MCPcopy Create free account

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

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

↓ 1 callersMethodgenerate_tma_descriptor_args
(self, desc_name_map: dict[str, str], desc_name_var_map: dict[str, tvm.tirx.Var])
tilelang/jit/adapter/wrapper.py:403
↓ 1 callersMethodgenerate_tma_descriptor_args
Generate Python code to initialize TMA descriptors. TMA (Tensor Memory Accelerator) descriptors are opaque CUDA objects that describe
tilelang/jit/adapter/nvrtc/wrapper.py:484
↓ 1 callersMethodgenerate_tma_descriptor_args
Generate TMA descriptor information for C++ code generation. Returns: List of descriptor variable names in the order they were pr
tilelang/jit/adapter/cutedsl/wrapper.py:1436
↓ 1 callersMethodgenerate_w_logprobs
( self, prompts: List[str], sampling_params: SamplingParams, images: Optional[
examples/bitnet-1.58b/vllm_workspace/conftest.py:473
↓ 1 callersFunctionget_abs_err
(x, y)
examples/kda/FLA_KDA/fla_utils.py:34
↓ 1 callersFunctionget_all_max_shared_mem
()
examples/kda/FLA_KDA/fla_utils.py:224
↓ 1 callersFunctionget_analyzer_by_tir
(block_analyzer, args)
tilelang/carver/roller/shape_inference/tir.py:370
↓ 1 callersFunctionget_args
()
examples/deepseek_mla/benchmark_mla.py:559
↓ 1 callersFunctionget_args
()
examples/deepseek_mla/amd/benchmark_mla_decode_amd_aiter.py:221
↓ 1 callersFunctionget_args
()
examples/deepseek_mla/amd/benchmark_mla_decode_amd_triton.py:440
↓ 1 callersFunctionget_autotune_inputs
Get the current autotune inputs from the stack.
tilelang/autotuner/capture.py:121
↓ 1 callersFunctionget_available_cpu_count
Gets the number of CPU cores available to the current process.
tilelang/autotuner/tuner.py:192
↓ 1 callersMethodget_base_tile
Gets the minimum tile configuration that satisfies no redundancy in computation. Returns ------- List[int]
tilelang/carver/roller/policy/default.py:136
↓ 1 callersFunctionget_best_config
( M, N, K, with_roller: bool = False, profile_backend: str = "event", )
examples/gemm/example_gemm_autotune.py:110
↓ 1 callersFunctionget_best_config
( M, N, K, with_roller: bool = False, profile_backend: str = "event", execution_backen
examples/gemm/example_gemm_advanced_autotune.py:118
↓ 1 callersMethodget_block_bindings
Returns all three block bindings.
tilelang/language/kernel.py:239
↓ 1 callersFunctionget_block_extents
Returns all three block extents.
tilelang/language/kernel.py:481
↓ 1 callersMethodget_block_extents
Returns the block extents for all three dimensions.
tilelang/language/kernel.py:189
↓ 1 callersMethodget_block_name
(self, block: SBlockRV)
tilelang/carver/roller/node.py:40
↓ 1 callersFunctionget_boolop_name
(boolop: ast.boolop)
tilelang/language/eager/ast.py:86
↓ 1 callersMethodget_buffers
(self, block: SBlockRV)
tilelang/carver/roller/node.py:77
↓ 1 callersFunctionget_bwd_configs
()
examples/attention_sink/example_gqa_sink_bwd_varlen.py:14
↓ 1 callersFunctionget_bwd_configs
()
examples/attention_sink/example_mha_sink_bwd_bhsd.py:11
↓ 1 callersFunctionget_bwd_configs
()
examples/attention_sink/example_gqa_sink_bwd_bhsd.py:11
↓ 1 callersFunctionget_col_major_tma_aligned_tensor
Returns TMA-aligned transposed format of the input tensor. `torch.transpose` will be called if necessary. If the input tensor is already colu
examples/cast/example_group_per_split_token_cast_to_fp8.py:92
↓ 1 callersFunctionget_configs
Generate a list of kernel tuning configuration dictionaries for a tiled matrix-multiply. When with_roller is True this queries the MatmulTem
examples/gemm/example_gemm_autotune.py:22
↓ 1 callersFunctionget_configs
Generate a list of kernel tuning configuration dictionaries for a tiled matrix-multiply. When with_roller is True this queries the MatmulTem
examples/gemm/example_gemm_advanced_autotune.py:30
↓ 1 callersMethodget_consumer_blocks
(self, block: SBlockRV)
tilelang/carver/roller/node.py:83
↓ 1 callersMethodget_cpu_init_func
(self)
tilelang/jit/adapter/wrapper.py:870
↓ 1 callersFunctionget_cta_rank_in_cluster
(cluster_size=4)
testing/python/language/test_tilelang_language_cluster.py:31
↓ 1 callersFunctionget_cuda_dll_search_dirs
Return CUDA_HOME-derived DLL search directories (Windows only). The CUDA_HOME value itself is auto-detected by ``_find_cuda_home`` (env vars,
tilelang/env.py:414
↓ 1 callersFunctionget_cuda_version
Utility function to get cuda version Parameters ---------- cuda_path : Optional[str] Path to cuda root. If None is passed, will
tilelang/contrib/nvcc.py:344
↓ 1 callersFunctionget_current_device
()
tilelang/utils/device.py:14
↓ 1 callersMethodget_current_device_functor
Return a callable that yields Torch's current device. Similar to the stream functor, we capture a callable that, when called, fetches
tilelang/jit/adapter/base.py:82
↓ 1 callersMethodget_declaration
(self, declare_kernel_code: str)
tilelang/jit/adapter/wrapper.py:382
↓ 1 callersMethodget_device_function
(self)
tilelang/carver/roller/rasterization.py:66
↓ 1 callersFunctionget_device_name
(device_id: int = 0)
tilelang/carver/arch/driver/cuda_driver.py:33
↓ 1 callersFunctionget_dummy_kernel
()
testing/python/language/test_tilelang_memory_leak.py:14
↓ 1 callersMethodget_dynamic_symbolic_set
(self, prim_func)
tilelang/jit/adapter/wrapper.py:527
↓ 1 callersFunctionget_e_replicate_factor
Return how many consecutive threads share the same logical metadata value.
tilelang/utils/sparse.py:25
↓ 1 callersFunctionget_fast_dequant_func
(in_dtype="fp4", out_dtype=T.bfloat16)
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_cdna4.py:124
↓ 1 callersFunctionget_fast_dequant_twiddling_func
Return a TileLang macro that performs fast dequantization of twiddled FP4-packed data into BF16. The returned macro has signature (B
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_hopper.py:173
↓ 1 callersFunctionget_fast_dequant_twiddling_func
Create a TileLang macro that performs fast, twiddling-based dequantization from packed FP4 to BF16 using an external runtime plugin.
examples/dequantize_gemm/example_dequant_gemm_bf16_fp4_hopper.py:124
↓ 1 callersMethodget_fileline_stack
(self, stacklevel=1)
tilelang/language/eager/builder.py:765
↓ 1 callersMethodget_grid_size
(self)
examples/autodd/tilelang_buggy.py:70
↓ 1 callersMethodget_hardware_aware_configs
Abstract method that must be implemented by subclasses. It should return a list of hardware-aware configurations (hints) base
tilelang/carver/template/base.py:35
↓ 1 callersFunctionget_host_func
Get the host function from the module.
testing/python/transform/test_tilelang_transform_split_host_device.py:30
↓ 1 callersFunctionget_index_map
Get index maps for the block Parameters ---------- block : tirx.SBlock The block to be analyzed layout : List[str] t
tilelang/carver/matmul_analysis.py:311
↓ 1 callersFunctionget_inf_kernel
(dtype: str)
testing/python/language/test_tilelang_language_infinity.py:8
↓ 1 callersMethodget_init_func
(self)
tilelang/jit/adapter/wrapper.py:560
↓ 1 callersMethodget_input_buffers
(self, block: SBlockRV)
tilelang/carver/roller/node.py:65
↓ 1 callersFunctionget_kernel
(m: int)
testing/python/issue/test_tilelang_issue_1257.py:19
↓ 1 callersMethodget_kernel_launch_code
(self, function_name, grid_str, block_str, smem_str, call_args, cluster_dims)
tilelang/jit/adapter/wrapper.py:554
↓ 1 callersMethodget_kernel_source
(self, *args: _P.args, **kwargs: _P.kwargs)
tilelang/jit/__init__.py:444
↓ 1 callersMethodget_kernel_source
Returns the source code of the compiled kernel.
tilelang/jit/adapter/cython/adapter.py:387
↓ 1 callersMethodget_kernel_source
Get the CuTeDSL device source, optionally with host wrapper source. Returns ------- str | None The kernel source
tilelang/jit/adapter/cutedsl/adapter.py:251
↓ 1 callersMethodget_key
(self, shape, dtype)
examples/deepseek_v32/fp8_lighting_indexer.py:65
↓ 1 callersFunctionget_layout_visual_formats
(pass_ctx: PassContext | None = None)
tilelang/backend/pass_pipeline/pipeline_utils.py:55
↓ 1 callersFunctionget_lop3_intrin_group
This function is used to get the intrinsic group of the LOP3 operation to avoid the overhead of fast decoding. LOP3 is a type of logic operat
tilelang/quantize/lop3.py:1091
↓ 1 callersFunctionget_mathop_lines
Extract lines containing the mathop from CUDA source for debugging
testing/python/fastmath/test_mathops_fastmath.py:9
↓ 1 callersFunctionget_mathop_lines
Extract lines containing the mathop from CUDA source for debugging
testing/python/math/test_math_fast_math.py:9
↓ 1 callersFunctionget_max_dynamic_shared_size_bytes
Get the maximum dynamic shared memory size in bytes, kilobytes, or megabytes.
tilelang/carver/arch/driver/cuda_driver.py:90
↓ 1 callersFunctionget_memory_requirements
Calculate memory requirements - completely useless
examples/autodd/tilelang_buggy.py:32
↓ 1 callersMethodget_name
(self)
tilelang/autodd.py:88
↓ 1 callersMethodget_next_task
(self)
tilelang/autodd.py:1159
↓ 1 callersMethodget_node_reduce_step_candidates
(self, node)
tilelang/carver/roller/policy/tensorcore.py:206
↓ 1 callersFunctionget_num_sms
Get the number of streaming multiprocessors (SMs) on the CUDA device. Args: device_id (int, optional): The CUDA device ID. Defaults
tilelang/carver/arch/driver/cuda_driver.py:111
↓ 1 callersMethodget_num_threads
Returns the thread indices from the topmost frame.
tilelang/language/kernel.py:223
↓ 1 callersFunctionget_operator_name
(operator: ast.operator)
tilelang/language/eager/ast.py:82
↓ 1 callersMethodget_parent_locals
(self)
tilelang/language/eager/ast.py:172
↓ 1 callersFunctionget_pipeline
Return the registered Pipeline for *name*.
tilelang/backend/pass_pipeline/pipeline.py:39
↓ 1 callersFunctionget_pytorch_reference
Get PyTorch reference implementation for the given operation.
maint/precision/compare_ops.py:313
↓ 1 callersFunctionget_qwq
(hidden: int)
testing/python/issue/test_tilelang_issue_1728.py:8
↓ 1 callersMethodget_reduce_inputs_dtype
(self)
tilelang/carver/roller/node.py:401
↓ 1 callersFunctionget_registers_per_block
Get the maximum number of 32-bit registers available per block.
tilelang/carver/arch/driver/cuda_driver.py:130
↓ 1 callersFunctionget_roller_hints_from_output_nodes
( output_nodes: list[OutputNode], arch: TileDevice, topk: int = 10, extra_tags: list[str] | None = None )
tilelang/carver/utils.py:68
↓ 1 callersFunctionget_root_block
(sch: Schedule, func_name: str = "main")
tilelang/carver/analysis.py:201
↓ 1 callersFunctionget_sample_kernel
()
testing/python/language/test_tilelang_language_frontend_v2.py:408
↓ 1 callersFunctionget_select_kernel_1
()
testing/python/language/test_tilelang_language_select.py:7
↓ 1 callersFunctionget_select_kernel_2
()
testing/python/language/test_tilelang_language_select.py:42
↓ 1 callersFunctionget_shared_kernel
()
testing/python/issue/test_tilelang_issue_1026.py:7
↓ 1 callersMethodget_shared_memory
(cls, arch: str)
examples/kda/FLA_KDA/fla_utils.py:217
↓ 1 callersFunctionget_shared_memory_per_block
(device_id: int = 0, format: str = "bytes")
tilelang/carver/arch/driver/cuda_driver.py:39
↓ 1 callersMethodget_shared_memory_size
(self)
examples/autodd/tilelang_buggy.py:75
↓ 1 callersFunctionget_simple_dequant_func
(in_dtype="fp4", out_dtype=T.bfloat16)
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_cdna4.py:172
↓ 1 callersFunctionget_simple_dequant_func
Create a simple (scalar) dequantization macro that converts 4-bit packed inputs to bfloat16. Returns a T.macro that, given shared-st
examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_hopper.py:265
↓ 1 callersFunctionget_simple_dequant_func
Create a simple TIR dequantization macro that converts packed 4-bit FP (FP4) stored in uint8 into bfloat16. The returned macro (name
examples/dequantize_gemm/example_dequant_gemm_bf16_fp4_hopper.py:196
↓ 1 callersFunctionget_single_device_function_name
(device_mod)
testing/python/language/test_tilelang_language_source_kernel.py:37
↓ 1 callersFunctionget_sparse_attn_mask_from_topk
(x, topk, use_dense_for_last_block=False)
benchmark/blocksparse_attention/benchmark_torch_block_sparse_fmha.py:9
↓ 1 callersFunctionget_sparse_attn_mask_from_topk
(x, topk, use_dense_for_last_block=False)
benchmark/blocksparse_attention/benchmark_triton_block_sparse_fmha.py:14
↓ 1 callersFunctionget_sparse_attn_mask_from_topk
(x, topk, use_dense_for_last_block=False)
benchmark/blocksparse_attention/benchmark_tilelang_block_sparse_fmha.py:14
↓ 1 callersMethodget_spm_processor
(self, from_slow=False)
examples/bitnet-1.58b/tokenization_bitnet.py:205
↓ 1 callersFunctionget_stack_str
(msg, stacklevel=1)
tilelang/language/print_op.py:132
↓ 1 callersMethodget_store_index_map
(self, inverse: bool = False)
tilelang/rocm/intrinsics/mfma_macro_generator.py:323
↓ 1 callersMethodget_store_index_map
(self, inverse: bool = False)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:150
↓ 1 callersMethodget_store_index_map
(self, inverse: bool = False)
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:189
↓ 1 callersMethodget_stream_type
(self)
tilelang/jit/adapter/wrapper.py:621
↓ 1 callersMethodget_stream_type
Return stream parameter spec for Python signature. NVRTC backend uses raw int for stream handle (not cudaStream_t pointer). Default t
tilelang/jit/adapter/nvrtc/wrapper.py:600
↓ 1 callersFunctionget_target_by_dump_machine
Functor of get_target_triple that can get the target triple using compiler. Parameters ---------- compiler : Optional[str] The co
tilelang/contrib/cc.py:263
↓ 1 callersFunctionget_test_dataset
(dataset_name, tokenizer, seqlen=2048)
examples/bitnet-1.58b/eval_utils.py:16
← previousnext →2,201–2,300 of 7,254, ranked by callers