MCPcopy Create free account

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

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

Functionscalar_check
(x: T.int32, flag: T.bool())
maint/host_checks/common.py:38
Methodscan
The scanning block axis defining function. Parameters ---------- dom : Union[Range, List[PrimExpr], Tuple[PrimExpr]]
tilelang/language/ast/ir.py:592
Methodscore_block_size
(self, n)
tilelang/carver/roller/policy/tensorcore.py:229
Methodscore_block_size
Scores a block size based on its efficiency and fit relative to the architecture's warp size and SM partition. Parameters --
tilelang/carver/roller/policy/default.py:202
Functionserial
The serial For statement. Parameters ---------- start : PrimExpr The minimum value of iteration. stop : PrimExpr The
tilelang/language/ast/ir.py:676
Functionserial
The serial For statement. Parameters ---------- start : PrimExpr The minimum value of iteration. stop : PrimExpr The
tilelang/language/tir/ir.py:10
Methodset_decoder
(self, decoder)
examples/bitnet-1.58b/modeling_bitnet.py:1154
Methodset_fileline
(self, filename: str, lineno: int, name: str)
testing/python/language/test_tilelang_language_eager_ast_signature.py:10
Methodset_fileline
(self, filename: str, lineno: int, name: str)
tilelang/language/eager/builder.py:760
Methodset_input_embeddings
(self, value)
examples/bitnet-1.58b/modeling_bitnet.py:980
Methodset_input_embeddings
(self, value)
examples/bitnet-1.58b/modeling_bitnet.py:1145
Methodset_input_embeddings
(self, value)
examples/bitnet-1.58b/modeling_bitnet.py:1495
Methodset_input_embeddings
(self, value)
examples/bitnet-1.58b/modeling_bitnet.py:1612
Methodset_lib_path
(self, libpath)
tilelang/jit/adapter/libgen.py:213
Methodset_output_embeddings
(self, new_embeddings)
examples/bitnet-1.58b/modeling_bitnet.py:1151
Methodset_outputs
(self, i: int, edge: Edge)
tilelang/carver/roller/node.py:140
Functionset_random_seed
(seed: int = 42)
tilelang/testing/__init__.py:98
Functionset_seed
(seed)
examples/bitnet-1.58b/eval_utils.py:11
Methodset_src_path
(self, srcpath)
tilelang/jit/adapter/libgen.py:216
Functionsetup_module_env
Setup and restore module-level environment and cache state.
testing/python/cache/test_tilelang_kernel_cache.py:79
Methodsf_k_start
(self)
tilelang/tileop/gemm/gemm_base.py:196
Methodsf_k_start
(self)
tilelang/tileop/gemm/__init__.py:118
Methodshape_equal
(a, b)
tilelang/autotuner/tuner.py:750
Functionshared_16x16_to_ldmatrix_64x4_layout
(ind)
tilelang/rocm/intrinsics/mfma_layout.py:35
Functionshared_16x16_to_local_32x8_layout_A_colmajor_gfx12
Forward: A_T[i=K, j=M] -> (thread=(i//8)*16+j, local=i%8).
tilelang/rocm/intrinsics/wmma_layout.py:83
Functionshared_16x16_to_local_32x8_layout_A_gfx12
Forward: A[i=M, j=K] -> (thread=(j//8)*16+i, local=j%8).
tilelang/rocm/intrinsics/wmma_layout.py:65
Functionshared_16x16_to_local_32x8_layout_B_colmajor_gfx12
Forward: B_T[i=N, j=K] -> (thread=(j//8)*16+i, local=j%8).
tilelang/rocm/intrinsics/wmma_layout.py:119
Functionshared_16x16_to_local_32x8_layout_B_gfx12
Forward: B[i=K, j=N] -> (thread=(i//8)*16+j, local=i%8).
tilelang/rocm/intrinsics/wmma_layout.py:101
Functionshared_16x16_to_local_32x8_layout_C_gfx11
Forward: D[i=M, j=N] -> (thread=(i%2)*16+j, local=i//2).
tilelang/rocm/intrinsics/wmma_layout.py:256
Functionshared_16x16_to_local_32x8_layout_C_gfx12
Forward: D[i=M, j=N] -> (thread=(i//8)*16+j, local=i%8).
tilelang/rocm/intrinsics/wmma_layout.py:137
Functionshared_16x16_to_local_64x4_layout_A
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:47
Functionshared_16x16_to_local_64x4_layout_B
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:59
Functionshared_16x16_to_mma_32x8_layout
(i, j)
tilelang/cuda/intrinsics/layout/utils.py:67
Functionshared_16x16_to_mma_32x8_smoothlayout
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:191
Functionshared_16x16_to_mma_a_32x8_layout_trans
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:87
Functionshared_16x16_to_mma_b_32x8_layout_trans
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:96
Functionshared_16x16_to_mma_sp_layout_sr_a
(i, j)
tilelang/cuda/intrinsics/layout/mma_sp_layout.py:14
Functionshared_16x16_to_mma_sp_layout_sr_b
(i, j)
tilelang/cuda/intrinsics/layout/mma_sp_layout.py:18
Functionshared_16x32_to_local_64x8_layout_A
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:89
Functionshared_16x32_to_local_64x8_layout_B
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:101
Functionshared_16x32_to_mma_32x16_layout
(i, j)
tilelang/cuda/intrinsics/layout/utils.py:72
Functionshared_16x32_to_mma_32x16_smoothlayout
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:195
Functionshared_16x32_to_mma_sp_layout_sr_a
(i, j)
tilelang/cuda/intrinsics/layout/mma_sp_layout.py:23
Functionshared_16x32_to_mma_sp_layout_sr_b
(i, j)
tilelang/cuda/intrinsics/layout/mma_sp_layout.py:27
Functionshared_16x4_to_local_64x1_layout_A
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:7
Functionshared_16x4_to_mma_a_32x4_layout
(row, col, rep)
tilelang/cuda/intrinsics/layout/mma_sm70_layout.py:1
Functionshared_16x4_to_mma_b_32x4_layout_trans
(row, col, rep)
tilelang/cuda/intrinsics/layout/mma_sm70_layout.py:13
Functionshared_16x64_to_local_64x16_layout_A
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:113
Functionshared_16x64_to_local_64x16_layout_B
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:125
Functionshared_16x64_to_mma_sp_layout_sr_a
(i, j)
tilelang/cuda/intrinsics/layout/mma_sp_layout.py:32
Functionshared_16x64_to_mma_sp_layout_sr_b
(i, j)
tilelang/cuda/intrinsics/layout/mma_sp_layout.py:36
Functionshared_16x8_to_mma_a_32x4_layout_trans
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:62
Functionshared_16x8_to_mma_b_32x4_layout_trans
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:72
Functionshared_32x16_to_mma_32x16_layout
(i, j)
tilelang/carver/matmul_analysis.py:749
Functionshared_32x16_to_mma_32x16_layout
(i, j)
tilelang/cuda/intrinsics/layout/utils.py:77
Functionshared_32x16_to_mma_32x16_smoothlayout
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:199
Functionshared_32x16_to_mma_a_32x16_layout_trans
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:111
Functionshared_32x16_to_mma_b_32x16_layout_trans
(i, j)
tilelang/cuda/intrinsics/layout/mma_layout.py:120
Functionshared_32x32_to_local_64x16_layout_A
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:162
Functionshared_32x32_to_local_64x16_layout_B
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:174
Functionshared_32x32_to_local_64x16_layout_C
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:131
Functionshared_32x8_to_mma_32x8_layout
(i, j)
tilelang/carver/matmul_analysis.py:744
Functionshared_4x16_to_local_64x1_layout_B
(i, j)
tilelang/rocm/intrinsics/mfma_layout.py:18
Functionshared_4x16_to_mma_b_32x4_layout
(row, col, rep)
tilelang/cuda/intrinsics/layout/mma_sm70_layout.py:7
Functionshared_kernel
()
testing/python/issue/test_tilelang_issue_1026.py:9
Functionshfl_down
Shift ``value`` down by ``delta`` lanes (``__shfl_down_sync`` on CUDA, ``__shfl_down`` on HIP).
tilelang/language/builtin.py:930
Functionshfl_sync
Broadcast ``value`` from ``srcLane`` to all lanes in the subgroup of ``width`` lanes (``__shfl_sync`` on CUDA, ``__shfl`` on HIP — mask ignored
tilelang/language/builtin.py:1000
Functionshfl_up
Shift ``value`` up by ``delta`` lanes (``__shfl_up_sync`` on CUDA, ``__shfl_up`` on HIP).
tilelang/language/builtin.py:942
Functionshfl_xor
XOR-swap ``value`` across lanes (``__shfl_xor_sync`` on CUDA, ``__shfl_xor`` on HIP — mask ignored on HIP).
tilelang/language/builtin.py:918
Functionshift_left
Return the result of x left shifted by y bits. Parameters ---------- x : PrimExpr Input argument. y : PrimExpr Input
tilelang/language/tir/op.py:3172
Functionshift_right
Return the result of x right shifted by y bits. Parameters ---------- x : PrimExpr Input argument. y : PrimExpr Inpu
tilelang/language/tir/op.py:3191
Functionshould_do_global_cleanup_after_test
Allow subdirectories to skip global cleanup by overriding this fixture. This can provide a ~10x speedup for non-GPU unit tests since they don't ne
examples/bitnet-1.58b/vllm_workspace/conftest.py:92
Methodshow_ptx
Print compiled PTX for the kernel (CUDA only). Examples -------- >>> jit_kernel.show_ptx()
tilelang/jit/kernel.py:726
Methodshow_sass
Print disassembled SASS for the kernel (CUDA only). Examples -------- >>> jit_kernel.show_sass()
tilelang/jit/kernel.py:788
Functionshuffle_elect
(thread_extent)
tilelang/contrib/cutedsl/utils.py:122
Functionshuffle_elect
Elect exactly one lane within a logical thread group. Parameters ---------- thread_extent : int Size (in threads) of the group in
tilelang/language/builtin.py:677
Functionshuffle_elect_kernel
(A: T.Tensor((num_threads,), T.int32))
testing/python/language/test_tilelang_language_get_warp_info.py:75
Functionsigmoid
Quick function to get sigmoid Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr
tilelang/language/tir/op.py:2305
Methodsim
(a: int, b: int)
tilelang/carver/roller/policy/default.py:290
Functionsimple
(A: T.Tensor((128,), T.float32), B: T.Tensor((128,), T.float32))
testing/python/cache/test_tilelang_kernel_cache.py:281
Functionsimple_dequant_bf16_fp4
(B_shared, B_dequantize_shared, Scale, k)
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_cdna4.py:177
Functionsimple_dequant_bf16_fp4
Dequantizes a packed 4-bit (FP4) block from B_shared into BF16 values in B_dequantize_shared using per-element scale exponents.
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_hopper.py:280
Functionsimple_dequant_bf16_fp4
Dequantize a packed FP4 uint8 shared buffer into BF16 and store the result into a shared dequantized buffer. This helper:
examples/dequantize_gemm/example_dequant_gemm_bf16_fp4_hopper.py:263
Functionsimplify_prim_func
(func: Callable)
tilelang/transform/simplify.py:53
Functionsin
(x: Union[TensorSSA, Numeric], fastmath: bool = False, **kwargs)
tilelang/contrib/cutedsl/math.py:110
Functionsin
Take sin of input x. Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr The res
tilelang/language/tir/op.py:2465
Methodsin_cached
(self)
examples/bitnet-1.58b/modeling_bitnet.py:128
Functionsinh
Take sinh of input x. Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr The re
tilelang/language/tir/op.py:2481
Functionsinkhorn_bwd_configs
Generate autotune configurations for different tilesize and threads
examples/deepseek_mhc/example_mhc_bwd.py:32
Functionsinkhorn_bwd_implicit_cg
(out, dout, tilesize: int = 32, threads: int = 128)
examples/deepseek_mhc/example_mhc_bwd.py:85
Methodsize
Return the number of items in the stack. Returns: int: The size of the stack.
tilelang/autotuner/capture.py:53
Methodskip_kernel_ctx
(self)
tilelang/language/eager/builder.py:769
Methodsmooth_a
(self)
tilelang/carver/roller/hint.py:138
Methodsmooth_b
(self)
tilelang/carver/roller/hint.py:142
Methodsource
(self)
tilelang/language/eager/builder.py:801
Functionsparse_attention_mask_copy
( Q: T.Tensor(q_shape, dtype), K: T.Tensor(kv_shape, dtype), BlockSparseMask: T.Tensor
testing/python/issue/test_tilelang_issue_tma_no_ws.py:536
Functionsparse_flash_attn
( Q: T.Tensor(shape, dtype), K: T.Tensor(shape, dtype), V: T.Tensor(shape, dtype),
testing/python/issue/test_tilelang_issue_tma_no_ws.py:424
Functionsparse_gqa_decode_varlen_indice
Args: query: [batch, heads, dim] key: [batch, max_cache_seqlen, heads_kv, dim] value: [batch, max_cache_seqlen, heads_kv,
examples/blocksparse_attention/example_tilelang_sparse_gqa_decode_varlen_indice.py:211
Functionsparse_gqa_decode_varlen_mask
Args: query: [batch, heads, dim] key: [batch, max_cache_seqlen, heads_kv, dim] value: [batch, max_cache_seqlen, heads_kv,
examples/blocksparse_attention/example_tilelang_sparse_gqa_decode_varlen_mask.py:203
Functionsparse_tensorcore_metadata_copy
( A_sparse: T.Tensor((M, K // 2), T.float16), E: T.Tensor((M, K // e_factor), T.uint8),
testing/python/issue/test_tilelang_issue_tma_no_ws.py:373
← previousnext →5,701–5,800 of 7,254, ranked by callers