Files
zl_bench/ops/inventory_vllm.csv
AharrypotterandClaude Opus 5 35a928a516 chore: pin shipped text formats to LF
Everything in this repository runs on Linux accelerator hosts, and campaign
receipts record the SHA-256 of the harness files that produced them. With Git's
default autocrlf on Windows a checkout rewrites the shell drivers with CRLF,
which breaks them outright on the target host and changes every recorded digest.

The two inventory CSVs are renormalized here because their content is unchanged;
the remaining files normalize as part of the commits that actually modify them.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-09 16:38:33 +08:00

11 KiB

1repoopop_filefilemarked_skipcollectedop_names
2vllmperf_chunk_gated_delta_rule_fwdFLA/test_chunk_gated_delta_rule_fwdbenchmark/test_FLA/test_chunk_gated_delta_rule_fwd.pyyeschunk_gated_delta_rule_fwd
3vllmperf_chunk_gated_delta_ruleFLA/test_chunk_gated_delta_rule_perfbenchmark/test_FLA/test_chunk_gated_delta_rule_perf.pyyeschunk_gated_delta_rule
4vllmchunk_gdn2FLA/test_chunk_gdn2benchmark/test_FLA/test_chunk_gdn2.pyyeschunk_gdn2
5vllmperf_chunk_glaFLA/test_chunk_gla_perfbenchmark/test_FLA/test_chunk_gla_perf.pyyes
6vllmchunk_kdaFLA/test_chunk_kdabenchmark/test_FLA/test_chunk_kda.pyyeschunk_kda
7vllmperf_fused_recurrent_gated_delta_ruleFLA/test_fused_recurrent_gated_delta_rule_perfbenchmark/test_FLA/test_fused_recurrent_gated_delta_rule_perf.pyyesfused_recurrent_gated_delta_rule
8vllmact_quant_perfact_quantbenchmark/test_act_quant.pyyesact_quant_triton
9vllmadd_rms_normadd_rms_normbenchmark/test_add_rms_norm.pyyesadd_rms_norm
10vllmapply_repetition_penaltiesapply_repetition_penaltiesbenchmark/test_apply_repetition_penalties.pyyesapply_repetition_penalties
11vllmapply_rotary_pos_embapply_rotary_pos_embbenchmark/test_apply_rotary_pos_emb.pyyesapply_rotary_pos_emb
12vllmbeam_search_scorebeam_search_scorebenchmark/test_beam_search_score.pyyesbeam_search_score
13vllmbeam_search_score_beam_search_scorebenchmark/test_beam_search_score.pyyesbeam_search_score_
14vllmbincountbincountbenchmark/test_bincount.pyyesbincount
15vllmbincount_weightedbincountbenchmark/test_bincount.pyyesbincount_weighted
16vllmblas_benchmarkblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyes
17vllmperf_w8a8_block_fp8_matmulblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyesw8a8_block_fp8_matmul
18vllmperf_w8a8_block_fp8_matmul_deepgemmblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyesw8a8_block_fp8_matmul_deepgemm
19vllmperf_sparse_attentionblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyessparse_attention
20vllmmv_and_outer_benchmarkblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyes
21vllmaddmv_benchmarkblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyes
22vllmvdot_benchmarkblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyesvdot
23vllmaddr_benchmarkblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyesaddr
24vllmperf_router_gemmblas_perf_parallelbenchmark/test_blas_perf_parallel.pyyesrouter_gemm
25vllmconcat_and_cache_mlaconcat_and_cache_mlabenchmark/test_concat_and_cache_mla.pyyesconcat_and_cache_mla
26vllmcp_gather_indexer_k_quant_cache_benchmarkcp_gather_indexer_k_quant_cachebenchmark/test_cp_gather_indexer_k_quant_cache.pyyes
27vllmcross_entropy_losscross_entropy_lossbenchmark/test_cross_entropy_loss.pyyescross_entropy_loss
28vllmcutlass_scaled_mm_benchmarkcutlass_scaled_mmbenchmark/test_cutlass_scaled_mm.pyyes
29vllmcombine_topk_swa_indices_benchmarkdeepseek_v4_attention_combine_topk_swa_indicesbenchmark/test_deepseek_v4_attention_combine_topk_swa_indices.pyyes
30vllmcompute_global_topk_indices_and_lens_benchmarkdeepseek_v4_attention_compute_global_topk_indices_and_lensbenchmark/test_deepseek_v4_attention_compute_global_topk_indices_and_lens.pyyes
31vllmdequantize_and_gather_k_cache_benchmarkdeepseek_v4_attention_dequantize_and_gather_k_cachebenchmark/test_deepseek_v4_attention_dequantize_and_gather_k_cache.pyyes
32vllmfused_q_kv_rmsnorm_benchmarkdeepseek_v4_attention_fused_q_kv_rmsnormbenchmark/test_deepseek_v4_attention_fused_q_kv_rmsnorm.pyyes
33vllmdgegludgeglubenchmark/test_dgeglu.pyyesdgeglu
34vllmdregludreglubenchmark/test_dreglu.pyyesdreglu
35vllmdswigludswiglubenchmark/test_dswiglu.pyyesdswiglu
36vllmflash_attention_forwardflash_attention_forwardbenchmark/test_flash_attention_forward.pyyesflash_attention_forward
37vllmflash_attn_varlen_funcflash_attn_varlen_funcbenchmark/test_flash_attn_varlen_func.pyyesflash_attn_varlen_func
38vllmflash_attn_varlen_opt_funcflash_attn_varlen_opt_init_funcbenchmark/test_flash_attn_varlen_opt_init_func.pyyesflash_attn_varlen_func
39vllmflash_mlaflash_mlabenchmark/test_flash_mla.pyyesflash_mla
40vllmflash_mla_sparse_fwdflash_mla_sparse_fwdbenchmark/test_flash_mla_sparse_fwd.pyyes
41vllmflash_mla_with_kvcacheflash_mla_with_kvcachebenchmark/test_flash_mla_with_kvcache.pyyes
42vllmfp8_fp4_mqa_logitsfp8_fp4_mqa_logitsbenchmark/test_fp8_fp4_mqa_logits.pyyesfp8_fp4_mqa_logits
43vllmfp8_fp4_paged_mqa_logitsfp8_fp4_paged_mqa_logitsbenchmark/test_fp8_fp4_paged_mqa_logits.pyyesfp8_fp4_paged_mqa_logits
44vllmfused_add_rms_normfused_add_rms_normbenchmark/test_fused_add_rms_norm.pyyesfused_add_rms_norm
45vllmfused_deepseek_v4_qnorm_rope_kv_rope_quant_insertfused_deepseek_v4_qnorm_rope_kv_rope_quant_insertbenchmark/test_fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert.pyyes
46vllmfused_indexer_q_rope_quantfused_indexer_q_rope_quantbenchmark/test_fused_indexer_q_rope_quant.pyyesfused_indexer_q_rope_quant
47vllmfused_inv_rope_fp8_quantfused_inv_rope_fp8_quantbenchmark/test_fused_inv_rope_fp8_quant.pyyesfused_inv_rope_fp8_quant
48vllmfused_marlin_moefused_marlin_moebenchmark/test_fused_marlin_moe.pyyesfused_marlin_moe
49vllmfused_moe_impl_gems_vs_vllmfused_moebenchmark/test_fused_moe.pyyesfused_experts_impl
50vllmfused_moe_fp8fused_moe_fp8benchmark/test_fused_moe_fp8.pyyesfused_experts_impl
51vllmfused_moe_fp8_blockwisefused_moe_fp8_blockwisebenchmark/test_fused_moe_fp8_blockwise.pyyesfused_experts_impl
52vllmfused_moe_int4_w4a16fused_moe_int4_w4a16benchmark/test_fused_moe_int4_w4a16.pyyesfused_experts_impl
53vllmfused_experts_impl_int8fused_moe_int8benchmark/test_fused_moe_int8.pyyesfused_experts_impl
54vllmfused_experts_impl_int8_w8a16fused_moe_int8_w8a16benchmark/test_fused_moe_int8_w8a16.pyyesfused_moe_int8_w8a16_gems_vs_bf16_deq
55vllmfused_moe_w8a16_mxqfused_moe_w8a16benchmark/test_fused_moe_w8a16.pyyesfused_moe_w8a16_mxq_gems_vs_bf16_deq
56vllmfused_moe_w8a16_mxq_gems_vs_vllmfused_moe_w8a16benchmark/test_fused_moe_w8a16.pyyesfused_moe_w8a16_mxq_gems_vs_vllm
57vllmgeglugeglubenchmark/test_geglu.pyyesgeglu
58vllmgelu_and_mulgelu_and_mulbenchmark/test_gelu_and_mul.pyyesgelu_and_mul
59vllmgrouped_topk_no_renormgrouped_topkbenchmark/test_grouped_topk.pyyesgrouped_topk
60vllmgrouped_topk_score_0grouped_topkbenchmark/test_grouped_topk.pyyesgrouped_topk
61vllmgrouped_topk_score_1grouped_topkbenchmark/test_grouped_topk.pyyesgrouped_topk
62vllmindexer_k_quant_and_cache_benchmarkindexer_k_quant_and_cachebenchmark/test_indexer_k_quant_and_cache.pyyes
63vllminplace_fused_experts_gems_vs_vllminplace_fused_expertsbenchmark/test_inplace_fused_experts.pyyesinplace_fused_experts
64vllminstance_norminstance_normbenchmark/test_instance_norm.pyyesinstance_norm
65vllmmhc_postmhcbenchmark/test_mhc.pyyesmhc_post
66vllmmhc_premhcbenchmark/test_mhc.pyyesmhc_pre
67vllmhc_split_sinkhorn_forwardmhcbenchmark/test_mhc.pyyeshc_split_sinkhorn_forward
68vllmmhc_bwdmhcbenchmark/test_mhc.pyyesmhc_bwd
69vllmhc_head_fused_kernelmhcbenchmark/test_mhc.pyyeshc_head_fused_kernel
70vllmmoe_align_block_size_tritonmoe_align_block_size_tritonbenchmark/test_moe_align_block_size_triton.pyyesmoe_align_block_size_triton
71vllmmoe_summoe_sumbenchmark/test_moe_sum.pyyesmoe_sum
72vllmmrope_gems_vs_torchmropebenchmark/test_mrope.pyyesmrope
73vllmouterouterbenchmark/test_outer.pyyesouter
74vllmoutplace_fused_experts_gems_vs_vllmoutplace_fused_expertsbenchmark/test_outplace_fused_experts.pyyesoutplace_fused_experts
75vllmpack_seqpack_seqbenchmark/test_pack_seq.pyyespack_seq_triton
76vllmpack_seq_fp8pack_seqbenchmark/test_pack_seq.pyyespack_seq_triton
77vllmperf_parallel_nsaparallel_nsabenchmark/test_parallel_nsa.pyyesparallel_nsa
78vllmperf_parallel_nsa_compressionparallel_nsa_compressionbenchmark/test_parallel_nsa_compression.pyyesparallel_nsa_compression
79vllmper_token_group_quant_fp8per_token_group_quant_fp8benchmark/test_per_token_group_quant_fp8.pyyesper_token_group_quant_fp8
80vllmpersistent_topkpersistent_topkbenchmark/test_persistent_topk.pyyespersistent_topk
81vllmreglureglubenchmark/test_reglu.pyyesreglu
82vllmreshape_and_cachereshape_and_cachebenchmark/test_reshape_and_cache.pyyesreshape_and_cache
83vllmreshape_and_cache_flashreshape_and_cache_flashbenchmark/test_reshape_and_cache_flash.pyyesreshape_and_cache_flash
84vllmperf_router_gemmrouter_gemmbenchmark/test_router_gemm.pyyesrouter_gemm
85vllmrwkv_ka_fusionrwkv_ka_fusionbenchmark/test_rwkv_ka_fusion.pyyesrwkv_ka_fusion
86vllmrwkv_mm_sparsityrwkv_mm_sparsitybenchmark/test_rwkv_mm_sparsity.pyyesrwkv_mm_sparsity
87vllmdynamic_scaled_int8_quantscaled_int8_quantbenchmark/test_scaled_int8_quant.pyyesdynamic_scaled_int8_quant
88vllmstatic_scaled_int8_quantscaled_int8_quantbenchmark/test_scaled_int8_quant.pyyesstatic_scaled_int8_quant
89vllmsilu_and_mulsilu_and_mulbenchmark/test_silu_and_mul.pyyessilu_and_mul
90vllmsilu_and_mul_outsilu_and_mulbenchmark/test_silu_and_mul.pyyessilu_and_mul_out
91vllmsilu_and_mul_with_clampsilu_and_mul_with_clampbenchmark/test_silu_and_mul_with_clamp.pyyessilu_and_mul_with_clamp
92vllmsilu_and_mul_with_clamp_outsilu_and_mul_with_clampbenchmark/test_silu_and_mul_with_clamp.pyyessilu_and_mul_with_clamp_out
93vllmskip_layernormskip_layer_normbenchmark/test_skip_layer_norm.pyyesskip_layer_norm
94vllmsparse_attn_tritonsparse_attentionbenchmark/test_sparse_attention.pyyesyessparse_attention
95vllmsparse_mla_fwd_interfacesparse_mla_fwd_interfacebenchmark/test_sparse_mla_fwd_interface.pyyessparse_mla_fwd_interface
96vllmstage_deepseek_v4_mega_moe_inputs_benchmarkstage_deepseek_v4_mega_moe_inputsbenchmark/test_stage_deepseek_v4_mega_moe_inputs.pyyes
97vllmswigluswiglubenchmark/test_swiglu.pyyesswiglu
98vllmtop_k_per_row_decodetop_k_per_row_decodebenchmark/test_top_k_per_row_decode.pyyestop_k_per_row_decode
99vllmtop_k_per_row_prefilltop_k_per_row_prefillbenchmark/test_top_k_per_row_prefill.pyyestop_k_per_row_prefill
100vllmtopk_softmaxtopk_softmaxbenchmark/test_topk_softmax.pyyestopk_softmax
101vllmtopk_softplus_sqrttopk_softplus_sqrtbenchmark/test_topk_softplus_sqrt.pyyestopk_softplus_sqrt
102vllmtriton_scaled_mm_benchmarktriton_scaled_mmbenchmark/test_triton_scaled_mm.pyyestriton_scaled_mm
103vllmtriton_unified_attention_perftriton_unified_attention_perfbenchmark/test_triton_unified_attention_perf.pyyes
104vllmunpack_sequnpack_seqbenchmark/test_unpack_seq.pyyesunpack_seq_triton
105vllmunpack_seq_fp8unpack_seqbenchmark/test_unpack_seq.pyyesunpack_seq_triton
106vllmweight_norm_dim0weight_normbenchmark/test_weight_norm.pyyesweight_norm
107vllmweight_norm_dim_lastweight_normbenchmark/test_weight_norm.pyyesweight_norm
108vllmweight_norm_interfaceweight_norm_interfacebenchmark/test_weight_norm_interface.pyyesweight_norm_interface
109vllmweight_norm_interface_backwardweight_norm_interfacebenchmark/test_weight_norm_interface.pyyesweight_norm_interface_backward