(self, m_dim=16, n_dim=16, k_dim=16, warp_size=32)
tilelang/rocm/intrinsics/mfma_macro_generator.py:156
↓ 1 callersMethod_initialize_local_size
(self, m_dim=16, n_dim=16, k_dim=16)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:98
↓ 1 callersMethod_initialize_local_size
(self, m_dim=16, n_dim=16, k_dim=16, warp_size=32)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:130
↓ 1 callersMethod_initialize_local_size
(self, m_dim=16, n_dim=16, k_dim=16, warp_size=32)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:125
↓ 1 callersMethod_initialize_m_dim
(self, a_dtype=T.float16)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:123
↓ 1 callersMethod_initialize_mfma_prefix
(self, k_dim=16)
tilelang/rocm/intrinsics/mfma_macro_generator.py:166
↓ 1 callersMethod_initialize_mfma_shape
Set ``(M_DIM, N_DIM, k_dim)`` from an explicit shape or auto-detect. Supported shapes on CDNA4 (gfx950) for int8: (16, 16, 32) —
tilelang/rocm/intrinsics/mfma_macro_generator.py:123
↓ 1 callersMethod_initialize_micro_size
(self, m_dim=16, n_dim=16, k_dim=16)
tilelang/rocm/intrinsics/mfma_macro_generator.py:208
↓ 1 callersMethod_initialize_micro_size
(self, m_dim: int = 16, k_dim: int = 16)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:121
↓ 1 callersMethod_initialize_micro_size
(self, m_dim: int = 16, k_dim: int = 16)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:183
↓ 1 callersMethod_initialize_micro_size
(self, m_dim: int = 16, k_dim: int = 16)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:153
↓ 1 callersMethod_initialize_mma_prefix
(self, k_dim: int = 16)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:114
↓ 1 callersMethod_initialize_mma_prefix
(self, k_dim: int = 16)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:157
↓ 1 callersMethod_initialize_mma_sp_prefix
(self, k_dim: int = 16)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:140
↓ 1 callersMethod_initialize_preshuffle
(self, a_preshuffle: bool | None, b_preshuffle: bool | None)
tilelang/rocm/intrinsics/mfma_macro_generator.py:840
↓ 1 callersMethod_initialize_transform_kind
(self, transform_kind_a, transform_kind_b)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:964
↓ 1 callersMethod_initialize_wgmma_prefix
(self, n_dim: int = 16)
tilelang/cuda/intrinsics/macro/wgmma_macro_generator.py:152
↓ 1 callersMethod_initialize_wgmma_prefix
(self, n_dim: int = 16)
tilelang/cuda/intrinsics/macro/wgmma_sp_macro_generator.py:89
↓ 1 callersFunction_invalid_fragment_write_owner_layout
()
testing/python/layout/test_tilelang_layout_inference.py:37
↓ 1 callersFunction_is_any_dtype
Check if obj is a dtype-like value. Use instead of isinstance(obj, AnyDType) to keep the runtime check explicit.
tilelang/language/dtypes.py:30
↓ 1 callersFunction_is_call_to
(expr, op_name)
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:44
↓ 1 callersFunction_is_cutedsl_available
()
testing/python/jit/test_tilelang_jit_cutedsl.py:9
↓ 1 callersFunction_is_int_zero
(expr)
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:48
↓ 1 callersMethod_is_lazy_style
Check if the function uses lazy style (explicitly returns PrimFunc). Lazy style functions define an inner @T.prim_func and return it
tilelang/language/eager/builder.py:1152
↓ 1 callersMethod_is_rename_collision
(exc: OSError)
tilelang/cache/kernel_cache.py:779
↓ 1 callersMethod_is_rename_collision
Return whether a rename failure came from a benign race.
tilelang/autotuner/param.py:594
↓ 1 callersFunction_is_running_autodd
Detect if we are running under `python -m tilelang.autodd`.
tilelang/env.py:99
← previousnext →1,601–1,700 of 7,254, ranked by callers