deepseek-ai/FlashMLA
File Size
The distribution of size of files (measured in lines of code).
File Size Overall
  • There are 97 files with 17,763 lines of code.
    • 4 very long files (5,848 lines of code)
    • 6 long files (4,483 lines of code)
    • 11 medium size files (3,318 lines of code)
    • 19 small files (2,776 lines of code)
    • 57 very small files (1,338 lines of code)
32% | 25% | 18% | 15% | 7%
Legend:
1001+
501-1000
201-500
101-200
1-100


explore: grouped by folders | grouped by size | sunburst | 3D view
File Size per Extension
1001+
501-1000
201-500
101-200
1-100
hpp39% | 22% | 20% | 12% | 5%
cuh31% | 56% | 5% | 5% | 1%
h43% | 0% | 6% | 40% | 8%
cpp0% | 0% | 78% | 17% | 3%
py0% | 0% | 69% | 0% | 30%
cu0% | 0% | 27% | 17% | 54%
File Size per Logical Decomposition
primary
1001+
501-1000
201-500
101-200
1-100
csrc33% | 25% | 17% | 16% | 6%
flash_mla0% | 0% | 76% | 0% | 23%
scripts0% | 0% | 0% | 0% | 100%
Longest Files (Top 50)
File# lines# units
kernel.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn
 1578 -
kernel_body.h
in csrc/ascend_kernels/prefill/sparse
 1440 16
sm100_fmha_bwd_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 1416 16
sm100_fmha_bwd_mla_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 1414 16
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128
 974 -
sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 814 20
sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 805 20
kernel.cuh
in csrc/cuda_kernels/sm100/decode/sparse/head64
 667 -
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128
 625 -
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64
 598 -
sm100_fmha_fwd_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 491 11
462 6
396 9
321 11
fmha_cutlass_fwd_sm100.cuh
in csrc/cuda_kernels/sm100/prefill/dense
 278 -
fmha_fusion.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 277 21
fmha_device_bwd.hpp
in csrc/cuda_kernels/sm100/prefill/dense/device
 237 3
common.h
in csrc/api
 223 14
sm100_fmha_mla_load_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 220 5
combine.cu
in csrc/cuda_kernels/smxx/decode/combine
 208 -
sm100_fmha_load_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 205 4
config.h
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn
 198 4
helpers.h
in csrc/cuda_kernels/sm100
 196 9
195 6
pipeline_mla.hpp
in csrc/cuda_kernels/sm100/prefill/dense/common
 174 12
fmha_cutlass_bwd_sm100.cuh
in csrc/cuda_kernels/sm100/prefill/dense
 171 -
fmha.hpp
in csrc/cuda_kernels/sm100/prefill/dense/device
 168 3
common_subroutine.h
in csrc/cuda_kernels/sm100
 168 5
config.h
in csrc/cuda_kernels/sm100/decode/sparse/head64
 151 -
sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 148 6
gather_tensor.hpp
in csrc/cuda_kernels/sm100/prefill/dense/common
 142 3
config.h
in csrc/ascend_kernels/prefill/sparse
 139 1
get_decoding_sched_meta.cu
in csrc/cuda_kernels/smxx/decode/get_decoding_sched_meta
 132 -
params.h
in csrc
 129 -
fmha_causal_tile_scheduler.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 117 8
config.h
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64
 115 -
config.h
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128
 113 -
config.h
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128
 107 -
dequant_utils.cuh
in csrc/cuda_kernels/sm100
 107 -
fmha_kernel_bwd_sum_OdO.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 106 7
fmha_kernel_bwd_convert.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 100 8
fmha_tile_scheduler.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 96 8
kernel.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_wv_proj
 88 -
fmha_common.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 84 5
fmha_cutlass_fwd_sm100.cu
in csrc/cuda_kernels/sm100/prefill/dense
 77 -
77 4
fmha_cutlass_bwd_sm100.cu
in csrc/cuda_kernels/sm100/prefill/dense
 76 -
kv_cache_utils.cuh
in csrc/cuda_kernels/sm100
 57 -
kernel.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_q_b_proj
 56 -
pow_2.hpp
in csrc/cuda_kernels/sm100/prefill/dense/common
 45 3
Large Files That Change Often (Top 20)
To change a file, an AI coding agent (or a person) has to read it: these files are read the most for the changes in the past year (lines × changes).
File# lines≈ tokens
to read		# changes
(1y)		≈ lines edited
per change		lines read per
line changed
kernel.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn
 1578 ~10k–30k 1 ~1,500 ~1
kernel_body.h
in csrc/ascend_kernels/prefill/sparse
 1440 ~10k–30k 1 ~1,500 ~1
sm100_fmha_bwd_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 1416 ~7k–20k 1 ~2,000 ~1
sm100_fmha_bwd_mla_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 1414 ~7k–20k 1 ~2,000 ~1
321 ~2k–5k 4 ~100 ~3
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128
 974 ~5k–15k 1 ~1,000 ~1
462 ~3k–7k 2 ~300 ~2
common.h
in csrc/api
 223 ~1.5k–5k 4 ~100 ~2
sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 814 ~5k–15k 1 ~1,000 ~1
sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 805 ~5k–15k 1 ~1,000 ~1
396 ~2k–7k 2 ~500 ~1
kernel.cuh
in csrc/cuda_kernels/sm100/decode/sparse/head64
 667 ~3k–10k 1 ~700 ~1
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128
 625 ~3k–10k 1 ~700 ~1
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64
 598 ~3k–10k 1 ~700 ~1
sm100_fmha_fwd_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 491 ~3k–7k 1 ~700 ~1
195 ~1k–3k 2 ~200 ~1
params.h
in csrc
 129 ~700–2k 3 ~150 ~1
fmha_cutlass_fwd_sm100.cuh
in csrc/cuda_kernels/sm100/prefill/dense
 278 ~1.5k–5k 1 ~300 ~1
fmha_fusion.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 277 ~1.5k–5k 1 ~300 ~1
fmha_device_bwd.hpp
in csrc/cuda_kernels/sm100/prefill/dense/device
 237 ~1.5k–5k 1 ~300 ~1

Tokens are lines × 7 to 14 (analysis.fileReadsForChanges). Lines edited per change are the file's lines added and deleted divided by its commits over the whole history. Agents read about 2,000 lines at a time: when a 17,000-line file was split into modules, the tokens an agent read for the same change fell by 83% (The Economic Benefit of Refactoring, 2026).

Files With Most Units (Top 36)
File# lines# units
fmha_fusion.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 277 21
sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 814 20
sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 805 20
sm100_fmha_bwd_mla_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 1414 16
sm100_fmha_bwd_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 1416 16
kernel_body.h
in csrc/ascend_kernels/prefill/sparse
 1440 16
common.h
in csrc/api
 223 14
pipeline_mla.hpp
in csrc/cuda_kernels/sm100/prefill/dense/common
 174 12
sm100_fmha_fwd_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 491 11
321 11
396 9
helpers.h
in csrc/cuda_kernels/sm100
 196 9
fmha_tile_scheduler.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 96 8
fmha_causal_tile_scheduler.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 117 8
fmha_kernel_bwd_convert.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 100 8
fmha_kernel_bwd_sum_OdO.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 106 7
462 6
195 6
sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 148 6
fmha_common.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 84 5
sm100_fmha_mla_load_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 220 5
common_subroutine.h
in csrc/cuda_kernels/sm100
 168 5
utils.h
in csrc/cuda_kernels
 37 4
sm100_fmha_load_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 205 4
config.h
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn
 198 4
77 4
pow_2.hpp
in csrc/cuda_kernels/sm100/prefill/dense/common
 45 3
gather_tensor.hpp
in csrc/cuda_kernels/sm100/prefill/dense/common
 142 3
fmha_device_bwd.hpp
in csrc/cuda_kernels/sm100/prefill/dense/device
 237 3
fmha.hpp
in csrc/cuda_kernels/sm100/prefill/dense/device
 168 3
kv_cache_format.h
in csrc/cuda_kernels
 45 2
fast_div_mod.h
in csrc/ascend_kernels/prefill/sparse
 28 2
45 2
dense_fwd.cpp
in csrc/api
 13 1
dense_bwd.cpp
in csrc/api
 13 1
config.h
in csrc/ascend_kernels/prefill/sparse
 139 1
Files With Long Lines (Top 50)

There are 62 files with lines longer than 120 characters. In total, there are 563 long lines.

File# lines# units# long lines
kernel.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn
 1578 - 137
kernel_body.h
in csrc/ascend_kernels/prefill/sparse
 1440 16 51
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128
 974 - 49
kernel.cuh
in csrc/cuda_kernels/sm100/decode/sparse/head64
 667 - 40
462 6 23
sm100_fmha_bwd_mla_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 1414 16 23
sm100_fmha_bwd_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 1416 16 21
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64
 598 - 20
dequant_utils.cuh
in csrc/cuda_kernels/sm100
 107 - 18
396 9 16
phase1.cuh
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128
 625 - 16
321 11 13
sm100_fmha_fwd_kernel_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 491 11 11
combine.cu
in csrc/cuda_kernels/smxx/decode/combine
 208 - 9
pipeline_mla.hpp
in csrc/cuda_kernels/sm100/prefill/dense/common
 174 12 9
common.h
in csrc/api
 223 14 7
sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 814 20 7
sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 805 20 7
config.h
in csrc/ascend_kernels/prefill/sparse
 139 1 7
kernel.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_wv_proj
 88 - 6
fmha_kernel_bwd_sum_OdO.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 106 7 5
fmha_kernel_bwd_convert.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 100 8 5
config.h
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128
 113 - 5
sm100_fmha_mla_load_tma_warpspecialized.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 220 5 4
config.h
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn
 198 4 4
config.h
in csrc/cuda_kernels/sm100/decode/sparse/head64
 151 - 4
kv_cache_utils.cuh
in csrc/cuda_kernels/sm100
 57 - 3
fmha_device_bwd.hpp
in csrc/cuda_kernels/sm100/prefill/dense/device
 237 3 3
kernel.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_q_b_proj
 56 - 3
195 6 2
fmha_fusion.hpp
in csrc/cuda_kernels/sm100/prefill/dense/collective
 277 21 2
fmha_cutlass_bwd_sm100.cu
in csrc/cuda_kernels/sm100/prefill/dense
 76 - 2
interface.h
in csrc/cuda_kernels/sm100/prefill/dense
 11 - 2
params.h
in csrc
 129 - 1
get_decoding_sched_meta.cu
in csrc/cuda_kernels/smxx/decode/get_decoding_sched_meta
 132 - 1
fmha_cutlass_bwd_sm100.cuh
in csrc/cuda_kernels/sm100/prefill/dense
 171 - 1
fmha_tile_scheduler.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 96 8 1
fmha_causal_tile_scheduler.hpp
in csrc/cuda_kernels/sm100/prefill/dense/kernel
 117 8 1
config.h
in csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128
 107 - 1
phase1_decode_k512_v41.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/instantiations
 5 - 1
phase1_decode_k512_v41fp4_splitkv.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/instantiations
 5 - 1
phase1_decode_k512_v41_splitkv.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/instantiations
 5 - 1
phase1_k512.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/instantiations
 5 - 1
phase1_decode_k512_v41fp4.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/instantiations
 5 - 1
v41_h64_decode_nonorm.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/instantiations
 5 - 1
v41_h128_decode_norm.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/instantiations
 5 - 1
v41fp4_h128_decode_nonorm.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/instantiations
 5 - 1
v41_h64_decode_norm.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/instantiations
 5 - 1
v41_h128_prefill_nonorm.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/instantiations
 5 - 1
v41_h128_decode_nonorm.cu
in csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/instantiations
 5 - 1
Correlations

File Size vs. Commits (all time): 97 points

csrc/api/api.cpp x: 3 commits (all time) y: 14 lines of code csrc/api/common.h x: 4 commits (all time) y: 223 lines of code csrc/api/dense_bwd.cpp x: 2 commits (all time) y: 13 lines of code csrc/api/fused_norm_rope_attn_rope_cast_fwd.cpp x: 2 commits (all time) y: 462 lines of code csrc/api/sparse_decode.cpp x: 2 commits (all time) y: 396 lines of code csrc/api/sparse_prefill.cpp x: 2 commits (all time) y: 195 lines of code csrc/ascend_kernels/prefill/sparse/config.h x: 1 commits (all time) y: 139 lines of code csrc/ascend_kernels/prefill/sparse/fast_div_mod.h x: 1 commits (all time) y: 28 lines of code csrc/ascend_kernels/prefill/sparse/kernel.h x: 1 commits (all time) y: 16 lines of code csrc/ascend_kernels/prefill/sparse/kernel_body.h x: 1 commits (all time) y: 1440 lines of code csrc/cuda_kernels/defines.h x: 1 commits (all time) y: 9 lines of code csrc/cuda_kernels/kv_cache_format.h x: 1 commits (all time) y: 45 lines of code csrc/cuda_kernels/sm100/common_subroutine.h x: 1 commits (all time) y: 168 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/config.h x: 1 commits (all time) y: 151 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/instantiations/v41_h64.cu x: 1 commits (all time) y: 5 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/kernel.cuh x: 1 commits (all time) y: 667 lines of code csrc/cuda_kernels/sm100/dequant_utils.cuh x: 1 commits (all time) y: 107 lines of code csrc/cuda_kernels/sm100/helpers.h x: 1 commits (all time) y: 196 lines of code csrc/cuda_kernels/sm100/kv_cache_utils.cuh x: 1 commits (all time) y: 57 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_common.hpp x: 1 commits (all time) y: 84 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_fusion.hpp x: 1 commits (all time) y: 277 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp x: 1 commits (all time) y: 148 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp x: 1 commits (all time) y: 805 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_load_tma_warpspecialized.hpp x: 1 commits (all time) y: 205 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp x: 1 commits (all time) y: 814 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_load_tma_warpspecialized.hpp x: 1 commits (all time) y: 220 lines of code csrc/cuda_kernels/sm100/prefill/dense/common/pipeline_mla.hpp x: 1 commits (all time) y: 174 lines of code csrc/cuda_kernels/sm100/prefill/dense/device/fmha_device_bwd.hpp x: 1 commits (all time) y: 237 lines of code csrc/cuda_kernels/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu x: 1 commits (all time) y: 76 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_causal_tile_scheduler.hpp x: 1 commits (all time) y: 117 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_kernel_bwd_convert.hpp x: 1 commits (all time) y: 100 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_options.hpp x: 1 commits (all time) y: 38 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_tile_scheduler.hpp x: 1 commits (all time) y: 96 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_bwd_kernel_tma_warpspecialized.hpp x: 1 commits (all time) y: 1416 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp x: 1 commits (all time) y: 491 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.cuh x: 1 commits (all time) y: 1578 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.h x: 1 commits (all time) y: 34 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_q_b_proj/kernel.h x: 1 commits (all time) y: 21 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_wv_proj/kernel.cu x: 1 commits (all time) y: 88 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128/phase1.cuh x: 1 commits (all time) y: 625 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64/phase1.cuh x: 1 commits (all time) y: 598 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/phase1.cuh x: 1 commits (all time) y: 974 lines of code csrc/cuda_kernels/smxx/decode/combine/combine.cu x: 1 commits (all time) y: 208 lines of code csrc/cuda_kernels/smxx/decode/get_decoding_sched_meta/get_decoding_sched_meta.cu x: 1 commits (all time) y: 132 lines of code csrc/params.h x: 5 commits (all time) y: 129 lines of code flash_mla/flash_mla_interface.py x: 11 commits (all time) y: 321 lines of code flash_mla/fused_norm_rope_attn_rope_cast.py x: 2 commits (all time) y: 77 lines of code flash_mla/__init__.py x: 5 commits (all time) y: 19 lines of code
1578.0
lines of code
  min: 2.0
  average: 183.12
  25th percentile: 6.0
  median: 56.0
  75th percentile: 184.5
  max: 1578.0
0 11.0
commits (all time)
min: 1.0 | average: 1.3 | 25th percentile: 1.0 | median: 1.0 | 75th percentile: 1.0 | max: 11.0

File Size vs. Contributors (all time): 97 points

csrc/api/api.cpp x: 1 contributors (all time) y: 14 lines of code csrc/api/common.h x: 2 contributors (all time) y: 223 lines of code csrc/api/fused_norm_rope_attn_rope_cast_fwd.cpp x: 1 contributors (all time) y: 462 lines of code csrc/api/sparse_decode.cpp x: 1 contributors (all time) y: 396 lines of code csrc/api/sparse_prefill.cpp x: 1 contributors (all time) y: 195 lines of code csrc/ascend_kernels/prefill/sparse/config.h x: 1 contributors (all time) y: 139 lines of code csrc/ascend_kernels/prefill/sparse/fast_div_mod.h x: 1 contributors (all time) y: 28 lines of code csrc/ascend_kernels/prefill/sparse/kernel_body.h x: 1 contributors (all time) y: 1440 lines of code csrc/cuda_kernels/defines.h x: 1 contributors (all time) y: 9 lines of code csrc/cuda_kernels/kv_cache_format.h x: 1 contributors (all time) y: 45 lines of code csrc/cuda_kernels/sm100/common_subroutine.h x: 1 contributors (all time) y: 168 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/config.h x: 1 contributors (all time) y: 151 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/instantiations/v41_h64.cu x: 1 contributors (all time) y: 5 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/kernel.cuh x: 1 contributors (all time) y: 667 lines of code csrc/cuda_kernels/sm100/dequant_utils.cuh x: 1 contributors (all time) y: 107 lines of code csrc/cuda_kernels/sm100/kv_cache_utils.cuh x: 1 contributors (all time) y: 57 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_common.hpp x: 1 contributors (all time) y: 84 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_fusion.hpp x: 1 contributors (all time) y: 277 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp x: 1 contributors (all time) y: 148 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp x: 1 contributors (all time) y: 805 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_load_tma_warpspecialized.hpp x: 1 contributors (all time) y: 205 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp x: 1 contributors (all time) y: 814 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_load_tma_warpspecialized.hpp x: 1 contributors (all time) y: 220 lines of code csrc/cuda_kernels/sm100/prefill/dense/common/pipeline_mla.hpp x: 1 contributors (all time) y: 174 lines of code csrc/cuda_kernels/sm100/prefill/dense/device/fmha_device_bwd.hpp x: 1 contributors (all time) y: 237 lines of code csrc/cuda_kernels/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu x: 1 contributors (all time) y: 76 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_causal_tile_scheduler.hpp x: 1 contributors (all time) y: 117 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_kernel_bwd_convert.hpp x: 1 contributors (all time) y: 100 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_options.hpp x: 1 contributors (all time) y: 38 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_tile_scheduler.hpp x: 1 contributors (all time) y: 96 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_bwd_kernel_tma_warpspecialized.hpp x: 1 contributors (all time) y: 1416 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp x: 1 contributors (all time) y: 491 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.cuh x: 1 contributors (all time) y: 1578 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.h x: 1 contributors (all time) y: 34 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_q_b_proj/kernel.h x: 1 contributors (all time) y: 21 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_wv_proj/kernel.cu x: 1 contributors (all time) y: 88 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128/phase1.cuh x: 1 contributors (all time) y: 625 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64/phase1.cuh x: 1 contributors (all time) y: 598 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/phase1.cuh x: 1 contributors (all time) y: 974 lines of code csrc/cuda_kernels/smxx/decode/combine/combine.cu x: 1 contributors (all time) y: 208 lines of code csrc/cuda_kernels/smxx/decode/get_decoding_sched_meta/get_decoding_sched_meta.cu x: 1 contributors (all time) y: 132 lines of code csrc/params.h x: 1 contributors (all time) y: 129 lines of code flash_mla/flash_mla_interface.py x: 6 contributors (all time) y: 321 lines of code flash_mla/__init__.py x: 2 contributors (all time) y: 19 lines of code
1578.0
lines of code
  min: 2.0
  average: 183.12
  25th percentile: 6.0
  median: 56.0
  75th percentile: 184.5
  max: 1578.0
0 6.0
contributors (all time)
min: 1.0 | average: 1.07 | 25th percentile: 1.0 | median: 1.0 | 75th percentile: 1.0 | max: 6.0

File Size vs. Commits (30 days): 97 points

csrc/api/api.cpp x: 2 commits (30d) y: 14 lines of code csrc/api/common.h x: 2 commits (30d) y: 223 lines of code csrc/api/fused_norm_rope_attn_rope_cast_fwd.cpp x: 2 commits (30d) y: 462 lines of code csrc/api/sparse_decode.cpp x: 2 commits (30d) y: 396 lines of code csrc/api/sparse_prefill.cpp x: 2 commits (30d) y: 195 lines of code csrc/ascend_kernels/prefill/sparse/config.h x: 1 commits (30d) y: 139 lines of code csrc/ascend_kernels/prefill/sparse/fast_div_mod.h x: 1 commits (30d) y: 28 lines of code csrc/ascend_kernels/prefill/sparse/kernel.h x: 1 commits (30d) y: 16 lines of code csrc/ascend_kernels/prefill/sparse/kernel_body.h x: 1 commits (30d) y: 1440 lines of code csrc/cuda_kernels/defines.h x: 1 commits (30d) y: 9 lines of code csrc/cuda_kernels/kv_cache_format.h x: 1 commits (30d) y: 45 lines of code csrc/cuda_kernels/sm100/common_subroutine.h x: 1 commits (30d) y: 168 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/config.h x: 1 commits (30d) y: 151 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/instantiations/v41_h64.cu x: 1 commits (30d) y: 5 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/kernel.cuh x: 1 commits (30d) y: 667 lines of code csrc/cuda_kernels/sm100/dequant_utils.cuh x: 1 commits (30d) y: 107 lines of code csrc/cuda_kernels/sm100/helpers.h x: 1 commits (30d) y: 196 lines of code csrc/cuda_kernels/sm100/kv_cache_utils.cuh x: 1 commits (30d) y: 57 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_common.hpp x: 1 commits (30d) y: 84 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_fusion.hpp x: 1 commits (30d) y: 277 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp x: 1 commits (30d) y: 148 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp x: 1 commits (30d) y: 805 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_load_tma_warpspecialized.hpp x: 1 commits (30d) y: 205 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp x: 1 commits (30d) y: 814 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_load_tma_warpspecialized.hpp x: 1 commits (30d) y: 220 lines of code csrc/cuda_kernels/sm100/prefill/dense/common/pipeline_mla.hpp x: 1 commits (30d) y: 174 lines of code csrc/cuda_kernels/sm100/prefill/dense/device/fmha_device_bwd.hpp x: 1 commits (30d) y: 237 lines of code csrc/cuda_kernels/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu x: 1 commits (30d) y: 76 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_causal_tile_scheduler.hpp x: 1 commits (30d) y: 117 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_kernel_bwd_convert.hpp x: 1 commits (30d) y: 100 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_options.hpp x: 1 commits (30d) y: 38 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_tile_scheduler.hpp x: 1 commits (30d) y: 96 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_bwd_kernel_tma_warpspecialized.hpp x: 1 commits (30d) y: 1416 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp x: 1 commits (30d) y: 491 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.cuh x: 1 commits (30d) y: 1578 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.h x: 1 commits (30d) y: 34 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_q_b_proj/kernel.h x: 1 commits (30d) y: 21 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_wv_proj/kernel.cu x: 1 commits (30d) y: 88 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128/phase1.cuh x: 1 commits (30d) y: 625 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64/phase1.cuh x: 1 commits (30d) y: 598 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/phase1.cuh x: 1 commits (30d) y: 974 lines of code csrc/cuda_kernels/smxx/decode/combine/combine.cu x: 1 commits (30d) y: 208 lines of code csrc/cuda_kernels/smxx/decode/get_decoding_sched_meta/get_decoding_sched_meta.cu x: 1 commits (30d) y: 132 lines of code csrc/params.h x: 2 commits (30d) y: 129 lines of code flash_mla/flash_mla_interface.py x: 2 commits (30d) y: 321 lines of code flash_mla/fused_norm_rope_attn_rope_cast.py x: 2 commits (30d) y: 77 lines of code
1578.0
lines of code
  min: 2.0
  average: 183.12
  25th percentile: 6.0
  median: 56.0
  75th percentile: 184.5
  max: 1578.0
0 2.0
commits (30d)
min: 1.0 | average: 1.1 | 25th percentile: 1.0 | median: 1.0 | 75th percentile: 1.0 | max: 2.0

File Size vs. Contributors (30 days): 97 points

csrc/api/api.cpp x: 1 contributors (30d) y: 14 lines of code csrc/api/common.h x: 1 contributors (30d) y: 223 lines of code csrc/api/fused_norm_rope_attn_rope_cast_fwd.cpp x: 1 contributors (30d) y: 462 lines of code csrc/api/sparse_decode.cpp x: 1 contributors (30d) y: 396 lines of code csrc/api/sparse_prefill.cpp x: 1 contributors (30d) y: 195 lines of code csrc/ascend_kernels/prefill/sparse/config.h x: 1 contributors (30d) y: 139 lines of code csrc/ascend_kernels/prefill/sparse/fast_div_mod.h x: 1 contributors (30d) y: 28 lines of code csrc/ascend_kernels/prefill/sparse/kernel_body.h x: 1 contributors (30d) y: 1440 lines of code csrc/cuda_kernels/defines.h x: 1 contributors (30d) y: 9 lines of code csrc/cuda_kernels/kv_cache_format.h x: 1 contributors (30d) y: 45 lines of code csrc/cuda_kernels/sm100/common_subroutine.h x: 1 contributors (30d) y: 168 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/config.h x: 1 contributors (30d) y: 151 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/instantiations/v41_h64.cu x: 1 contributors (30d) y: 5 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/kernel.cuh x: 1 contributors (30d) y: 667 lines of code csrc/cuda_kernels/sm100/dequant_utils.cuh x: 1 contributors (30d) y: 107 lines of code csrc/cuda_kernels/sm100/kv_cache_utils.cuh x: 1 contributors (30d) y: 57 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_common.hpp x: 1 contributors (30d) y: 84 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_fusion.hpp x: 1 contributors (30d) y: 277 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp x: 1 contributors (30d) y: 148 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp x: 1 contributors (30d) y: 805 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_load_tma_warpspecialized.hpp x: 1 contributors (30d) y: 205 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp x: 1 contributors (30d) y: 814 lines of code csrc/cuda_kernels/sm100/prefill/dense/common/pipeline_mla.hpp x: 1 contributors (30d) y: 174 lines of code csrc/cuda_kernels/sm100/prefill/dense/device/fmha_device_bwd.hpp x: 1 contributors (30d) y: 237 lines of code csrc/cuda_kernels/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu x: 1 contributors (30d) y: 76 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_causal_tile_scheduler.hpp x: 1 contributors (30d) y: 117 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_kernel_bwd_convert.hpp x: 1 contributors (30d) y: 100 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_options.hpp x: 1 contributors (30d) y: 38 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_tile_scheduler.hpp x: 1 contributors (30d) y: 96 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_bwd_kernel_tma_warpspecialized.hpp x: 1 contributors (30d) y: 1416 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp x: 1 contributors (30d) y: 491 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.cuh x: 1 contributors (30d) y: 1578 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.h x: 1 contributors (30d) y: 34 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_q_b_proj/kernel.h x: 1 contributors (30d) y: 21 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_wv_proj/kernel.cu x: 1 contributors (30d) y: 88 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128/phase1.cuh x: 1 contributors (30d) y: 625 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64/phase1.cuh x: 1 contributors (30d) y: 598 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/phase1.cuh x: 1 contributors (30d) y: 974 lines of code csrc/cuda_kernels/smxx/decode/combine/combine.cu x: 1 contributors (30d) y: 208 lines of code csrc/cuda_kernels/smxx/decode/get_decoding_sched_meta/get_decoding_sched_meta.cu x: 1 contributors (30d) y: 132 lines of code csrc/params.h x: 1 contributors (30d) y: 129 lines of code flash_mla/flash_mla_interface.py x: 1 contributors (30d) y: 321 lines of code
1578.0
lines of code
  min: 2.0
  average: 183.12
  25th percentile: 6.0
  median: 56.0
  75th percentile: 184.5
  max: 1578.0
0 1.0
contributors (30d)
min: 1.0 | average: 1.0 | 25th percentile: 1.0 | median: 1.0 | 75th percentile: 1.0 | max: 1.0

File Size vs. Commits (90 days): 97 points

csrc/api/api.cpp x: 2 commits (90d) y: 14 lines of code csrc/api/common.h x: 2 commits (90d) y: 223 lines of code csrc/api/fused_norm_rope_attn_rope_cast_fwd.cpp x: 2 commits (90d) y: 462 lines of code csrc/api/sparse_decode.cpp x: 2 commits (90d) y: 396 lines of code csrc/api/sparse_prefill.cpp x: 2 commits (90d) y: 195 lines of code csrc/ascend_kernels/prefill/sparse/config.h x: 1 commits (90d) y: 139 lines of code csrc/ascend_kernels/prefill/sparse/fast_div_mod.h x: 1 commits (90d) y: 28 lines of code csrc/ascend_kernels/prefill/sparse/kernel.h x: 1 commits (90d) y: 16 lines of code csrc/ascend_kernels/prefill/sparse/kernel_body.h x: 1 commits (90d) y: 1440 lines of code csrc/cuda_kernels/defines.h x: 1 commits (90d) y: 9 lines of code csrc/cuda_kernels/kv_cache_format.h x: 1 commits (90d) y: 45 lines of code csrc/cuda_kernels/sm100/common_subroutine.h x: 1 commits (90d) y: 168 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/config.h x: 1 commits (90d) y: 151 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/instantiations/v41_h64.cu x: 1 commits (90d) y: 5 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/kernel.cuh x: 1 commits (90d) y: 667 lines of code csrc/cuda_kernels/sm100/dequant_utils.cuh x: 1 commits (90d) y: 107 lines of code csrc/cuda_kernels/sm100/helpers.h x: 1 commits (90d) y: 196 lines of code csrc/cuda_kernels/sm100/kv_cache_utils.cuh x: 1 commits (90d) y: 57 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_common.hpp x: 1 commits (90d) y: 84 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_fusion.hpp x: 1 commits (90d) y: 277 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp x: 1 commits (90d) y: 148 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp x: 1 commits (90d) y: 805 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_load_tma_warpspecialized.hpp x: 1 commits (90d) y: 205 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp x: 1 commits (90d) y: 814 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_load_tma_warpspecialized.hpp x: 1 commits (90d) y: 220 lines of code csrc/cuda_kernels/sm100/prefill/dense/common/pipeline_mla.hpp x: 1 commits (90d) y: 174 lines of code csrc/cuda_kernels/sm100/prefill/dense/device/fmha_device_bwd.hpp x: 1 commits (90d) y: 237 lines of code csrc/cuda_kernels/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu x: 1 commits (90d) y: 76 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_causal_tile_scheduler.hpp x: 1 commits (90d) y: 117 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_kernel_bwd_convert.hpp x: 1 commits (90d) y: 100 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_options.hpp x: 1 commits (90d) y: 38 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_tile_scheduler.hpp x: 1 commits (90d) y: 96 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_bwd_kernel_tma_warpspecialized.hpp x: 1 commits (90d) y: 1416 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp x: 1 commits (90d) y: 491 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.cuh x: 1 commits (90d) y: 1578 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.h x: 1 commits (90d) y: 34 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_q_b_proj/kernel.h x: 1 commits (90d) y: 21 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_wv_proj/kernel.cu x: 1 commits (90d) y: 88 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128/phase1.cuh x: 1 commits (90d) y: 625 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64/phase1.cuh x: 1 commits (90d) y: 598 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/phase1.cuh x: 1 commits (90d) y: 974 lines of code csrc/cuda_kernels/smxx/decode/combine/combine.cu x: 1 commits (90d) y: 208 lines of code csrc/cuda_kernels/smxx/decode/get_decoding_sched_meta/get_decoding_sched_meta.cu x: 1 commits (90d) y: 132 lines of code csrc/params.h x: 2 commits (90d) y: 129 lines of code flash_mla/flash_mla_interface.py x: 2 commits (90d) y: 321 lines of code flash_mla/fused_norm_rope_attn_rope_cast.py x: 2 commits (90d) y: 77 lines of code
1578.0
lines of code
  min: 2.0
  average: 183.12
  25th percentile: 6.0
  median: 56.0
  75th percentile: 184.5
  max: 1578.0
0 2.0
commits (90d)
min: 1.0 | average: 1.1 | 25th percentile: 1.0 | median: 1.0 | 75th percentile: 1.0 | max: 2.0

File Size vs. Contributors (90 days): 97 points

csrc/api/api.cpp x: 1 contributors (90d) y: 14 lines of code csrc/api/common.h x: 1 contributors (90d) y: 223 lines of code csrc/api/fused_norm_rope_attn_rope_cast_fwd.cpp x: 1 contributors (90d) y: 462 lines of code csrc/api/sparse_decode.cpp x: 1 contributors (90d) y: 396 lines of code csrc/api/sparse_prefill.cpp x: 1 contributors (90d) y: 195 lines of code csrc/ascend_kernels/prefill/sparse/config.h x: 1 contributors (90d) y: 139 lines of code csrc/ascend_kernels/prefill/sparse/fast_div_mod.h x: 1 contributors (90d) y: 28 lines of code csrc/ascend_kernels/prefill/sparse/kernel_body.h x: 1 contributors (90d) y: 1440 lines of code csrc/cuda_kernels/defines.h x: 1 contributors (90d) y: 9 lines of code csrc/cuda_kernels/kv_cache_format.h x: 1 contributors (90d) y: 45 lines of code csrc/cuda_kernels/sm100/common_subroutine.h x: 1 contributors (90d) y: 168 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/config.h x: 1 contributors (90d) y: 151 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/instantiations/v41_h64.cu x: 1 contributors (90d) y: 5 lines of code csrc/cuda_kernels/sm100/decode/sparse/head64/kernel.cuh x: 1 contributors (90d) y: 667 lines of code csrc/cuda_kernels/sm100/dequant_utils.cuh x: 1 contributors (90d) y: 107 lines of code csrc/cuda_kernels/sm100/kv_cache_utils.cuh x: 1 contributors (90d) y: 57 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_common.hpp x: 1 contributors (90d) y: 84 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/fmha_fusion.hpp x: 1 contributors (90d) y: 277 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_epilogue_tma_warpspecialized.hpp x: 1 contributors (90d) y: 148 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp x: 1 contributors (90d) y: 805 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_load_tma_warpspecialized.hpp x: 1 contributors (90d) y: 205 lines of code csrc/cuda_kernels/sm100/prefill/dense/collective/sm100_fmha_mla_fwd_mainloop_tma_warpspecialized.hpp x: 1 contributors (90d) y: 814 lines of code csrc/cuda_kernels/sm100/prefill/dense/common/pipeline_mla.hpp x: 1 contributors (90d) y: 174 lines of code csrc/cuda_kernels/sm100/prefill/dense/device/fmha_device_bwd.hpp x: 1 contributors (90d) y: 237 lines of code csrc/cuda_kernels/sm100/prefill/dense/fmha_cutlass_bwd_sm100.cu x: 1 contributors (90d) y: 76 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_causal_tile_scheduler.hpp x: 1 contributors (90d) y: 117 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_kernel_bwd_convert.hpp x: 1 contributors (90d) y: 100 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_options.hpp x: 1 contributors (90d) y: 38 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/fmha_tile_scheduler.hpp x: 1 contributors (90d) y: 96 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_bwd_kernel_tma_warpspecialized.hpp x: 1 contributors (90d) y: 1416 lines of code csrc/cuda_kernels/sm100/prefill/dense/kernel/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp x: 1 contributors (90d) y: 491 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.cuh x: 1 contributors (90d) y: 1578 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/core_attn/kernel.h x: 1 contributors (90d) y: 34 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_q_b_proj/kernel.h x: 1 contributors (90d) y: 21 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fused_norm_rope_attn_rope_cast_fwd/permute_wv_proj/kernel.cu x: 1 contributors (90d) y: 88 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head128/phase1.cuh x: 1 contributors (90d) y: 625 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd/head64/phase1.cuh x: 1 contributors (90d) y: 598 lines of code csrc/cuda_kernels/sm100/prefill/sparse/fwd_for_small_topk/head128/phase1.cuh x: 1 contributors (90d) y: 974 lines of code csrc/cuda_kernels/smxx/decode/combine/combine.cu x: 1 contributors (90d) y: 208 lines of code csrc/cuda_kernels/smxx/decode/get_decoding_sched_meta/get_decoding_sched_meta.cu x: 1 contributors (90d) y: 132 lines of code csrc/params.h x: 1 contributors (90d) y: 129 lines of code flash_mla/flash_mla_interface.py x: 1 contributors (90d) y: 321 lines of code
1578.0
lines of code
  min: 2.0
  average: 183.12
  25th percentile: 6.0
  median: 56.0
  75th percentile: 184.5
  max: 1578.0
0 1.0
contributors (90d)
min: 1.0 | average: 1.0 | 25th percentile: 1.0 | median: 1.0 | 75th percentile: 1.0 | max: 1.0
About This Analysis
  • File size measurements show the distribution of size of files.
  • Files are classified in four categories based on their size (lines of code): 1-100 (very small files), 101-200 (small files), 201-500 (medium size files), 501-1000 (long files), 1001+(very long files).
  • It is a good practice to keep files small. Long files may become "bloaters", code that have increased to such gargantuan proportions that they are hard to work with.
Learn more...