MCPcopy Create free account

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

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

↓ 1 callersFunctionmodule_has_tma
Check if any function in the module was lowered with TMA operations. This reads the ``tl.has_tma`` attribute set by ``LowerTileOp`` during ``
tilelang/cuda/pipeline.py:39
↓ 1 callersMethodmoe_infer
(self, x: torch.Tensor, flat_expert_indices: torch.Tensor, flat_expert_weights: torch.Tensor)
examples/fusedmoe/example_fusedmoe_torch.py:66
↓ 1 callersFunctionmqa_attn_return_logits
( IndexQ, IndexK, IndexKScale, Weights, CuSeqLenKS, CuSeqLenKE, block_N=256, n
examples/deepseek_v32/fp8_lighting_indexer.py:90
↓ 1 callersFunctionmulti_kernels_with_pdl
(N, block_size=256, dtype=T.float32)
testing/python/jit/test_tilelang_jit_tvm_ffi.py:455
↓ 1 callersFunctionmulti_kernels_with_pdl
(N, block_size=256, dtype=T.float32)
testing/python/jit/test_tilelang_jit_cython.py:16
↓ 1 callersFunctionmulti_kernels_with_pdl
(N, block_size=256, dtype=T.float32)
testing/python/jit/test_tilelang_jit_nvrtc.py:450
↓ 1 callersFunctionnaive_nsa_simple_inference
r""" Args: q (torch.Tensor): queries of shape `[B, 1, HQ, K]` if `head_first=False` else `[B, HQ, T, K]`. k (torch.Ten
examples/deepseek_nsa/reference.py:230
↓ 1 callersFunctionnative_sparse_attention_varlen
( Q, K, V, O_slc, BlockIndices, BlockCounts, Offsets, TokenIndices, batch,
examples/deepseek_nsa/example_tilelang_nsa_fwd_varlen.py:26
↓ 1 callersFunctionnested_continuous_parallels
(length=256, block=16, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:32
↓ 1 callersFunctionnested_continuous_ps
(length=256, block=16, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:283
↓ 1 callersFunctionnested_continuous_psp
(length=256, block1=8, block2=2, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:298
↓ 1 callersFunctionnested_continuous_serials
(length=256, block=16, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:211
↓ 1 callersFunctionnested_continuous_sp
(length=256, block=16, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:268
↓ 1 callersFunctionnested_continuous_sps
(length=256, block1=8, block2=2, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:314
↓ 1 callersFunctionnested_invalid_loop
(dtype: T.dtype = T.bfloat16, accum_dtype: T.dtype = T.float32, num_threads: int = 128)
testing/python/analysis/test_tilelang_fragment_loop_checker.py:29
↓ 1 callersFunctionnested_noncontinuous_parallels
(length=256, block=16, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:63
↓ 1 callersFunctionnested_noncontinuous_serials
(length=256, block=16, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:226
↓ 1 callersFunctionnested_triple_continuous_parallels
(length=256, block1=8, block2=2, dtype=T.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:47
↓ 1 callersFunctionno_set_max_nreg
Disable the maximum register limit setting.
tilelang/language/builtin.py:426
↓ 1 callersFunctionnormalize_flat_binds
Return ``stmt`` with dominating flat Bind values substituted into uses.
tilelang/transform/decouple_type_cast.py:272
↓ 1 callersFunctionnormalize_pass_configs
Canonicalize known pass-config keys and emit compatibility warnings.
tilelang/transform/pass_config.py:297
↓ 1 callersFunctionnormalize_prim_func
Normalize the primfunc to normal form
tilelang/carver/analysis.py:125
↓ 1 callersFunctionnormalize_to_matmul
(sch: Schedule, main_block: SBlockRV, layout: list[str] | None = None)
tilelang/carver/matmul_analysis.py:488
↓ 1 callersFunctionoob_store_legalize
(M: int = 64, N: int = 64, M_offset: int = 2, N_offset: int = 2)
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:167
↓ 1 callersFunctionpack_half2_impl
(x_val, y_val, *, loc=None, ip=None)
tilelang/contrib/cutedsl/utils.py:148
↓ 1 callersFunctionpack_rows_to_group_major_flat
(packed_rows)
examples/blockscaled_gemm_sm100/grouped_gemm_mxfp8_blockscaled_1d1d.py:437
↓ 1 callersFunctionpack_sf_u8_to_u32_rows
(sf_u8)
examples/blockscaled_gemm_sm100/grouped_gemm_mxfp8_blockscaled_1d1d.py:429
↓ 1 callersFunctionpack_sfb_to_group_major_flat
(packed_sfb)
examples/blockscaled_gemm_sm100/grouped_gemm_mxfp8_blockscaled_1d1d.py:441
↓ 1 callersFunctionpar_compile
Parallel compile multiple TileLang PrimFunc with TVM and build JITKernels. Parameters ---------- funcs : Iterable[tvm.tirx.PrimFunc]
tilelang/jit/__init__.py:123
↓ 1 callersFunctionparallel_elementwise_dynamic
(max_len=512, threads=256, dtype=T.float32)
testing/python/language/test_tilelang_language_parallel.py:25
↓ 1 callersFunctionparallel_elementwise_static
(length=256, dtype=T.float32)
testing/python/language/test_tilelang_language_parallel.py:11
↓ 1 callersFunctionparallel_nsa
r""" Args: q (torch.Tensor): queries of shape `[B, T, HQ, K]` if `head_first=False` else `[B, HQ, T, K]`. k (torch.Ten
examples/deepseek_nsa/example_triton_nsa_fwd_varlen.py:254
↓ 1 callersFunctionparallel_nsa
r""" Args: q (torch.Tensor): queries of shape `[B, T, HQ, K]` if `head_first=False` else `[B, HQ, T, K]`. k (torch.Ten
examples/deepseek_nsa/example_tilelang_nsa_fwd_varlen.py:227
↓ 1 callersFunctionparallel_nsa
r""" Args: q (torch.Tensor): queries of shape `[B, SEQLEN, HQ, K]` if `head_first=False` else `[B, HQ, SEQLEN, K]`. k
examples/deepseek_nsa/example_tilelang_nsa_bwd.py:702
↓ 1 callersFunctionparallel_nsa
r""" Args: q (torch.Tensor): queries of shape `[B, T, HQ, K]` if `head_first=False` else `[B, HQ, T, K]`. k (torch.Ten
examples/deepseek_nsa/example_triton_nsa_bwd.py:872
↓ 1 callersFunctionparallel_nsa
r""" Args: q (torch.Tensor): queries of shape `[B, T, HQ, K]` if `head_first=False` else `[B, HQ, T, K]`. k (torch.Ten
examples/deepseek_nsa/example_triton_nsa_fwd.py:235
↓ 1 callersFunctionparallel_nsa_block_mask
( block_indices: torch.LongTensor, block_counts: Union[torch.LongTensor, int], offsets: torch.Long
examples/deepseek_nsa/example_triton_nsa_bwd.py:668
↓ 1 callersFunctionparallel_nsa_bwd
( q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, o_slc: torch.Tensor, lse_slc: torch.T
examples/deepseek_nsa/example_tilelang_nsa_bwd.py:551
↓ 1 callersFunctionparallel_nsa_fwd
( q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, block_indices: torch.LongTensor, bloc
examples/deepseek_nsa/example_triton_nsa_fwd_varlen.py:156
↓ 1 callersFunctionparallel_nsa_fwd
( q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, block_indices: torch.LongTensor, bloc
examples/deepseek_nsa/example_tilelang_nsa_fwd_varlen.py:160
↓ 1 callersFunctionparallel_vectorized_cast_kernel
(M: int, dtype_A: str, dtype_B: str)
testing/python/language/test_tilelang_language_vectorized_cast.py:23
↓ 1 callersMethodparams_as_dict
Returns the template parameters as a dictionary. Returns: dict: Dictionary containing template parameter values.
tilelang/carver/template/gemv.py:117
↓ 1 callersMethodparams_as_dict
Returns the template parameters as a dictionary. Returns: dict: Dictionary containing template parameter values.
tilelang/carver/template/flashattention.py:134
↓ 1 callersMethodparams_as_dict
Returns the template parameters as a dictionary. Returns: dict: Dictionary containing template parameter values.
tilelang/carver/template/matmul.py:132
↓ 1 callersMethodparams_as_dict
(self)
tilelang/carver/template/general_reduce.py:111
↓ 1 callersMethodparams_as_dict
Returns the template parameters as a dictionary. Returns: dict: Dictionary containing template parameter values.
tilelang/carver/template/conv.py:165
↓ 1 callersMethodparams_as_dict
Returns the parameters of the template as a dictionary. Returns: dict: A dictionary containing shape and dtype.
tilelang/carver/template/elementwise.py:67
↓ 1 callersFunctionparse_args
()
examples/blockscaled_gemm_sm100/grouped_gemm_mxfp8_blockscaled_1d1d.py:580
↓ 1 callersFunctionparse_args
()
examples/blockscaled_gemm_sm100/gemm_mxfp8_blockscaled_1d1d.py:635
↓ 1 callersFunctionparse_args
()
examples/deepseek_v4/fp8_fp4_gemm_1d1d_sm100.py:386
↓ 1 callersFunctionparse_args
()
cmake/generate_windows_import_lib.py:13
↓ 1 callersFunctionparse_arguments
Parse command line arguments.
maint/precision/compare_ops.py:50
↓ 1 callersFunctionparse_expr
(s: str)
tilelang/autodd.py:81
↓ 1 callersMethodparse_source_information
(self)
tilelang/jit/adapter/wrapper.py:449
↓ 1 callersMethodparse_source_information
(self)
tilelang/jit/adapter/wrapper.py:843
↓ 1 callersFunctionparse_stmts
(s: str)
tilelang/autodd.py:76
↓ 1 callersFunctionper_block_cast_to_fp8
(x: torch.Tensor)
examples/deepseek_deepgemm/example_deepgemm_fp8_2xAcc.py:95
↓ 1 callersFunctionper_token_cast_to_fp8
(x: torch.Tensor)
examples/deepseek_deepgemm/example_deepgemm_fp8_2xAcc.py:87
↓ 1 callersFunctionper_token_group_quant_fp8
Function to perform per-token-group quantization on an input tensor `x`. It converts the tensor values into signed float8 values and returns the
examples/cast/example_triton_cast_to_fp8.py:111
↓ 1 callersFunctionper_warp_tma_copy_kernel
()
testing/python/language/test_tilelang_language_tma_copy.py:115
↓ 1 callersMethodplan_rasterization
(self, td: TileDict)
tilelang/carver/roller/policy/tensorcore.py:342
↓ 1 callersMethodplan_rasterization
Plans the rasterization for the given TileDict. This function is not implemented yet. Parameters ---------- td : Til
tilelang/carver/roller/policy/default.py:802
↓ 1 callersFunctionpointer_table_copy_test
(N, dtype=T.float16)
testing/python/language/test_tilelang_language_ptr.py:42
↓ 1 callersFunctionpointer_table_grouped_matmul_test
(batch_sizes_list, N, K, block_M, block_N, block_K, dtype=T.float16, accum_dtype=T.float32)
testing/python/language/test_tilelang_language_ptr.py:70
↓ 1 callersFunctionpool_mqa_attn_return_logits_fp8
( IndexQ, IndexBlockedK, IndexBlockedKScale, Logits, Weights, CuSeqLenBlockedKS, C
examples/dsa_hisa/pool_mqa_fp8.py:31
↓ 1 callersMethodpop
Pops and returns the top of the stack, or returns None if the stack is empty.
tilelang/language/kernel.py:42
↓ 1 callersFunctionpop_recorded
Return everything recorded since the last ``reset_recorder`` and clear the buffer.
tilelang/contrib/hip_resource_info.py:69
↓ 1 callersMethodpost_process_weights
(self)
examples/bitnet-1.58b/utils_quant.py:138
↓ 1 callersMethodpost_quant_process
(self, input, si, sw)
examples/bitnet-1.58b/utils_quant.py:166
↓ 1 callersFunctionpostprocess
( dKV, D, D_tail, kv_group=1, block_N=64, threads=128, dtype=T.bfloat16, accum
examples/deepseek_v32/sparse_mla_bwd.py:44
↓ 1 callersFunctionpostprocess
(o)
examples/linear_attention/example_retention_fwd.py:81
↓ 1 callersFunctionpostprocess
( dKV, D, D_tail, kv_group=1, block_N=64, threads=128, dtype=T.bfloat16, accum
examples/dsa_sparse_finetune/sparse_mla_bwd.py:49
↓ 1 callersFunctionpre_order_traverse
(block_analyzer, blocks, func)
tilelang/carver/roller/node.py:20
↓ 1 callersFunctionprecompute_freqs_cis
Precomputes frequency-based complex exponential values for rotary positional embeddings. Args: args (ModelArgs): Model arguments con
examples/deepseek_v32/inference/model.py:340
↓ 1 callersFunctionprelude_tma_bound_index
Pre-loop TMA load uses a scalar bind that is also consumed in the WS branch.
testing/python/transform/test_tilelang_transform_producer_consumer_ws.py:117
↓ 1 callersFunctionprelude_tma_wait_sink
A tiled-WS kernel with pre-loop TMA loads consumed at different points.
testing/python/transform/test_tilelang_transform_producer_consumer_ws.py:82
↓ 1 callersFunctionprepare_cu_seqlens_from_lens
( lens: torch.LongTensor, dtype: torch.dtype | None = torch.int32, )
examples/dsa_hisa/tilelang_utils.py:54
↓ 1 callersFunctionprepare_input
(B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, gate_dtype=torch.float32)
examples/gdn/example_wy_fast.py:24
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/gdn/example_wy_fast_bwd_split.py:51
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/gdn/example_chunk_delta_h.py:39
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/gdn/example_chunk_o.py:24
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/gdn/example_chunk_o_bwd.py:55
↓ 1 callersFunctionprepare_input
( B, S, H, DK, input_dtype, output_dtype, accum_dtype, )
examples/gdn/example_chunk_scaled_dot_kkt.py:25
↓ 1 callersFunctionprepare_input
( B, S, H, DK, chunk_size, sub_chunk_size, input_dtype, output_dtype, accu
examples/kda/chunk_inter_solve_fused.py:15
↓ 1 callersFunctionprepare_input
( B, S, H, DK, chunk_size, input_dtype, output_dtype, accum_dtype, gate_dt
examples/kda/chunk_intra_token_parallel.py:13
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/kda/wy_fast_bwd.py:17
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/kda/chunk_delta_bwd.py:16
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/kda/chunk_delta_h_fwd.py:23
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, gate_dtype, )
examples/kda/chunk_bwd_dqkwg.py:13
↓ 1 callersFunctionprepare_input
(B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, gate_dtype=torch.float32)
examples/kda/wy_fast.py:14
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/kda/chunk_o.py:13
↓ 1 callersFunctionprepare_input
( B, S, H, DK, chunk_size, input_dtype, output_dtype, accum_dtype, gate_dt
examples/kda/chunk_bwd_intra.py:16
↓ 1 callersFunctionprepare_input
( B, S, H, DV, chunk_size, input_dtype, do_dtype, )
examples/kda/chunk_bwd_gla_dA.py:13
↓ 1 callersFunctionprepare_input
( B, S, H, DK, DV, chunk_size, input_dtype, do_dtype, )
examples/kda/chunk_bwd_dv.py:14
↓ 1 callersFunctionprepare_lens
(cu_seqlens: torch.LongTensor)
examples/dsa_sparse_finetune/index.py:50
↓ 1 callersFunctionprepare_lens
(cu_seqlens: torch.LongTensor)
examples/kda/FLA_KDA/fla_utils.py:102
↓ 1 callersFunctionprepare_output
( B, S, H, DK, DV, chunk_size, output_dtype, gate_dtype, state_dtype, )
examples/kda/wy_fast_bwd.py:44
↓ 1 callersFunctionprepare_output
( B, S, H, DK, DV, chunk_size, output_dtype, gate_dtype, state_dtype, )
examples/kda/chunk_delta_bwd.py:46
↓ 1 callersFunctionprepare_output
( B, S, H, DK, DV, chunk_size, gate_dtype, )
examples/kda/chunk_bwd_dqkwg.py:37
↓ 1 callersFunctionprepare_output
( B, S, H, DK, chunk_size, NK, output_dtype, gate_dtype, state_dtype, )
examples/kda/chunk_bwd_intra.py:48
← previousnext →2,701–2,800 of 7,254, ranked by callers