MCPcopy Create free account

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

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

↓ 1 callersMethod_safe_write_executable
(cls, executable: Executable, path: str)
tilelang/cache/kernel_cache.py:459
↓ 1 callersMethod_safe_write_executable
Atomically export one runtime executable to disk.
tilelang/autotuner/param.py:183
↓ 1 callersMethod_save_cubin_to_cache_if_needed
Save cubin to cache directory after first execution. This is called after the first kernel execution to ensure the generated cubin fi
tilelang/jit/adapter/cutedsl/adapter.py:286
↓ 1 callersMethod_save_kernel_source_code_to_disk
(self, kernel: JITKernel, cache_path: str, verbose: bool = False)
tilelang/cache/kernel_cache.py:730
↓ 1 callersMethod_save_kernel_to_disk
Persists a compiled kernel to disk cache. Args: cache_path (Path): The root path for the cache files. kernel
tilelang/autotuner/param.py:191
↓ 1 callersMethod_save_result_to_disk
(self, key, result: AutotuneResult)
tilelang/autotuner/tuner.py:465
↓ 1 callersMethod_save_so_cubin_to_disk
(self, kernel: JITKernel, cache_path: str, verbose: bool = False)
tilelang/cache/kernel_cache.py:743
↓ 1 callersMethod_save_wrapper_kernel_code_to_disk
(self, kernel: JITKernel, cache_path: str, verbose: bool = False)
tilelang/cache/kernel_cache.py:737
↓ 1 callersFunction_scalar_arg_type
(arg)
tilelang/contrib/cutedsl/math.py:29
↓ 1 callersFunction_scalar_reduce_accumulator_owner_layout
()
testing/python/layout/test_tilelang_layout_inference.py:64
↓ 1 callersMethod_set_adapter_cache_path
(self, kernel: JITKernel, cache_path: str)
tilelang/cache/kernel_cache.py:801
↓ 1 callersFunction_shuffle_elect_kernel
(num_threads: int = 128, thread_extent: int = 64)
testing/python/language/test_tilelang_language_get_warp_info.py:73
↓ 1 callersFunction_simple_add_kernel
()
testing/python/transform/test_readonly_param_const_codegen.py:6
↓ 1 callersFunction_smem_optin_bytes
Per-block opt-in dynamic shared memory limit (bytes) of the current CUDA device, queried via cudaDevAttrMaxSharedMemoryPerBlockOptin. Referen
testing/python/language/test_tilelang_language_transpose.py:98
↓ 1 callersMethod_spec_key
(spec: object)
maint/scripts/pytest_cuda_scheduler.py:82
↓ 1 callersMethod_specialize_func
(self, func: PrimFunc)
tilelang/carver/roller/node.py:231
↓ 1 callersFunction_strip_mvb_attrs
Remove intermediate MVB attributes that are consumed by later passes.
testing/python/transform/test_tilelang_transform_Inject_software_pipeline.py:19
↓ 1 callersMethod_substitute_dynamic_symbols
Substitute dynamic symbolic variables in param shape with concrete values. Args: param: The kernel parameter with potentially dyn
tilelang/profiler/__init__.py:72
↓ 1 callersMethod_target_arch
Return the CUDA SM architecture requested by the TileLang target.
tilelang/jit/adapter/cutedsl/wrapper.py:823
↓ 1 callersFunction_tcgen05_cp_warpx4_chunked
(buffer, tmem_ptr, tmem_col_offset, base_offset)
tilelang/language/builtin.py:1330
↓ 1 callersFunction_tcgen05_sf_warp_transpose_chunked
(buffer, base_offset)
tilelang/language/builtin.py:1356
↓ 1 callersFunction_test_compress
(dtype, meta_dtype)
testing/python/utils/test_compress_utils.py:88
↓ 1 callersFunction_tilelang_issue_layout_free_inference_choose_smallest_replication
()
testing/python/layout/test_tilelang_layout_inference.py:8
↓ 1 callersFunction_tir_packed_int_to_int_convert
(storage_type="uint", storage_nbit=8)
tilelang/quantize/quantization.py:281
↓ 1 callersFunction_tir_packed_to_unsigned_convert
(storage_type="uint", storage_nbit=8)
tilelang/quantize/quantization.py:258
↓ 1 callersFunction_tir_u8_to_f4_to_bf16
Convert a packed 4-bit FP4 field to bfloat16 (mirrors Hopper example, ignoring overflow clamp).
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_cdna4.py:26
↓ 1 callersFunction_tir_u8_to_f4_to_bf16
Convert a 4-bit field packed in a uint8 into a bfloat16 value, applying an exponent scale. This helper extracts a 4-bit nibble from `val` at
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_hopper.py:10
↓ 1 callersFunction_tir_u8_to_f4_to_bf16
Convert a 4-bit FP4 value packed in a uint8 byte into a bfloat16 value. This helper extracts the 4-bit field located at the
examples/dequantize_gemm/example_dequant_gemm_bf16_fp4_hopper.py:217
↓ 1 callersFunction_tmp_var_kernel
(N, block_N, dtype=T.float32)
testing/python/issue/test_tilelang_issue_814.py:8
↓ 1 callersFunction_to_ir_value
Convert value to MLIR IR, handling both cutlass types and raw MLIR Values
tilelang/contrib/cutedsl/ldsm.py:31
↓ 1 callersFunction_traverse
(block)
tilelang/carver/roller/node.py:23
↓ 1 callersFunction_triton_mixed_sparse_attention
( q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, seqlens: torch.Tensor, block_count: t
examples/minference/example_vertical_slash_sparse_attn.py:420
↓ 1 callersMethod_upad_input
(self, query_layer, key_layer, value_layer, attention_mask, query_length)
examples/bitnet-1.58b/modeling_bitnet.py:701
↓ 1 callersMethod_update_causal_mask
(self, attention_mask, input_tensor, cache_position)
examples/bitnet-1.58b/modeling_bitnet.py:1091
↓ 1 callersMethod_validate_input_supply_requirements
(self, prim_func: PrimFunc, out_idx: list[int] | int | None)
tilelang/autotuner/tuner.py:417
↓ 1 callersFunction_vs_install_roots
()
tilelang/contrib/msvc.py:31
↓ 1 callersFunction_workers_per_device
Return number of workers per GPU from env or default.
maint/scripts/pytest_cuda_scheduler.py:31
↓ 1 callersMethod_wrap_with_allocations
Wrap statement with buffer declarations and allocations.
tilelang/transform/decouple_type_cast.py:541
↓ 1 callersFunctionact_quant_kernel
(N, in_dtype=BF16, out_dtype=FP8, scale_dtype=FP32, round_scale=False)
examples/deepseek_v32/inference/kernel.py:34
↓ 1 callersFunctionadapt_tensor
(arg)
tilelang/contrib/dlpack.py:47
↓ 1 callersMethodadd_next
(self, node)
tilelang/carver/roller/shape_inference/tir.py:57
↓ 1 callersMethodadd_prev
(self, node)
tilelang/carver/roller/shape_inference/tir.py:61
↓ 1 callersFunctionafter
()
testing/python/transform/test_tilelang_transform_layout_inference.py:48
↓ 1 callersFunctionalloc_global
( N, block_N, dtype, )
testing/python/language/test_tilelang_language_alloc.py:166
↓ 1 callersFunctionalloc_global_eagerjit
(A, block_N, dtype)
testing/python/language/test_tilelang_language_alloc.py:201
↓ 1 callersFunctionalloc_multi_vars_with_initializer
( N, block_N, dtype, )
testing/python/language/test_tilelang_language_alloc.py:125
↓ 1 callersFunctionalloc_tcgen05_instruction_desc
(dtype: DType = _dtypes.uint32)
tilelang/language/allocate.py:329
↓ 1 callersFunctionalloc_var
( N, block_N, dtype, )
testing/python/language/test_tilelang_language_alloc.py:6
↓ 1 callersFunctionalloc_var_add
( N, block_N, dtype, )
testing/python/language/test_tilelang_language_alloc.py:44
↓ 1 callersFunctionalloc_var_with_initializer
( N, block_N, dtype, init_value, )
testing/python/language/test_tilelang_language_alloc.py:84
↓ 1 callersFunctionanalysis_tensorcore_tags
(sch: Schedule, block: SBlockRV, target: Target)
tilelang/carver/matmul_analysis.py:552
↓ 1 callersMethodanalyze
(self)
tilelang/carver/roller/shape_inference/tir.py:122
↓ 1 callersFunctionannotated_layout_on_dtype_changing_view_test
()
testing/python/language/test_tilelang_language_view.py:127
↓ 1 callersFunctionapply_simplify
Apply Simplify pass to a PrimFunc or IRModule.
tilelang/transform/simplify.py:61
↓ 1 callersFunctionassert_atomic_add_return_access_ptr_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:457
↓ 1 callersFunctionassert_atomic_load_access_ptr_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:442
↓ 1 callersFunctionassert_atomic_store_access_ptr_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:462
↓ 1 callersFunctionassert_bitnet_158_int8xint2_decode_correctness
(M, N, K, in_dtype, out_dtype, accum_dtype, fast_decoding=True)
examples/bitnet-1.58b/kernel_benchmark/tilelang_bitnet_158_int8xint2_decode.py:237
↓ 1 callersFunctionassert_bitnet_158_int8xint2_prefill_correctness
(M, N, K, in_dtype, out_dtype, accum_dtype, fast_decoding=True)
examples/bitnet-1.58b/kernel_benchmark/tilelang_bitnet_158_int8xint2_prefill.py:360
↓ 1 callersFunctionassert_call_extern_multiple_access_ptrs_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:478
↓ 1 callersFunctionassert_close
(name, expect, actual, atol=1e-3, rtol=1e-3)
examples/blocksparse_attention/example_tilelang_sparse_gqa_decode_varlen_indice.py:312
↓ 1 callersFunctionassert_close
(name, expect, actual, atol=1e-3, rtol=1e-3)
examples/blocksparse_attention/example_tilelang_sparse_gqa_decode_varlen_mask.py:303
↓ 1 callersFunctionassert_cp_async_access_ptr_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:238
↓ 1 callersFunctionassert_cp_async_access_ptr_nonzero_safe_value_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:430
↓ 1 callersFunctionassert_gemm_codegen
( M, N, K, block_M, block_N, block_K, dtype=T.float16, accum_dtype=T.float32,
testing/python/webgpu/test_webgpu_codegen.py:34
↓ 1 callersFunctionassert_matmul_codegen
(M=1024, N=1024, K=1024, block_M=128, block_N=128, block_K=32)
testing/python/cpu/test_tilelang_cpu_gemm.py:48
↓ 1 callersFunctionassert_oob_store_legalize
(M: int = 64, N: int = 64)
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:193
↓ 1 callersFunctionassert_ss_gemm
Compile and run an SS GEMM, asserting correctness against a CPU reference.
testing/python/kernel/test_tilelang_kernel_bf16_gemm_tcgen5_ts.py:142
↓ 1 callersFunctionassert_tl_matmul_correctness
(M, N, K, k_pack=1, a_transposed=False, b_transposed=True)
examples/gemm_fp8/example_tilelang_gemm_amd_fp8_preshuffle.py:169
↓ 1 callersFunctionassert_tl_matmul_correctness
(M, N, K, in_dtype, out_dtype, accum_dtype=T.float32, a_transposed=False, b_transposed=True, k_pack=1)
testing/python/amd/test_tilelang_gemm_mfma_intrinsic.py:165
↓ 1 callersFunctionassert_tl_matmul_with_ladder_weight_only_transform_block_reduce_int4_correctness
( M, N, K, in_dtype, out_dtype, accum_dtype, transform_b, )
examples/dequantize_gemm/example_dequant_gemm_fine_grained.py:357
↓ 1 callersFunctionassert_vectorize_access
(M: int = 64, N: int = 64)
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:110
↓ 1 callersFunctionassert_vectorize_access
(M: int = 64, N: int = 64)
testing/python/transform/test_tilelang_transform_legalize_vectorized_loop.py:36
↓ 1 callersFunctionassert_vectorize_access_with_atmoic_add
(M: int = 64, N: int = 64)
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:155
↓ 1 callersMethodassign_block_size
Assigns block sizes to the TileDict based on the recommended block sizes. Parameters ---------- td : TileDict
tilelang/carver/roller/policy/default.py:640
↓ 1 callersFunctionast_replace
(node: ast.AST, **changes)
tilelang/autodd.py:69
↓ 1 callersFunctionatomic_add_auto_vectorized_unit_test
(vec_size: int, dtype=T.float32)
testing/python/language/test_tilelang_language_atomic.py:225
↓ 1 callersFunctionatomic_add_complicated_parallel_program
(K, M, N, block_M, block_N, dtype=T.float32)
testing/python/language/test_tilelang_language_atomic.py:242
↓ 1 callersFunctionatomic_add_return_access_ptr_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:320
↓ 1 callersFunctionatomic_addx2_program
(M, N, block_M, block_N, dtype=T.float16)
testing/python/language/test_tilelang_language_atomic.py:63
↓ 1 callersFunctionatomic_addx4_program
(M, N, block_M, block_N)
testing/python/language/test_tilelang_language_atomic.py:128
↓ 1 callersFunctionatomic_different_memory_orders_program
(M, N, block_M, block_N, dtype=T.float32)
testing/python/language/test_tilelang_language_atomic.py:93
↓ 1 callersFunctionatomic_load_access_ptr_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:294
↓ 1 callersFunctionatomic_load_store_program
(M, N, block_M, block_N, dtype=T.float32)
testing/python/language/test_tilelang_language_atomic.py:484
↓ 1 callersFunctionatomic_max_program
(K, M, N, block_M, block_N, dtype=T.float32)
testing/python/language/test_tilelang_language_atomic.py:418
↓ 1 callersFunctionatomic_memory_order_program
(K, M, N, block_M, block_N, dtype=T.float32)
testing/python/language/test_tilelang_language_atomic.py:37
↓ 1 callersFunctionatomic_min_program
(K, M, N, block_M, block_N, dtype=T.float32)
testing/python/language/test_tilelang_language_atomic.py:451
↓ 1 callersFunctionatomic_return_prev_program
(M, N, block_M, block_N, dtype=T.float32)
testing/python/language/test_tilelang_language_atomic.py:160
↓ 1 callersFunctionatomic_store_access_ptr_legalize
()
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:346
↓ 1 callersFunctionauto_tma_store_copy
(M, N, block_M, block_N, dtype, threads)
testing/python/language/test_tilelang_language_tma_store.py:103
↓ 1 callersFunctionbarrier_kernel
()
testing/python/language/test_tilelang_language_cluster.py:48
↓ 1 callersFunctionbefore
()
testing/python/transform/test_tilelang_transform_layout_inference.py:21
↓ 1 callersFunctionbench_tilelang
(M, N, K, block_M, block_N, block_K, warmup, repeats)
benchmark/matmul_metal/benchmark_matmul_metal.py:65
↓ 1 callersFunctionbench_torch_mps
(M, N, K, warmup, repeats)
benchmark/matmul_metal/benchmark_matmul_metal.py:58
↓ 1 callersFunctionbenchmark
(kernel, inputs, warmup=50, rep=100)
examples/grouped_gemm/example_grouped_gemm_fwd_ptr.py:104
↓ 1 callersFunctionbenchmark
(M=512, N=8192, K=8192, seed=0, warmup=25, rep=100)
examples/deepseek_v4/fp8_fp4_gemm_1d1d_sm100.py:373
↓ 1 callersFunctionbenchmark_compress
( m: int, k: int, dtype: torch.dtype, meta_dtype: torch.dtype | None = None, # noqa: FA100 )
benchmark/matmul/benchmark_matmul_sp_compress.py:32
↓ 1 callersFunctionbenchmark_fwd
( BATCH: int = 1, H: int = 8, N_CTX: int = 512, N_KV: int = 1024, D_HEAD: int = 128, T
examples/deepseek_v4/sparse_attn_fwd_sm90.py:225
↓ 1 callersFunctionbenchmark_topk_sparse_attention
()
benchmark/blocksparse_attention/benchmark_torch_block_sparse_fmha.py:29
↓ 1 callersFunctionbenchmark_topk_sparse_attention
()
benchmark/blocksparse_attention/benchmark_triton_block_sparse_fmha.py:254
← previousnext →1,801–1,900 of 7,254, ranked by callers