MCPcopy Create free account

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

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

Methodparse_cache_key
(self, *args: _P.args, **kwargs: _P.kwargs)
tilelang/jit/__init__.py:422
Methodparse_macro
(self, parser: Parser)
tilelang/language/parser/entry.py:100
Functionpass_fn
(func: PrimFunc, mod, ctx)
tilelang/analysis/nested_loop_checker.py:115
Functionpass_fn
(func: PrimFunc, mod, ctx)
tilelang/analysis/ast_printer.py:123
Functionpass_fn
(func: tirx.PrimFunc, mod, ctx)
tilelang/analysis/layout_visual.py:104
Functionpass_fn
(func: PrimFunc, mod, ctx)
tilelang/analysis/fragment_loop_checker.py:111
Functionpass_fn
(func, mod, ctx)
tilelang/metal/transform/mark_host_metal_context.py:51
Functionpass_fn
(func: PrimFunc, mod, ctx)
tilelang/transform/decouple_type_cast.py:601
Functionpass_fn
(func: PrimFunc, mod, ctx)
tilelang/transform/add_bufstore_wrapper.py:19
Functionpass_fn
(func: PrimFunc, mod, ctx)
tilelang/transform/hoist_broadcast_values.py:90
Functionpdl_sync
()
tilelang/language/pdl.py:17
Functionpdl_trigger
()
tilelang/language/pdl.py:10
Functionper_custom_dims_cast_to_fp8
(x: torch.Tensor, dims: Tuple[int], use_ue8m0: bool)
examples/dsa_hisa/tilelang_utils.py:133
Functionplot_fragment_tv
Plot fragment in terms of thread and local index mapping. Parameters ---------- frag : T.Fragment The fragment object that de
tilelang/tools/plot_layout.py:616
Methodpolicy
(self)
tilelang/tileop/gemm/gemm_base.py:168
Methodpolicy
(self)
tilelang/tileop/gemm_sp/gemm_sp_base.py:136
Functionpopcount
Count the number of set bits in input x. Parameters ---------- x : PrimExpr Input argument. Returns ------- y : Prim
tilelang/language/tir/op.py:3091
Functionpost_visit
Post-order visitor that applies transformations and updates counters.
tilelang/transform/add_bufstore_wrapper.py:113
Functionpow
x power y Parameters ---------- x : PrimExpr Input argument. y : PrimExpr The exponent span : Optional[Span]
tilelang/language/tir/op.py:3067
Functionpower
x power y Parameters ---------- x : PrimExpr Input argument. y : PrimExpr The exponent span : Optional[Span]
tilelang/language/tir/op.py:3043
Functionpre_visit
Pre-order visitor that tracks thread bindings and tile operation depth.
tilelang/transform/add_bufstore_wrapper.py:102
Functionprefetch
The prefetch hint for a buffer. Parameters ---------- buffer : Buffer The buffer to be prefetched. bounds : List[Range]
tilelang/language/ast/ir.py:1300
Functionprefetch_tma_descriptor
Prefetch a TMA descriptor. Corresponds to PTX instruction: prefetch.tensormap;
tilelang/contrib/cutedsl/cpasync.py:282
Functionprepare_cu_seqlens_from_lens
( lens: torch.LongTensor, dtype: torch.dtype | None = torch.int32, )
examples/dsa_sparse_finetune/index.py:55
Functionprepare_cu_seqlens_from_position_ids
( position_ids: torch.LongTensor, dtype: torch.dtype | None = torch.int32, )
examples/dsa_hisa/tilelang_utils.py:85
Functionprepare_input_fake
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/gdn/example_wy_fast_bwd_split.py:27
Functionprepare_input_fake
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/gdn/example_chunk_delta_bwd.py:66
Functionprepare_input_fake
( B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, accum_dtype,
examples/gdn/example_chunk_o_bwd.py:29
Methodprepare_inputs_for_generation
( self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, cache_position=None,
examples/bitnet-1.58b/modeling_bitnet.py:1248
Functionprepare_ks_ke_from_cu_seqlens_qk
( cu_seqlens_q: torch.LongTensor, cu_seqlens_k: torch.LongTensor, )
examples/dsa_hisa/tilelang_utils.py:111
Functionprepare_lens_from_cu_seqlens
( cu_seqlens: torch.LongTensor, )
examples/dsa_sparse_finetune/index.py:63
Functionprepare_lens_from_cu_seqlens
( cu_seqlens: torch.LongTensor, )
examples/dsa_hisa/tilelang_utils.py:62
Functionprepare_wy_repr_bwd_kernel
( k, v, beta, gk, A, dA, dw, du, dk, dk2, dv, db, dg,
examples/kda/FLA_KDA/fla_wy_fast.py:114
Functionprim_buffer_slice_step
(A: T.Buffer((10,), T.int32), B: T.Buffer((5,), T.int32))
testing/python/language/test_tilelang_language_frontend_v2.py:434
Functionprim_call_macro
()
testing/python/language/test_tilelang_language_frontend_v2.py:348
Functionprim_func
The parsing method for tir prim func, by using `@prim_func` as decorator. Parameters ---------- func : Callable The function to b
tilelang/language/parser/entry.py:37
Functionprim_func
The primitive function statement. Parameters ---------- is_private : bool Whether the PrimFunc is annotated as private (i
tilelang/language/ast/ir.py:171
Functionprim_func
The parsing method for tir prim func, by using `@prim_func` as decorator. Parameters ---------- func : Callable The function to b
tilelang/language/tir/entry.py:10
Methodprim_func
(self)
tilelang/jit/adapter/wrapper.py:625
Methodprim_func
(self)
tilelang/jit/adapter/wrapper.py:896
Methodprim_func
Returns the primary TIR function from the IR module.
tilelang/jit/adapter/tvm_ffi.py:337
Methodprim_func
Returns the primary TIR function from the IR module.
tilelang/jit/adapter/cython/adapter.py:363
Methodprim_func
Returns the primary TIR function from the IR module.
tilelang/jit/adapter/nvrtc/adapter.py:289
Methodprim_func
Returns the primary TIR function from the IR module.
tilelang/jit/adapter/cutedsl/adapter.py:465
Functionprint_bit
Print the 32-bit binary representation of a CPU scalar extracted from a PyTorch tensor. Converts `val` to CPU, reads its Python scalar with
examples/dequantize_gemm/dequantize_utils.py:87
Methodprint_dependencies
(self)
tilelang/carver/roller/shape_inference/tir.py:127
Methodprint_source_code
Deprecated: use show_source() or export_sources() instead. Parameters ---------- which : Literal["kernel", "host", "
tilelang/jit/kernel.py:546
Functionprint_var_with_condition
Conditionally prints a TIR primitive expression (PrimExpr) if a given condition is True. Parameters: condition (tirx.PrimExpr): A TI
tilelang/language/print_op.py:30
Methodprio
(td: TileDict)
tilelang/carver/roller/policy/default.py:111
Functionprocess_func
Execute a single perf function and record its latency. `func` is expected to return a positive latency scalar (seconds or ms; we treat it as
tilelang/testing/perf_regression.py:52
Functionprofile
(model, input_data)
examples/bitnet-1.58b/benchmark_generate.py:51
Functionprofile
(model, input_data)
examples/bitnet-1.58b/eval_gpu_memory.py:12
Functionprofile
(model, input_data)
examples/bitnet-1.58b/benchmark_model_10k_loops.py:19
Functionprofile
(model, input_data)
examples/bitnet-1.58b/eval_correctness.py:49
Functionprogram
(Q: T.Tensor((M, N), dtype))
testing/python/debug/test_tilelang_debug_print.py:11
Functionprogram
()
testing/python/debug/test_device_assert.py:9
Methodprompts
Convenience method to define the prompt for each test image. The order of the returned prompts matches the order of the asse
examples/bitnet-1.58b/vllm_workspace/conftest.py:67
Functionptr
Create a TIR var that represents a pointer. Parameters ---------- dtype: DType The data type of the pointer. storage_scope:
tilelang/language/proxy.py:269
Functionptx_commit_group
TVM intrinsic for ptx async copy commit https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#data-movement-and-conversion-instruction
tilelang/language/tir/op.py:1598
Functionptx_cp_async
TVM intrinsic for ptx async copy from global to shared memory using cp.async https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#dat
tilelang/language/tir/op.py:1507
Functionptx_cp_async_barrier
TVM intrinsic for ptx async copy barrier using cp.async.mbarrier.arrive https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#parallel
tilelang/language/tir/op.py:1877
Functionptx_cp_async_bulk
TVM intrinsic for ptx async copy from global to shared memory using cp.async.bulk https://docs.nvidia.com/cuda/parallel-thread-execution/index.htm
tilelang/language/tir/op.py:1563
Functionptx_fence_barrier_init
TVM intrinsic for ptx fence barrier initialization. Returns ------- call : PrimExpr The call expression.
tilelang/language/tir/op.py:1914
Functionptx_init_barrier_thread_count
TVM intrinsic for ptx barrier initialization of thread count using mbarrier.init https://docs.nvidia.com/cuda/parallel-thread-execution/index.html
tilelang/language/tir/op.py:1894
Functionptx_ldmatrix
TileLang intrinsic for ptx load matrix from shared memory Uses `tl.ptx_ldmatrix` which expects access pointers created via `T.access_ptr` (i.
tilelang/language/tir/op.py:1470
Functionptx_ldmatrix_x1
Load 1 matrix (8x8) from shared memory
tilelang/contrib/cutedsl/ldsm.py:64
Functionptx_ldmatrix_x1_trans
Load 1 matrix (8x8) with transpose from shared memory
tilelang/contrib/cutedsl/ldsm.py:85
Functionptx_ldmatrix_x2
Load 2 matrices (8x8 each) from shared memory
tilelang/contrib/cutedsl/ldsm.py:73
Functionptx_ldmatrix_x2_trans
Load 2 matrices (8x8 each) with transpose from shared memory
tilelang/contrib/cutedsl/ldsm.py:93
Functionptx_ldmatrix_x4
Load 4 matrices (8x8 each) from shared memory
tilelang/contrib/cutedsl/ldsm.py:79
Functionptx_ldmatrix_x4_trans
Load 4 matrices (8x8 each) with transpose from shared memory
tilelang/contrib/cutedsl/ldsm.py:99
Functionptx_mma
Generic PTX MMA dispatcher. Dispatches to the appropriate specialized MMA function based on shape and data types.
tilelang/contrib/cutedsl/ptx_mma.py:303
Functionptx_mma
TVM intrinsic for ptx tensor core mma instructions https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#warp-level-matrix-instruction
tilelang/language/tir/op.py:887
Functionptx_mma_sm70
TVM intrinsic for ptx tensor core mma instructions on SM70 (Volta). This intrinsic provides SM70-specific MMA operations that support m16n16k4 sh
tilelang/language/builtin.py:1364
Functionptx_mma_sp
Generic PTX sparse MMA dispatcher. Dispatches to the appropriate specialized sparse MMA function based on shape and data types.
tilelang/contrib/cutedsl/ptx_mma.py:381
Functionptx_mma_sp
TVM intrinsic for sparse tensor core ptx instructions https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#warp-level-matrix-instruct
tilelang/language/tir/op.py:978
Functionptx_stmatrix_x1
Store 1 matrix (8x8) to shared memory
tilelang/contrib/cutedsl/ldsm.py:110
Functionptx_stmatrix_x1_trans
Store 1 matrix (8x8) with transpose to shared memory
tilelang/contrib/cutedsl/ldsm.py:128
Functionptx_stmatrix_x2
Store 2 matrices (8x8 each) to shared memory
tilelang/contrib/cutedsl/ldsm.py:116
Functionptx_stmatrix_x2_trans
Store 2 matrices (8x8 each) with transpose to shared memory
tilelang/contrib/cutedsl/ldsm.py:134
Functionptx_stmatrix_x4
Store 4 matrices (8x8 each) to shared memory
tilelang/contrib/cutedsl/ldsm.py:122
Functionptx_stmatrix_x4_trans
Store 4 matrices (8x8 each) with transpose to shared memory
tilelang/contrib/cutedsl/ldsm.py:140
Functionptx_tcgen05_mma_blockscaled_ss
TVM intrinsic for tcgen05.mma block-scaled (mxf8f6f4.block_scale) instructions. Block-scaled TCGEN05 is explicit-async and carries an explicit ``
tilelang/language/tir/op.py:1359
Functionptx_tcgen05_mma_ss
TVM intrinsic for tcgen05.mma shared-memory x shared-memory instructions. Expects 14 or 15 positional arguments: (kind_dtype, desc_a, A_offse
tilelang/language/tir/op.py:1246
Functionptx_tcgen05_mma_ts
TVM intrinsic for tcgen05.mma tensor-memory x shared-memory instructions. Expects 13 positional arguments: (kind_dtype, A_ptr, A_offset, desc
tilelang/language/tir/op.py:1315
Functionptx_wait_barrier
TVM intrinsic for ptx barrier wait using mbarrier.try_wait https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#parallel-synchronizat
tilelang/language/tir/op.py:1966
Functionptx_wait_group
TVM intrinsic for ptx async copy wait https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#data-movement-and-conversion-instructions-
tilelang/language/tir/op.py:1610
Functionptx_wgmma_rs
( dtype, wgmma_prefix, b_is_k_major, a_dtype_abbrv, b_dtype_abbrv, accum_dtype_abbrv,
tilelang/language/tir/op.py:1121
Functionptx_wgmma_sp_rs
( dtype, wgmma_prefix, b_is_k_major, a_dtype_abbrv, b_dtype_abbrv, accum_dtype_abbrv,
tilelang/language/tir/op.py:1203
Functionptx_wgmma_sp_ss
( dtype, wgmma_prefix, a_is_k_major, b_is_k_major, a_dtype_abbrv, b_dtype_abbrv, a
tilelang/language/tir/op.py:1158
Functionptx_wgmma_ss
TVM intrinsic for ptx tensor core wmma instructions https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#warp-level-matrix-instructio
tilelang/language/tir/op.py:1079
Functionpytest_addoption
(parser)
testing/conftest.py:34
Functionpytest_collection_modifyitems
When TILELANG_TARGET=cutedsl, annotate known-bad tests automatically.
examples/conftest.py:46
Functionpytest_collection_modifyitems
(config, items)
testing/conftest.py:43
Functionpytest_configure
(config: Config)
maint/scripts/pytest_cuda_scheduler.py:168
Functionpytest_configure_node
(node)
maint/scripts/pytest_cuda_scheduler.py:197
Functionpytest_runtest_protocol
(item, nextitem)
maint/scripts/pytest_cuda_scheduler.py:214
Functionpytest_terminal_summary
Ensure that at least one test is collected. Error out if all tests are skipped.
examples/conftest.py:62
Functionpytest_terminal_summary
Ensure that at least one test is collected. Error out if all tests are skipped.
testing/conftest.py:57
Functionpytest_xdist_make_scheduler
(config: Config, log)
maint/scripts/pytest_cuda_scheduler.py:204
← previousnext →5,301–5,400 of 7,254, ranked by callers