MCPcopy Create free account

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

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

↓ 1 callersFunctionfind_lld
Find ld.lld in system. Parameters ---------- required : bool Whether it is required, runtime error will be raised if the
tilelang/contrib/rocm.py:33
↓ 1 callersMethodfind_path_from_source
Finds the path (if it exists) from a starting node (source) to a target node. Returns the path as a list of nodes.
tilelang/carver/roller/shape_inference/tir.py:131
↓ 1 callersFunctionfind_topo_sort
(output_node_list)
tilelang/carver/roller/node.py:576
↓ 1 callersFunctionflash_mla
()
examples/deepseek_mla/benchmark_mla.py:70
↓ 1 callersFunctionflash_mla_tilelang
()
examples/deepseek_mla/example_mla_decode_paged.py:274
↓ 1 callersFunctionflash_mla_tilelang
()
examples/deepseek_mla/benchmark_mla.py:430
↓ 1 callersFunctionflash_mla_triton
()
examples/deepseek_mla/benchmark_mla.py:391
↓ 1 callersFunctionflash_mla_triton
()
examples/deepseek_mla/amd/benchmark_mla_decode_amd_triton.py:312
↓ 1 callersFunctionflash_split_ref
(Q, K, V, mask)
examples/flash_decoding/example_gqa_decode.py:291
↓ 1 callersFunctionflashattn
(batch_size, groups, UQ, UKV, heads, dim, is_causal, block_M=64, block_N=64, num_stages=1, threads=128)
examples/flash_attention/example_gqa_fwd_varlen.py:17
↓ 1 callersFunctionflashattn
(batch, heads, kv_head_num, seqlen_kv, dim, pe_dim, block_N, block_H, num_split)
examples/warp_specialize/example_warp_specialize_flashmla.py:11
↓ 1 callersFunctionflashattn
Flash Attention forward. variant='ss': mma_ss (128t, P via shared); 'ts': mma_ts (256t, P via TMEM).
examples/flash_attention_sm100/mha_fwd_bshd.py:23
↓ 1 callersFunctionflashattn
GQA forward. variant='ss': mma_ss (128t, P via shared); 'ts': mma_ts (256t, P via TMEM).
examples/flash_attention_sm100/gqa_fwd_bshd.py:21
↓ 1 callersFunctionflashattn
(batch, heads, kv_head_num, seqlen_kv, dim, pe_dim, block_N, block_H, num_split, softmax_scale)
examples/deepseek_mla/example_mla_decode_ws.py:28
↓ 1 callersFunctionflashattn
(batch, heads, kv_head_num, seqlen_kv, dim, pe_dim, block_N, block_H, num_split)
examples/deepseek_mla/example_mla_decode_persistent.py:17
↓ 1 callersFunctionflashattn
(batch, heads, kv_head_num, seqlen_kv, dim, pe_dim, block_N, block_H)
examples/deepseek_mla/experimental/example_mla_decode_kv_fp8.py:16
↓ 1 callersFunctionflashattn_bwd
( batch, heads, seq_len, dim, is_causal, groups, block_M: int, block_N: int,
examples/amd/example_amd_flash_attn_bwd.py:309
↓ 1 callersFunctionflashattn_bwd
( batch_size, groups, UQ, UKV, N_CTX, heads, max_seq_len, dim, is_causal,
examples/attention_sink/example_gqa_sink_bwd_varlen.py:301
↓ 1 callersFunctionflashattn_bwd_atomic_add
(batch, heads, seq_len, dim_qk, dim_v, is_causal, block_M, block_N, threads=256, num_stages=2, groups=1)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:171
↓ 1 callersFunctionflashattn_bwd_atomic_add
(batch, heads, seq_len, dim_qk, dim_v, is_causal, block_M, block_N, threads=256, num_stages=2, groups=1)
examples/flash_attention/example_gqa_bwd.py:156
↓ 1 callersFunctionflashattn_bwd_atomic_add
( batch, total_q, total_kv, N_CTX, heads, max_seq_len, dim_qk, dim_v, is_causal, block_M, block_N, threads
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:236
↓ 1 callersFunctionflashattn_bwd_dsink
(batch_size, heads, N_CTX, max_seq_len, block=256, dtype: T.dtype = T.float16)
examples/attention_sink/example_gqa_sink_bwd_varlen.py:464
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, seq_len, dim)
examples/amd/example_amd_flash_attn_bwd.py:430
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, seq_len, dim)
examples/flash_attention/example_mha_bwd_bhsd.py:132
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, head_kv, seq_len, dim_qk, dim_v)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:133
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, seq_len, dim_qk)
examples/flash_attention/example_gqa_bwd.py:130
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, seq_len, dim)
examples/flash_attention/example_mha_bwd_bshd.py:129
↓ 1 callersFunctionflashattn_bwd_postprocess
(total_q, total_kv, heads, head_kv, dim_qk, dim_v)
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:198
↓ 1 callersFunctionflashattn_bwd_postprocess
(UQ, heads, dim, dtype: T.dtype = T.float16)
examples/attention_sink/example_gqa_sink_bwd_varlen.py:276
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, seq_len, dim, dtype: T.dtype = T.float16)
examples/attention_sink/example_mha_sink_bwd_bhsd.py:178
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, seq_len, dim, dtype: T.dtype = T.float16)
examples/attention_sink/example_gqa_sink_bwd_bhsd.py:179
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, seq_len, dim)
examples/flash_attention_sm100/gqa_bwd_bshd.py:131
↓ 1 callersFunctionflashattn_bwd_postprocess
(batch, heads, seq_len, dim)
examples/flash_attention_sm100/mha_bwd_bshd.py:126
↓ 1 callersFunctionflashattn_bwd_preprocess
(batch, heads, seq_len, dim)
examples/amd/example_amd_flash_attn_bwd.py:282
↓ 1 callersFunctionflashattn_bwd_preprocess
(batch, heads, seq_len, dim_v)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:93
↓ 1 callersFunctionflashattn_bwd_preprocess
(batch_size, heads, UQ, N_CTX, max_seq_len, dim, dtype: T.dtype = T.float16)
examples/attention_sink/example_gqa_sink_bwd_varlen.py:223
↓ 1 callersFunctionflashattn_bwd_preprocess
(batch, heads, seq_len, dim)
examples/flash_attention_sm100/gqa_bwd_bshd.py:94
↓ 1 callersFunctionflashattn_bwd_preprocess
(batch, heads, seq_len, dim)
examples/flash_attention_sm100/mha_bwd_bshd.py:89
↓ 1 callersFunctionflashattn_bwd_split_novarlen
(batch, heads, seq_len, dim_qk, dim_v, is_causal, block_M, block_N, threads=256, num_stages=2, groups=1)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:258
↓ 1 callersFunctionflashattn_fwd
(batch, heads, seq_len, dim, is_causal, block_M, block_N)
examples/flash_attention/example_mha_bwd_bhsd.py:15
↓ 1 callersFunctionflashattn_fwd
(batch, heads, seq_len, dim_qk, dim_v, is_causal, block_M, block_N, groups=1)
examples/flash_attention/example_gqa_bwd_tma_reduce.py:15
↓ 1 callersFunctionflashattn_fwd
(batch, heads, seq_len, dim_qk, dim_v, is_causal, block_M, block_N, groups=1)
examples/flash_attention/example_gqa_bwd.py:14
↓ 1 callersFunctionflashattn_fwd
(batch, heads, seq_len, dim, is_causal, block_M, block_N)
examples/flash_attention/example_mha_bwd_bshd.py:15
↓ 1 callersFunctionflashattn_fwd
(batch, total_q, total_kv, N_CTX, heads, max_seq_len, dim_qk, dim_v, is_causal, block_M, block_N, groups=1)
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:29
↓ 1 callersFunctionflashattn_fwd
( batch_size, groups, UQ, UKV, N_CTX, heads, max_seq_len, dim, is_causal,
examples/attention_sink/example_gqa_sink_bwd_varlen.py:29
↓ 1 callersFunctionflashattn_fwd
Forward for LSE; K/V indexed by by // groups.
examples/flash_attention_sm100/gqa_bwd_bshd.py:19
↓ 1 callersFunctionflashattn_fwd
Forward to get O and LSE (for backward).
examples/flash_attention_sm100/mha_bwd_bshd.py:18
↓ 1 callersFunctionflashattn_sink
( batch_size, groups, UQ, UKV, heads, dim, is_causal, window_size=None, # Non
examples/attention_sink/example_gqa_sink_fwd_varlen.py:24
↓ 1 callersFunctionflashattn_wasp
Warp-specialized pipeline: softmax(0-127)/DMA(128-159)/BMM(160-191); GEMM2 mma_ts.
examples/flash_attention_sm100/mha_fwd_bshd.py:177
↓ 1 callersFunctionflashattn_wasp
GQA warp-specialized pipeline: softmax(0-127)/DMA(128-159)/BMM(160-191); GEMM2 mma_ts.
examples/flash_attention_sm100/gqa_fwd_bshd.py:177
↓ 1 callersFunctionflashinfer
()
examples/deepseek_mla/benchmark_mla.py:127
↓ 1 callersFunctionfloat_init_readback
()
testing/python/amd/test_tilelang_hip_codegen.py:516
↓ 1 callersFunctionfn
()
examples/deepseek_v32/sparse_mla_fwd.py:264
↓ 1 callersFunctionfn
()
examples/deepseek_v32/sparse_mla_fwd_pipelined.py:443
↓ 1 callersFunctionfn
()
examples/dsa_sparse_finetune/sparse_mla_fwd.py:296
↓ 1 callersMethodforward
Forward pass for the Multi-Head Latent Attention (MLA) Layer. Args: x (torch.Tensor): Input tensor of shape (batch_size,
examples/deepseek_v32/inference/model.py:579
↓ 1 callersMethodforward
(i: PrimExpr, j: PrimExpr)
tilelang/cuda/intrinsics/macro/tcgen05_macro_generator.py:526
↓ 1 callersFunctionfp4_act_quant_ref
PyTorch reference for block-wise FP4 quantization with E8M0 scale. Returns: quant: shape (M, N//2), dtype torch.float4_e2m1fn_x2 (packed)
examples/deepseek_v4/act_quant.py:249
↓ 1 callersFunctionfp4_quant_kernel
Block-wise FP4 quantization: bf16 -> float4_e2m1fn, E8M0 power-of-2 scale.
examples/deepseek_v4/act_quant.py:112
↓ 1 callersFunctionfp4_tma_copy_roundtrip
( M=128, N=256, block_M=64, block_N=128, )
testing/python/language/test_tilelang_language_tma_copy.py:265
↓ 1 callersFunctionfp4_tma_copy_unpacked_smem_store
(M=128, N=256, block_M=64, block_N=128)
testing/python/language/test_tilelang_language_tma_copy.py:307
↓ 1 callersFunctionfp4_to_uint8_view_test
(rows_per_cta=16, mask_k=256)
testing/python/language/test_tilelang_language_view.py:94
↓ 1 callersFunctionfp4_unpacked_x_fp8_gemm
(M: int, N: int, K: int)
maint/gemm/correctness_evaluation_tcgen05_fp4_unpacked.py:26
↓ 1 callersFunctionfp8_act_quant_ref
PyTorch reference for block-wise FP8 quantization. Returns: quant: shape (M, N), dtype torch.float8_e4m3fn. scale: shape (M, num_
examples/deepseek_v4/act_quant.py:191
↓ 1 callersFunctionfp8_fp4_gemm_1d1d_persistent
( M: int, N: int, K: int, out_dtype, )
examples/deepseek_v4/fp8_fp4_gemm_1d1d_sm100.py:17
↓ 1 callersFunctionfp8_fp4_gemm_ref
(a_fp4, b_fp8, sfa_packed, sfb_packed, sf_granularity_k=128)
examples/deepseek_v4/fp8_fp4_gemm_1d1d_sm100.py:310
↓ 1 callersFunctionfp8_gemm
Perform a matrix multiplication using FP8 precision. Args: a (torch.Tensor): The first input matrix, must be contiguous. a_s
examples/deepseek_v32/inference/kernel.py:165
↓ 1 callersFunctionfp8_gemm_kernel
(N, K, out_dtype=BF16, accum_dtype=T.float32)
examples/deepseek_v32/inference/kernel.py:108
↓ 1 callersFunctionfp8_index
Perform index score using FP8 precision. Args: q (torch.Tensor): The Q tensor, must be contiguous. q_s (torch.Tensor): The s
examples/deepseek_v32/inference/kernel.py:246
↓ 1 callersFunctionfp8_index_kernel
(h: int, d: int)
examples/deepseek_v32/inference/kernel.py:192
↓ 1 callersFunctionfp8_matmul
(M, N, K, block_M, block_N, block_K, num_stages, num_threads, k_pack, gemm_type)
examples/gemm_fp8/example_tilelang_gemm_amd.py:57
↓ 1 callersFunctionfp8_native_block_mean_pooling
( K, KScale, dim: int = 128, pooling_block_size: int = 128, block_N: int = 64, num_sta
examples/dsa_hisa/fp8_block_mean_pooling.py:12
↓ 1 callersFunctionfp8_native_block_sparse_mqa_attn_return_logits
( IndexQ, IndexK, IndexKScale, TopKBlockIndex, Weights, CuSeqLenKS, CuSeqLenKE,
examples/dsa_hisa/block_sparse_mqa_fp8.py:14
↓ 1 callersFunctionfp8_quant_kernel
Block-wise FP8 quantization.
examples/deepseek_v4/act_quant.py:29
↓ 1 callersMethodfrom_buffer
Creates a KernelParam instance from a TVM Buffer object. Args: buffer: TVM Buffer object containing dtype and shape info
tilelang/engine/param.py:27
↓ 1 callersMethodfrom_ptr
( cls, pointer_var: Var, shape: ShapeType, dtype: DType = "float32", strides: tuple[PrimExpr, ...]
tilelang/language/proxy.py:242
↓ 1 callersMethodfrom_source
(cls, source: str, *args, **kwargs)
tilelang/autodd.py:721
↓ 1 callersMethodfrom_var
Creates a KernelParam instance from a TVM Variable object. Used for scalar parameters. Args: var: TVM Variable o
tilelang/engine/param.py:54
↓ 1 callersFunctionfrontend_cached_kernel
(block_m: int = 16)
testing/python/cache/test_tilelang_kernel_cache_atomic_save.py:245
↓ 1 callersFunctionfull_shape_tma_store_1d
(dtype, threads)
testing/python/language/test_tilelang_language_tma_store.py:117
↓ 1 callersFunctionfunc
()
testing/python/transform/test_tilelang_transform_thread_sync.py:30
↓ 1 callersMethodfunc
()
tilelang/autotuner/tuner.py:725
↓ 1 callersFunctionfused_index_kernel
(B: int, M: int, N: int, BLOCK_MN: int, BLOCK_K: int)
testing/python/layout/test_tilelang_layout_fused_replicate.py:14
↓ 1 callersFunctiongate_reduce
(n_tok: int, n_exp: int)
testing/python/amd/test_tilelang_hip_codegen.py:56
↓ 1 callersFunctiongather_scatter_program
(N: int, K: int, K_box: int, in_dtype: str = "float16")
testing/python/language/test_tilelang_language_tma_gather_scatter.py:25
↓ 1 callersFunctiongather_scatter_swizzled_program
(N: int, K: int, K_box: int, swizzle_kind: str, in_dtype: str = "float16")
testing/python/language/test_tilelang_language_tma_gather_scatter.py:100
↓ 1 callersFunctiongemm
( A, B, C, dtype: T.dtype = T.float16, accum_dtype: T.dtype = T.float3
testing/python/language/test_tilelang_language_eager_jit.py:10
↓ 1 callersFunctiongemm_ptr
( A: T.ptr, B: T.ptr, C: T.ptr, M: int, N: int, K: int,
testing/python/language/test_tilelang_language_eager_jit.py:49
↓ 1 callersFunctiongemv_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:14
↓ 1 callersFunctiongeneral_compress
(lowprecision_weight, source_bits=4, storage_dtype=np.int8)
examples/bitnet-1.58b/kernel_benchmark/tilelang_bitnet_158_int8xint2_decode.py:178
↓ 1 callersFunctiongeneral_compress
(lowprecision_weight, source_bits=4, storage_dtype=np.int8)
examples/bitnet-1.58b/kernel_benchmark/tilelang_bitnet_158_int8xint2_prefill.py:301
↓ 1 callersMethodgenerate_cache_key
Generate a cache key for the auto-tuning process.
tilelang/autotuner/tuner.py:439
↓ 1 callersFunctiongenerate_dense_input
(N, trans_A, trans_B, in_dtype, seed=0)
testing/python/utils/test_compress_utils.py:66
↓ 1 callersMethodgenerate_l2_persistent_map
(self, function_name: str)
tilelang/jit/adapter/wrapper.py:385
↓ 1 callersMethodgenerate_l2_persistent_map
Generate Python code to configure L2 cache persistence for a kernel. L2 persistence pins frequently-accessed data in L2 cache to reduce
tilelang/jit/adapter/nvrtc/wrapper.py:445
↓ 1 callersMethodgenerate_pdl_sync_code
Generate Python code to insert PDL synchronization for a given kernel.
tilelang/jit/adapter/nvrtc/wrapper.py:475
↓ 1 callersFunctiongenerate_random_padding_mask
(max_seqlen, batch_size, device, mode="random")
examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py:11
↓ 1 callersFunctiongenerate_test_data
Generate appropriate test data for each operation.
maint/precision/compare_ops.py:417
↓ 1 callersFunctiongenerate_text
(model, tokenizer, prompt, max_length=100)
examples/bitnet-1.58b/load_from_quantized.py:19
↓ 1 callersFunctiongenerate_text_batch
(model, tokenizer, prompts, max_length=100)
examples/bitnet-1.58b/benchmark_generate.py:13
← previousnext →2,101–2,200 of 7,254, ranked by callers