MCPcopy Create free account

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

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

↓ 4 callersFunction_fp4_tma_stack_int
(block, index)
testing/python/language/test_tilelang_language_tma_copy.py:338
↓ 4 callersFunction_is_int_type
Check if a value is an integer Numeric type.
tilelang/contrib/cutedsl/reduce.py:36
↓ 4 callersFunction_is_linux_like
()
tilelang/contrib/cc.py:33
↓ 4 callersFunction_is_windows_like
()
tilelang/contrib/cc.py:37
↓ 4 callersFunction_ldmatrix
Internal helper for ldmatrix operations
tilelang/contrib/cutedsl/ldsm.py:40
↓ 4 callersMethod_load_binary
(path: str)
tilelang/cache/kernel_cache.py:443
↓ 4 callersFunction_lower_cutedsl
Lower a TileLang program and build its CuTeDSL host wrapper.
testing/python/jit/test_tilelang_jit_cutedsl_host_codegen.py:21
↓ 4 callersFunction_make_auto_vec_binary_kernel
Build a kernel that uses T.Parallel to let the vectoriser emit tl::<op>2.
testing/python/cuda/test_cuda_f32x2_intrinsics.py:110
↓ 4 callersFunction_make_fake_kernel
(tmp_path)
testing/python/cache/test_tilelang_kernel_cache_atomic_save.py:41
↓ 4 callersFunction_make_result
(tmp_path, execution_backend: str = "cython")
testing/python/autotune/test_tilelang_autotune_atomic_save.py:55
↓ 4 callersFunction_matmul
(A, B)
testing/python/tilelibrary/test_tilelang_tilelibrary_gemm_sp.py:112
↓ 4 callersFunction_normalize_for_json
(value)
tilelang/cache/__init__.py:32
↓ 4 callersFunction_normalize_tensor_dtype
(dtype: DType)
tilelang/language/proxy.py:74
↓ 4 callersFunction_normalize_value
(value, sort_dict_items: bool = False)
tilelang/autotuner/tuner.py:202
↓ 4 callersFunction_require_cuda_tensor
(shape, dtype=torch.float32)
testing/python/analysis/test_tilelang_nested_loop_checker.py:10
↓ 4 callersFunction_requirement_spec
()
tilelang/jit/adapter/cutedsl/checks.py:30
↓ 4 callersFunction_resolve_warp_size
(warp_size: Optional[int])
testing/python/language/test_tilelang_language_get_warp_info.py:12
↓ 4 callersFunction_run_and_check
(kernel, n)
testing/python/runtime/test_tilelang_runtime_dynamic_shared_memory.py:34
↓ 4 callersFunction_run_version_command
Return the first non-empty version command line, or a readable failure.
testing/python/language/test_tilelang_language_pdl.py:179
↓ 4 callersFunction_test_infinity
(dtype: str)
testing/python/language/test_tilelang_language_infinity.py:17
↓ 4 callersFunction_wgmma_ab_dtype
Map A/B operand dtype for WGMMA: fp32 -> tf32 (SM90 compatibility).
tilelang/contrib/cutedsl/gemm_v2.py:118
↓ 4 callersFunctionadd
(path: pathlib.Path)
cmake/find_pip_cuda.py:90
↓ 4 callersFunctionadd_import_lib
(lib_path: str | None)
tilelang/contrib/msvc.py:293
↓ 4 callersFunctionalign_up
(xv, yv)
maint/gemm/correctness_evaluation_tcgen05_fp4_unpacked.py:128
↓ 4 callersFunctionalloc_fragment
Allocate a fragment memory buffer for specialized operations. Args: shape (tuple): The shape of the buffer to allocate dtype (str
tilelang/language/allocate.py:66
↓ 4 callersFunctionallow_vectorize
(pass_ctx: PassContext | None = None)
tilelang/backend/pass_pipeline/pipeline_utils.py:10
↓ 4 callersFunctionapply_rotary_emb
Applies rotary positional embeddings to the input tensor. Args: x (torch.Tensor): Input tensor with positional embeddings to be appl
examples/deepseek_v32/inference/model.py:421
↓ 4 callersMethodarg
(self, name: str, value: Any)
tilelang/language/eager/ast.py:241
↓ 4 callersFunctionassert_gemm_v2
( M, N, K, block_M, block_N, block_K, dtype=T.float16, accum_dtype=T.float32,
testing/python/metal/test_metal_gemm_v2.py:40
↓ 4 callersFunctionassert_metal_gemm_v2_codegen
( M, N, K, block_M, block_N, block_K, dtype=T.float16, accum_dtype=T.float32,
testing/python/metal/test_metal_gemm_v2_linux.py:39
↓ 4 callersFunctionassert_simdgroup_store_codegen
(M, N, K, block_M, block_N, block_K, dtype=T.float16, accum_dtype=T.float32)
testing/python/metal/test_metal_simdgroup_store.py:59
↓ 4 callersMethodbackward
(ctx, do_slc, do_swa)
examples/deepseek_nsa/example_triton_nsa_bwd.py:139
↓ 4 callersMethodbackward
( ctx, do: torch.Tensor, _1: torch.Tensor, )
examples/dsa_sparse_finetune/dsa.py:138
↓ 4 callersMethodbackward
(ctx, do)
examples/flash_attention/example_mha_bwd_bhsd.py:259
↓ 4 callersMethodbackward
(ctx, do)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:364
↓ 4 callersMethodbackward
(ctx, do)
examples/flash_attention/example_gqa_bwd.py:354
↓ 4 callersMethodbackward
(ctx, do)
examples/flash_attention/example_mha_bwd_bshd.py:256
↓ 4 callersMethodbackward
(ctx, do)
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:517
↓ 4 callersMethodbackward
(ctx, do)
examples/attention_sink/example_gqa_sink_bwd_varlen.py:547
↓ 4 callersMethodbackward
(ctx, do)
examples/attention_sink/example_gqa_sink_bwd_bhsd.py:351
↓ 4 callersFunctionbar_sync_ptx
(barrier_id, number_of_threads)
tilelang/contrib/cutedsl/reduce.py:188
↓ 4 callersFunctionbits_product
Compute the number of bits in a Buffer (shape with dtype). For vector types (e.g. ``bfloat16x2``) ``DataType.bits`` returns the per-lane
tilelang/utils/language.py:448
↓ 4 callersFunctionblocksparse_flashattn
(batch, heads, seq_q, seq_kv, dim, downsample_len, is_causal)
examples/seer_attention/block_sparse_attn_tilelang.py:35
↓ 4 callersFunctionbuffer_pair
(shape=(16,))
testing/python/transform/test_tilelang_transform_simplify.py:15
↓ 4 callersFunctioncheck
(x, ty)
testing/python/language/test_tilelang_language_frontend_v2.py:241
↓ 4 callersFunctioncheck_expr
(expr)
testing/python/arith/test_arith_hard.py:19
↓ 4 callersFunctioncheck_hip_availability
Check if HIP (ROCm) is available on the system by locating the ROCm path. Returns: bool: True if HIP is available, False otherwise.
tilelang/utils/target.py:131
↓ 4 callersMethodchunk
(self)
tilelang/tileop/gemm/gemm_base.py:108
↓ 4 callersFunctionclean_and_maintain_logits_interface
In-place: applies +inf/-inf mask based on per-row [ks, ke).
examples/dsa_hisa/clean_and_maintain_logits.py:40
↓ 4 callersFunctioncoalesced_factor
(subtensor: list[int], tensor: list[int])
tilelang/carver/roller/policy/common.py:32
↓ 4 callersFunctioncompile
Compile the given TileLang PrimFunc with TVM and build a JITKernel. Parameters ---------- func : tvm.tirx.PrimFunc, optional
tilelang/jit/__init__.py:43
↓ 4 callersMethodcompute_tcgen05_b_desc_params
Compute B descriptor parameters from the B shared buffer. This is a pure-Python helper -- no TIR code is emitted. The returned ``TCGE
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:663
↓ 4 callersMethoddisable
Disable kernel caching globally.
tilelang/env.py:192
↓ 4 callersFunctiondisplay_error_message
(msg)
examples/deepseek_v32/fp8_lighting_indexer.py:9
↓ 4 callersFunctiondo_bench
Do benchmark for a function.
examples/kda/chunk_o.py:144
↓ 4 callersFunctionevaluate_gemv_simt
( M: int, N: int, K: int, in_dtype: str, out_dtype: str, accum_dtype: str, trans_A
testing/python/kernel/test_tilelang_kernel_gemv_simt.py:115
↓ 4 callersFunctionextract_tensormap_ptr
extract the tensormap pointer from a TMA Copy Atom. :param tma_atom: The TMA Copy Atom :type tma_atom: CopyAtom
tilelang/contrib/cutedsl/cpasync.py:78
↓ 4 callersMethodextract_thread_binding
Return (lane_id, warp_n, warp_m).
tilelang/rocm/intrinsics/wmma_macro_generator.py:133
↓ 4 callersMethodfrom_bit_linear
(cls, bitlinear, weight_group=1)
examples/bitnet-1.58b/utils_quant.py:100
↓ 4 callersMethodfrom_output_nodes
(cls, nodes: list[OutputNode], arch: TileDevice, tags: dict | None = None)
tilelang/carver/roller/policy/default.py:44
↓ 4 callersFunctiongen_inputs
(B, H, Sq, Skv, D, dtype=torch.float16)
examples/attention_sink/example_mha_sink_fwd_bhsd.py:183
↓ 4 callersFunctiongenerate_dense_input
(M, N, K, trans_A, trans_B, in_dtype, seed=0)
testing/python/tilelibrary/test_tilelang_tilelibrary_gemm_sp.py:133
↓ 4 callersFunctionget_cuda_device_properties
(device_id: int = 0)
tilelang/carver/arch/driver/cuda_driver.py:22
↓ 4 callersFunctionget_device_func
Get the device kernel function from the module.
testing/python/transform/test_tilelang_transform_split_host_device.py:22
↓ 4 callersFunctionget_env_path
(compiler_env: dict[str, str])
tilelang/contrib/msvc.py:22
↓ 4 callersMethodget_ldmatrix_index_map
(self, is_b=False)
tilelang/rocm/intrinsics/mfma_macro_generator.py:236
↓ 4 callersFunctionget_let_value
Get the value bound to a variable in the current let frame stack. Args: var (Var): The variable to look up Returns: Optional
tilelang/language/frame.py:221
↓ 4 callersFunctionget_param_by_name
Get a parameter by name_hint.
testing/python/transform/test_tilelang_transform_split_host_device.py:72
↓ 4 callersFunctionget_sparse_attn_mask_from_topk
(x, topk, use_dense_for_last_block=False)
examples/seer_attention/block_sparse_attn_tilelang.py:9
↓ 4 callersMethodget_spatial_axis
(self, block: SBlockRV)
tilelang/carver/roller/node.py:49
↓ 4 callersFunctionget_swizzled_block_idx
(tile_id, group_size, m_clusters, cta_id)
examples/gemm_sm100/gemm_tcgen5mma_ws_clc.py:7
↓ 4 callersMethodget_thread_binding
(self)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:181
↓ 4 callersMethodget_thread_extent
Returns the thread extent for the given dimension. dim=0 corresponds to threadIdx.x, dim=1 to threadIdx.y, and dim=2 to threadIdx.z.
tilelang/language/kernel.py:195
↓ 4 callersMethodget_tmp
(self)
tilelang/language/eager/ast.py:264
↓ 4 callersFunctiongroup_per_split_token_cast_to_fp8
(X, batch_sizes, M_max, blk_m)
examples/cast/example_group_per_split_token_cast_to_fp8.py:13
↓ 4 callersFunctionhas_let_value
Check if a variable has a binding in the current let frame stack. Args: var (Var): The variable to check Returns: bool: True
tilelang/language/frame.py:209
↓ 4 callersFunctionindex_to_coordinates
Convert a flat (linear) index into multi-dimensional coordinates for a given shape. Given a linear index and a shape (sequence of dimension
tilelang/language/utils.py:29
↓ 4 callersFunctioninfer_wgmma_shared_layout
(continuity, dtype)
testing/python/language/test_tilelang_language_atom_mma.py:31
↓ 4 callersMethodinit_tcgen05_b_desc
Emit TIR to initialize a pre-allocated TCGEN05 B descriptor. Parameters ---------- desc_b : Buffer A descriptor b
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:767
↓ 4 callersFunctionis_cpu_device_backend
(target: Target)
tilelang/engine/lower.py:22
↓ 4 callersMethodis_gemm_ss
(self)
tilelang/cuda/op/gemm/gemm_mma.py:221
↓ 4 callersMethodis_output
(self)
tilelang/carver/roller/node.py:125
↓ 4 callersFunctionis_rdna_wmma_target
(target: Target)
tilelang/carver/matmul_analysis.py:549
↓ 4 callersFunctionis_tensor_memory
Check if the buffer is in tensor memory scope (e.g., shared.tmem). Args: buffer: The TVM buffer, BufferLoad, or BufferRegion to chec
tilelang/utils/language.py:80
↓ 4 callersFunctionis_var
(v: Any)
tilelang/language/eager/builder.py:191
↓ 4 callersMethodldmatrix_e
(self, E_local_buf: Buffer, E_shared_buf: Buffer | BufferRegion | BufferLoad, ki: PrimExpr, rk: PrimExpr = 0)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:299
↓ 4 callersFunctionlegalize_arguments
Convert let-bound variables to their corresponding buffers. Args: arg (Union[tirx.Buffer, tirx.Var]): Input argument to legalize
tilelang/language/gemm_op.py:41
↓ 4 callersFunctionlegalize_arguments
(arg: BufferLikeType | tirx.Var)
tilelang/language/experimental/gemm_sp_op.py:39
↓ 4 callersFunctionlinear
Applies a linear transformation to the incoming data: y = xA^T + b. This function supports specialized implementations based on quantization
examples/deepseek_v32/inference/model.py:137
↓ 4 callersMethodload_lib
(self, lib_path: str | None = None)
tilelang/jit/adapter/cutedsl/libgen.py:46
↓ 4 callersFunctionmacro_with_var
(x: T.Var)
testing/python/language/test_tilelang_language_frontend_v2.py:344
↓ 4 callersFunctionmain
( A: T.Tensor(A_shape, in_dtype), B: T.Tensor(B_shape, storage_dtype), Scale: T.Tensor
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_cdna4.py:202
↓ 4 callersFunctionmain
Tiled, pipelined kernel entry that multiplies A with a quantized B (with per-block Scale) producing C. This prim-level kernel implem
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_hopper.py:318
↓ 4 callersFunctionmake_dq_layout
(dQ)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:122
↓ 4 callersFunctionmake_dq_layout
(dQ)
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:187
↓ 4 callersFunctionmake_half_bank_swizzled_layout
Args: buffer: BufferLikeType Examples: make_half_bank_swizzled_layout(buffer)
tilelang/layout/swizzle.py:106
↓ 4 callersFunctionmake_iter_fusion_index_map
( traits: list[IterTrait], kind_order: list[IterKind], )
tilelang/carver/matmul_analysis.py:214
↓ 4 callersMethodmake_mfma_load_layout
Create a layout function for storing MFMA results into a fragment buffer. Parameters ---------- local_buf : tirx.Buf
tilelang/rocm/intrinsics/mfma_macro_generator.py:564
↓ 4 callersMethodmake_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_sp_macro_generator.py:621
↓ 4 callersMethodmake_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_sp_macro_generator.py:750
← previousnext →401–500 of 7,254, ranked by callers