MCPcopy Create free account

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

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

Function_patched_buffer_getitem
(self, indices)
tilelang/language/overrides/buffer.py:13
Function_per_token_group_quant_fp8
A Triton-accelerated function to perform per-token-group quantization on a tensor. This function converts the tensor values into float8 values
examples/cast/example_triton_cast_to_fp8.py:12
Function_per_token_group_quant_fp8_colmajor
A Triton-accelerated function to perform per-token-group quantization on a tensor. This function converts the tensor values into float8 values
examples/cast/example_triton_cast_to_fp8.py:58
Method_post_visit
Post-visit callback for IR nodes. Args: stmt: The current IR node being visited.
tilelang/tools/Analyzer.py:146
Function_postorder
(op)
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:11
Function_pre_order
(stmt)
tilelang/metal/transform/metal_fragment_to_simdgroup.py:107
Method_pre_visit
Pre-visit callback for IR nodes. Args: stmt: The current IR node being visited.
tilelang/tools/Analyzer.py:119
Method_process_benchmark_result
(result_item, progress_bar)
tilelang/autotuner/tuner.py:1058
Method_record_benchmark_result
(latency: float, config: dict[str, Any], jit_kernel: tilelang.JITKernel, idx: int, progress_bar)
tilelang/autotuner/tuner.py:1020
Function_register
(fn: Callable[[str, Target], str])
tilelang/engine/callback.py:79
Method_reorder_cache
(past_key_values, beam_idx)
examples/bitnet-1.58b/modeling_bitnet.py:1327
Method_replace_weight_param_with_qweight
(self)
examples/bitnet-1.58b/modeling_bitnet.py:1379
Method_reset_cache
(self)
examples/bitnet-1.58b/modeling_bitnet.py:873
Method_rs_atom
(A_buf, desc_b, C_buf)
tilelang/cuda/intrinsics/macro/wgmma_macro_generator.py:574
Method_run_benchmark_target
( _jit_kernel: tilelang.JITKernel = jit_kernel, _worker_state:
tilelang/autotuner/tuner.py:644
Method_save_kernel_source_code_to_disk
Save the original CuTeDSL device source for source inspection.
tilelang/jit/adapter/cutedsl/kernel_cache.py:21
Method_save_so_cubin_to_disk
(self, kernel: JITKernel, cache_path: str, verbose: bool = False)
tilelang/jit/adapter/kernel_cache.py:16
Method_save_so_cubin_to_disk
(self, kernel: JITKernel, cache_path: str, verbose: bool = False)
tilelang/jit/adapter/nvrtc/kernel_cache.py:14
Method_save_so_cubin_to_disk
Save launcher shared objects and optional debugging artifacts.
tilelang/jit/adapter/cutedsl/kernel_cache.py:49
Method_save_wrapper_kernel_code_to_disk
(self, kernel: JITKernel, cache_path: str, verbose: bool = False)
tilelang/jit/adapter/kernel_cache.py:10
Method_save_wrapper_kernel_code_to_disk
Save both the generated host wrapper and importable kernel module.
tilelang/jit/adapter/cutedsl/kernel_cache.py:31
Method_score
(rstep_id)
tilelang/carver/roller/policy/tensorcore.py:144
Method_score
(rstep_id)
tilelang/carver/roller/policy/default.py:293
Method_set_adapter_cache_path
Record the cache path so first execution can persist generated cubin.
tilelang/jit/adapter/cutedsl/kernel_cache.py:73
Method_setup_cache
(self, cache_cls, max_batch_size, max_cache_len: Optional[int] = None)
examples/bitnet-1.58b/modeling_bitnet.py:858
Function_setup_tilelang_environment
()
maint/gemm/correctness_evaluation_sm70.py:258
Function_setup_tilelang_environment
()
maint/gemm/correctness_evaluation.py:520
Method_shared_memory_usage
(td: TileDict)
tilelang/carver/roller/policy/tensorcore.py:141
Method_simdgroup_copy
Copy tiles between simdgroup local and shared memory.
tilelang/metal/intrinsics/metal_macro_generator.py:201
Method_smem_elems_in_bytes
Byte width of one SMEM element for TCGEN05 descriptor offset math.
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:120
Function_split_kernel
( q_ptr, k_cache_ptr, v_cache_ptr, cache_seqlens_ptr, o_partial_ptr, lse_partial_ptr,
examples/blocksparse_attention/example_triton_sparse_gqa_decode_varlen_mask.py:17
Function_split_kernel
( q_ptr, k_cache_ptr, v_cache_ptr, cache_seqlens_ptr, o_partial_ptr, lse_partial_ptr,
examples/blocksparse_attention/example_triton_sparse_gqa_decode_varlen_indice.py:19
Method_ss_atom
(desc_a, desc_b, C_local_buf)
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:923
Method_ss_atom
(desc_a, desc_b, C_buf)
tilelang/cuda/intrinsics/macro/wgmma_macro_generator.py:674
Function_timeout_handler
(signum, frame)
tilelang/autotuner/tuner.py:47
Function_tir_f16_to_uint_to_f4
(val: tirx.PrimExpr)
tilelang/quantize/quantization.py:158
Function_tir_f32_to_uint_to_f4
(val: tirx.PrimExpr)
tilelang/quantize/quantization.py:141
Function_tir_f32x2_to_bf16x2_to_u32
Convert two float32 values to bfloat16 and pack them into a single uint32. The two inputs v0 and v1 (float32 PrimExpr) are reinterpreted as
tilelang/quantize/quantization.py:75
Function_tir_packed_int_to_int_to_float
(storage_nbit: int)
tilelang/quantize/quantization.py:128
Function_tir_packed_to_fp4_to_f16
(nbit: int, val: tirx.PrimExpr, pos: tirx.PrimExpr, dtype: str)
tilelang/quantize/quantization.py:187
Function_tir_packed_to_signed_convert
(storage_type="uint", storage_nbit=8)
tilelang/quantize/quantization.py:246
Function_tir_packed_to_unsigned_convert_with_zeros
(storage_type="uint", storage_nbit=8)
tilelang/quantize/quantization.py:269
Function_tir_packed_uint_to_uint_to_float
(storage_nbit: int)
tilelang/quantize/quantization.py:116
Function_tir_u32_to_bf16x2_to_f32x2
(x: tirx.PrimExpr)
tilelang/quantize/quantization.py:103
Function_tir_u32_to_f4_to_f32
(nbit: int, val: tirx.PrimExpr, pos: tirx.PrimExpr, dtype: str)
tilelang/quantize/quantization.py:171
Function_tir_u32_to_int_to_float
(nbit: int, val: tirx.PrimExpr, pos: tirx.PrimExpr, dtype: str)
tilelang/quantize/quantization.py:110
Function_tir_u8_to_f4_to_bf16
Convert a packed 4-bit field stored in a uint8 into a bfloat16 value using an exponent scale. This function expects a storage field
tilelang/quantize/quantization.py:30
Function_tir_u8_to_f8_e4m3_to_f16
(nbit: int, val: tirx.PrimExpr, dtype: str)
tilelang/quantize/quantization.py:230
Function_tir_u8_to_f8_e4m3_to_f16_naive
(nbit: int, val: tirx.PrimExpr, dtype: str)
tilelang/quantize/quantization.py:219
Function_tir_u8_to_f8_e5m2_to_f16
(nbit: int, val: tirx.PrimExpr, dtype: str)
tilelang/quantize/quantization.py:240
Method_tokenize
Returns a tokenized string. We de-activated the `add_dummy_prefix` option, thus the sentencepiece internals will always strip any
examples/bitnet-1.58b/tokenization_bitnet.py:264
Function_triton_mixed_sparse_attn_fwd_kernel
( Q, K, V, seqlens, sm_scale, block_count, block_offset, column_count, col
examples/minference/example_vertical_slash_sparse_attn.py:289
Method_ts_atom
(a_data, desc_b, C_local_buf)
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:1019
Function_var_iter
(self)
tilelang/language/kernel.py:20
Function_visit
(node)
testing/python/issue/test_tilelang_issue_sm120_tma_smem_alignment.py:133
Function_visit
(node)
testing/python/issue/test_tilelang_issue_2123.py:40
Function_visit
(node)
testing/python/transform/test_tilelang_transform_flatten_buffer.py:10
Function_visit
(node)
testing/python/transform/test_tilelang_transform_lower_tile_op.py:13
Function_visit
(node)
testing/python/transform/test_tilelang_transform_lexical_alloc_scope.py:24
Function_visit
(node)
testing/python/transform/test_tilelang_transform_plan_update_buffer_allocation_location.py:22
Function_visit
(node)
testing/python/transform/test_tilelang_transform_legalize_safe_memory_access.py:36
Function_visit
(node)
testing/python/transform/test_tilelang_transform_pipeline_planning.py:28
Function_visit
(node)
testing/python/transform/test_tilelang_transform_lower_ptx_async_copy.py:13
Function_visit
(stmt)
testing/python/transform/test_tilelang_transform_Inject_software_pipeline.py:22
Function_visit
(node)
testing/python/transform/test_tilelang_transform_inject_set_max_nreg.py:10
Function_visit_loop_body
(node)
testing/python/transform/test_tilelang_transform_plan_update_buffer_allocation_location.py:60
Method_visit_target
(target: ast.expr)
tilelang/language/eager/ast.py:353
Function_visitor
(node)
testing/python/transform/test_tilelang_transform_make_packed_api.py:32
Function_visitor
(node)
tilelang/jit/adapter/utils.py:201
Function_visitor
(stmt)
tilelang/metal/transform/metal_fragment_to_simdgroup.py:43
Method_wait
()
tilelang/cuda/intrinsics/macro/wgmma_macro_generator.py:503
Method_warp_ld_a_fp64
( A_local_buf, A_shared_buf, ki, thread_bindin
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:290
Method_warp_ld_b_fp64
( B_local_buf, B_shared_buf, ki, thread_bindin
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:402
Method_warp_ldmatrix_a
(A_local_buf, A_shared_buf, ki, thread_binding, rk=0)
tilelang/rocm/intrinsics/wmma_macro_generator.py:196
Method_warp_ldmatrix_a
( A_local_buf, A_shared_buf, ki, thread_binding, r
tilelang/rocm/intrinsics/mfma_macro_generator.py:395
Method_warp_ldmatrix_a
Load A matrix tiles via simdgroup_load per warp row.
tilelang/metal/intrinsics/metal_macro_generator.py:102
Method_warp_ldmatrix_a
( A_local_buf, A_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:214
Method_warp_ldmatrix_a
( A_local_buf, A_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:344
Method_warp_ldmatrix_a
( A_local_buf, A_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:996
Method_warp_ldmatrix_a
( A_local_buf, A_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:261
Method_warp_ldmatrix_a_global
( A_local_buf, A_buf, ki, thread_binding, rk=0,
tilelang/rocm/intrinsics/mfma_macro_generator.py:862
Method_warp_ldmatrix_a_shared
( A_local_buf, A_shared_buf, ki, thread_binding, r
tilelang/rocm/intrinsics/mfma_macro_generator.py:890
Method_warp_ldmatrix_b
(B_local_buf, B_shared_buf, ki, thread_binding, rk=0)
tilelang/rocm/intrinsics/wmma_macro_generator.py:250
Method_warp_ldmatrix_b
( B_local_buf, B_shared_buf, ki, thread_binding, r
tilelang/rocm/intrinsics/mfma_macro_generator.py:439
Method_warp_ldmatrix_b
Load B matrix tiles via simdgroup_load per warp column.
tilelang/metal/intrinsics/metal_macro_generator.py:139
Method_warp_ldmatrix_b
( B_local_buf, B_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:253
Method_warp_ldmatrix_b
( B_local_buf, B_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:456
Method_warp_ldmatrix_b
( B_local_buf, B_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:1092
Method_warp_ldmatrix_b
( B_local_buf, B_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:408
Method_warp_ldmatrix_b_global
( B_local_buf, B_buf, ki, thread_binding, rk=0,
tilelang/rocm/intrinsics/mfma_macro_generator.py:936
Method_warp_ldmatrix_b_shared
( B_local_buf, B_shared_buf, ki, thread_binding, r
tilelang/rocm/intrinsics/mfma_macro_generator.py:964
Method_warp_ldmatrix_e
( E_local_buf, E_shared_buf, ki, thread_binding, r
tilelang/cuda/intrinsics/macro/mma_sp_macro_generator.py:352
Method_warp_ldmatrix_e
( E_local_buf, E_shared_buf, inst_i, ki, thread_bi
tilelang/cuda/intrinsics/macro/wgmma_sp_macro_generator.py:482
Method_warp_mfma
(A_local_buf, B_local_buf, C_local_buf)
tilelang/rocm/intrinsics/mfma_macro_generator.py:488
Method_warp_mma
Execute simdgroup matrix multiply-accumulate across all warp tiles.
tilelang/metal/intrinsics/metal_macro_generator.py:170
Method_warp_mma
(A_local_buf, B_local_buf, C_local_buf)
tilelang/cuda/intrinsics/macro/mma_sm70_macro_generator.py:301
Method_warp_mma
(A_local_buf, B_local_buf, C_local_buf)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:506
Method_warp_mma
(A_local_buf, B_local_buf, C_local_buf)
tilelang/cuda/intrinsics/macro/mma_macro_generator.py:1189
Method_warp_mma
(C_buf)
tilelang/cuda/intrinsics/macro/wgmma_macro_generator.py:223
Method_warp_mma
(A_ptr, B_ptr, C_buf)
tilelang/cuda/intrinsics/macro/wgmma_sp_macro_generator.py:221
← previousnext →3,901–4,000 of 7,254, ranked by callers