MCPcopy Create free account

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

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

Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor((M, N),
testing/python/jit/test_tilelang_jit_parcompile.py:34
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor((M, N),
testing/python/jit/test_tilelang_jit_tvm_ffi.py:30
Functionmain
( A: T.Tensor((M, K), dtype), B: T.Tensor((K, N), dtype), C: T.Tensor((M, N), accum_dt
testing/python/jit/test_tilelang_jit_nullptr.py:11
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor((M, N),
testing/python/jit/test_tilelang_jit_gemm_cython.py:30
Functionmain
Write one output tensor from one device kernel.
testing/python/jit/test_tilelang_jit_cutedsl_host_codegen.py:49
Functionmain
( A: T.Tensor((N,), dtype), B: T.Tensor((N,), dtype), C: T.Tensor((N,), dt
testing/python/jit/test_tilelang_jit_cython.py:18
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor((M, N),
testing/python/jit/test_tilelang_jit_gemm.py:31
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor((M, N),
testing/python/jit/test_tilelang_jit_callback.py:30
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor((M, N),
testing/python/jit/test_tilelang_jit_nvrtc.py:30
Functionmain
( A: T.Tensor((M, K), dtype), B: T.Tensor((K, N), dtype), C: T.Tensor((M, N), dtype),
testing/python/metal/test_metal_codegen_linux.py:15
Functionmain
( A: T.Tensor((M, K), dtype), B: T.Tensor((K, N), dtype), C: T.Tensor((M, N), accum_dt
testing/python/metal/test_metal_gemm_v2_linux.py:16
Functionmain
( A: T.Tensor((M, K), dtype), B: T.Tensor((K, N), dtype), C: T.Tensor((M, N), dtype),
testing/python/webgpu/test_webgpu_codegen.py:9
Functionmain
(global_tensor: T.Tensor[(length,), T.int32])
testing/python/runtime/test_tilelang_runtime_dynamic_shared_memory.py:15
Functionmain
( A_sparse: T.Tensor(A_sparse_shape, in_dtype), E: T.Tensor(E_shape, metadata_dtype),
testing/python/tilelibrary/test_tilelang_tilelibrary_gemm_sp.py:35
Functionmain
The compiled TVM function for block-level matrix multiplication. - We divide the entire (M, N) domain into blocks of shape
testing/python/autotune/test_tilelang_autotune_with_inputs.py:46
Functionmain
The compiled TVM function for block-level matrix multiplication. - We divide the entire (M, N) domain into blocks of shape
testing/python/autotune/test_tilelang_autotune.py:134
Functionmain
( A: T.Tensor((M, K), dtype), B: T.Tensor((N, K), dtype), C: T.Tensor((M, N), dtype),
testing/python/autotune/test_tilelang_autotune_do_not_specialize.py:41
Functionmain
( A: T.Buffer((M, K), dtype), B: T.Buffer((K, N), dtype), C: T.Buffer((M, N), dtype),
testing/python/transform/test_tilelang_transform_producer_consumer_ws.py:15
Functionmain
(A: T.Tensor((128,), T.float32), l: T.int32)
testing/python/transform/test_tilelang_transform_inject_assumes.py:14
Functionmain
( x: T.Tensor((128,), T.float32), y: T.Tensor((128,), T.float32), )
testing/python/transform/test_readonly_param_const_codegen.py:8
Functionmain
( A: T.Tensor((32,), dtype), B: T.Tensor((16,), dtype), C: T.Tensor((8,), dtype),
testing/python/transform/test_tilelang_transform_hoist_broadcast_values.py:15
Functionmain
( A: T.Tensor((M, N), dtype=dtype), )
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:85
Functionmain
( B: T.Tensor((K, N), dtype), )
testing/python/transform/test_tilelang_transform_layout_inference.py:23
Functionmain
( A: T.Tensor((1024,), dtype=T.bfloat16), B: T.Tensor((1024,), dtype=T.bfloat16),
testing/python/transform/test_tilelang_transform_decouple_type_cast.py:336
Functionmain
(a: T.Tensor[(n,), T.int32])
testing/python/transform/test_tilelang_transform_split_host_device.py:93
Functionmain
( Q: T.Tensor(shape, dtype), K: T.Tensor(shape, dtype), V: T.Tensor(shape,
testing/python/transform/test_tilelang_transform_config_index_bitwidth.py:100
Functionmain
( A: T.Tensor((M, N, vec_len), dtype=T.float32), )
testing/python/transform/test_tilelang_transform_legalize_vectorized_loop.py:12
Functionmain
( A: T.Tensor((M, N), dtype), B: T.Tensor((M, N), dtype), )
testing/python/math/test_math_fast_math.py:69
Functionmain
( A: T.Tensor((32, 32), T.float32), B: T.Tensor((32, 32), T.float32), )
testing/python/math/test_math_ieee_math.py:191
Functionmain
( A: T.Tensor((M * 2,), dtype=dtype_tl), B: T.Tensor((M * 2,), dtype=dtype_tl), C: T.T
testing/python/cuda/test_cuda_f32x2_intrinsics.py:43
Functionmain
( A: T.Tensor((M, K), in_dtype), B: T.Tensor((N, K), in_dtype), C: T.Tensor((M, N), ou
testing/python/kernel/test_tilelang_kernel_bf16_gemm_tcgen5_ts.py:21
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), Bias: T.Tensor(Bias_
testing/python/kernel/test_tilelang_kernel_gemv_simt.py:51
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor((M, N),
testing/python/kernel/test_tilelang_kernel_gemm.py:27
Functionmain
( A: T.Tensor((M, K), in_dtype), B: T.Tensor((N, K), in_dtype), C: T.Tensor((M, N), ou
testing/python/kernel/test_tilelang_kernel_gemm_sm75.py:10
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor((batch,
testing/python/kernel/test_tilelang_kernel_gemm_batched.py:28
Functionmain
( A: T.Tensor((M, K), in_dtype), B: T.Tensor((N, K), in_dtype), C: T.Tensor((M, N), ou
testing/python/kernel/test_tilelang_kernel_int8_gemm_tcgen5.py:15
Functionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, in_dtype), C: T.Tensor(C_shape,
testing/python/kernel/test_tilelang_kernel_gemm_simt.py:78
Functionmain
( A: T.Tensor((M, K), dtype), B: T.Tensor((K, N), dtype), C: T.Tensor((M, N), dtype),
testing/python/kernel/test_tilelang_kernel_gemm_with_stride.py:9
Functionmain
()
tilelang/testing/__init__.py:93
Functionmain_func
( A: T.Tensor((M, N), dtype), B: T.Tensor((M, N), dtype), C: T.Tensor((M,
testing/python/math/test_math_ieee_math.py:29
Functionmain_no_split
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/warp_specialize/example_warp_specialize_flashmla.py:21
Functionmain_no_split
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/deepseek_mla/example_mla_decode.py:121
Functionmain_no_split
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/deepseek_mla/example_mla_decode_ws.py:287
Functionmain_no_split
( Q: T.Tensor([batch, h_q, dv], dtype), Q_pe: T.Tensor([batch, h_q, dpe], dtype), KV:
examples/deepseek_mla/example_mla_decode_paged.py:134
Functionmain_no_split
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/deepseek_mla/amd/benchmark_mla_decode_amd_tilelang.py:139
Functionmain_no_split
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/deepseek_mla/experimental/example_mla_decode_kv_fp8.py:26
Functionmain_split
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, storage_dtype), Ct: T.Tensor((N
examples/dequantize_gemm/example_dequant_gemm_fp4_hopper.py:136
Functionmain_split
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/deepseek_mla/example_mla_decode.py:23
Functionmain_split
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/deepseek_mla/example_mla_decode_ws.py:37
Functionmain_split
( Q: T.Tensor([batch, h_q, dv], dtype), Q_pe: T.Tensor([batch, h_q, dpe], dtype), KV:
examples/deepseek_mla/example_mla_decode_paged.py:28
Functionmain_split
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/deepseek_mla/amd/benchmark_mla_decode_amd_tilelang.py:46
Functionmain_split_persistent
( Q: T.Tensor([batch, heads, dim], dtype), Q_pe: T.Tensor([batch, heads, pe_dim], dtype),
examples/deepseek_mla/example_mla_decode_persistent.py:27
Functionmake_blockscaled_gemm_layout
Build the TMEM store layout for the C accumulator of a block-scaled GEMM. Users must call ``T.annotate_layout({C_tmem: layout})`` with the return
tilelang/language/gemm_op.py:445
Functionmake_cutlass_metadata_layout
(buffer: tvm.tirx.Buffer, mma_dtype: str = T.float16, arch: str | None = None, **extra_args)
tilelang/layout/gemm_sp.py:150
Functionmake_dq_layout
(dQ)
examples/deepseek_nsa/example_tilelang_nsa_bwd.py:306
Functionmake_filled_tensor
(shape, value)
tilelang/contrib/cutedsl/utils.py:97
Functionmake_fully_replicated_layout_fragment
Create a fully replicated layout for a fragment buffer. A fully replicated fragment means all threads hold identical copies of the entir
tilelang/layout/swizzle.py:168
Functionmake_gemm_fragment_8x8
Create a standard 8x8 GEMM fragment layout for ldmatrix/stmatrix. This layout matches the warp-level matrix multiplication pattern used in t
tilelang/layout/swizzle.py:143
Functionmake_gemm_fragment_8x8_transposed
Create a transposed 8x8 GEMM fragment layout for ldmatrix/stmatrix. This layout is the transposed version of make_gemm_fragment_8x8, useful
tilelang/layout/swizzle.py:155
Functionmake_mfma_swizzle_layout
(shared_buf, vecSize=8)
tilelang/rocm/intrinsics/mfma_layout.py:186
Methodmake_mma_load_layout
Create a layout function for storing MMA results into a fragment buffer. This layout is used in conjunction with `inverse_mma_store_l
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:320
Methodmake_mma_load_layout
(self, local_buf: Buffer, matrix: str = "A")
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:481
Methodmake_mma_load_layout
Create a layout function for storing MMA results into a fragment buffer. This layout is used in conjunction with `inverse_mma_store_l
tilelang/cuda/intrinsics/macro/wgmma_macro_generator.py:718
Methodmake_mma_store_layout
Create a layout function for storing MMA results into a fragment buffer. This layout is used in conjunction with `inverse_mma_store_l
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:434
Methodmake_mma_store_layout
Create the TCGEN5 tensor-memory layout used to store MMA accumulators. Parameters ---------- tmem_buf : tir.Buffer
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:484
Methodmake_mma_store_layout
Create a layout function for storing MMA results into a fragment buffer. This layout is used in conjunction with `inverse_mma_store_l
tilelang/cuda/intrinsics/macro/wgmma_macro_generator.py:821
Functionmake_swizzle_layout
(shared_buf)
testing/python/kernel/test_tilelang_kernel_gemm_simt.py:13
Functionmake_tcgen05mma_swizzled_layout
(buffer: BufferLikeType, continuity: int = None, k_major: bool = True)
tilelang/layout/swizzle.py:86
Functionmake_tensor
( ptr: Var | PrimExpr, shape: ShapeType, dtype: DType = "float32", strides: tuple[PrimExpr, ...] | None =
tilelang/language/proxy.py:329
Functionmake_wgmma_swizzled_layout
(buffer: BufferLikeType, continuity: int = None, k_major: bool = True)
tilelang/layout/swizzle.py:78
Methodmatch
(self, node: ast.AST, parent: ast.AST, field: str, inside_list: bool)
tilelang/autodd.py:92
Methodmatch
(self, node: ast.AST, parent: ast.AST, field: str, inside_list: bool)
tilelang/autodd.py:112
Methodmatch
(self, node: ast.AST, parent: ast.AST, field: str, inside_list: bool)
tilelang/autodd.py:134
Methodmatch
(self, node: ast.AST, parent: ast.AST, field: str, inside_list: bool)
tilelang/autodd.py:149
Methodmatch
(self, node: ast.AST, parent: ast.AST, field: str, inside_list: bool)
tilelang/autodd.py:175
Methodmatch
(self, node: ast.AST, parent: ast.AST, field: str, inside_list: bool)
tilelang/autodd.py:193
Methodmatch
(self, node: ast.AST, parent: ast.AST, field: str, inside_list: bool)
tilelang/autodd.py:369
Functionmatch_all_sync
Return ``mask`` if all lanes in ``mask`` agree on ``value``, else 0. Lowers to ``__match_all_sync`` on CUDA (compute capability >= 7.0); the
tilelang/language/builtin.py:1139
Functionmatch_any_sync
Return a ``uint32`` bitmask of lanes in ``mask`` whose ``value`` equals the calling lane's value. Lowers to ``__match_any_sync`` on CUDA (comp
tilelang/language/builtin.py:1128
Functionmatch_declare_kernel_cutedsl
(source: str, annotation: str = "@cute.kernel")
tilelang/jit/adapter/utils.py:40
Functionmatch_global_kernel
(source: str, annotation: str = "__global__")
tilelang/jit/adapter/utils.py:20
Functionmatmul
(A, B, block_M: int, block_N: int, block_K: int)
examples/quickstart.py:9
Functionmatmul
(A, B, block_M, block_N, block_K, dtype=T.float16, accum_dtype=T.float32)
examples/gemm/example_gemm.py:6
Functionmatmul
( A, B, block_M, block_N, block_K, trans_A, trans_B, in_dtype, out_dtype,
examples/gemm_fp8/example_tilelang_gemm_fp8_sm100.py:14
Functionmatmul_kernel_jit
( M, N, K, block_M, block_N, block_K, trans_A=False, trans_B=True, in_dtyp
testing/python/jit/test_tilelang_jit_parcompile.py:11
Functionmatmul_kernel_with_barrier
( M, N, K, block_M, block_N, block_K, mbars, trans_A, trans_B, in_dtyp
testing/python/jit/test_tilelang_jit_cutedsl.py:114
Functionmatmul_non_persistent
(A, B, block_M, block_N, block_K, threads, num_stages, dtype=T.float16, accum_dtype=T.float32)
examples/gemm/example_gemm_persistent.py:8
Functionmatmul_persistent
( A, B, block_M, block_N, block_K, threads, num_stages, dtype=T.float16, accum_dtype=T.float32, use_persis
examples/gemm/example_gemm_persistent.py:36
Functionmax2
Packed element-wise maximum.
tilelang/language/math_intrinsics.py:391
Methodmax_gen_toks
(self)
examples/bitnet-1.58b/eval_utils.py:89
Methodmax_length
(self)
examples/bitnet-1.58b/eval_utils.py:67
Functionmax_value
maximum value of dtype Parameters ---------- dtype : str The data type. span : Optional[Span] The location of this o
tilelang/language/tir/op.py:2158
Methodmaybe_contiguous
(x)
examples/grouped_gemm/example_grouped_gemm_bwd.py:100
Methodmaybe_contiguous
(x)
examples/flash_attention/example_mha_bwd_bhsd.py:263
Methodmaybe_contiguous
(x)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:373
Methodmaybe_contiguous
(x)
examples/flash_attention/example_gqa_bwd.py:363
Methodmaybe_contiguous
(x)
examples/flash_attention/example_mha_bwd_bshd.py:260
Methodmaybe_contiguous
(x)
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:527
Methodmaybe_contiguous
(x)
examples/attention_sink/example_gqa_sink_bwd_varlen.py:506
Methodmaybe_contiguous
(x)
examples/attention_sink/example_mha_sink_bwd_bhsd.py:356
← previousnext →5,101–5,200 of 7,254, ranked by callers