alibaba / MNN - File Size

Intro

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...

File Size Overall

There are 2,794 files with 531,272 lines of code.

75 very long files (214,322 lines of code)
106 long files (71,521 lines of code)
374 medium size files (116,492 lines of codeclsfd_ftr_w_mp_ins)
419 small files (60,420 lines of code)
1,820 very small files (68,517 lines of code)

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

File Size per Logical Decomposition

primary

1001+

501-1000

201-500

101-200

1-100

Longest Files (Top 50)

File	# lines	# units
fmha_v2_flash_attention_fp16_128_16_S_32_sm80.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	9916	-
fmha_v2_flash_attention_fp16_128_16_S_32_sm89.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	9916	-
fmha_v2_flash_attention_fp16_128_16_S_32_sm86.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	9916	-
fmha_v2_flash_attention_fp16_64_32_S_32_sm89.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	8593	-
fmha_v2_flash_attention_fp16_64_32_S_32_sm86.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	8593	-
fmha_v2_flash_attention_fp16_64_32_S_32_sm80.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	8583	-
fmha_v2_flash_attention_fp16_128_16_S_16_sm80.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	8337	-
fmha_v2_flash_attention_fp16_128_16_S_16_sm89.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	8327	-
fmha_v2_flash_attention_fp16_128_16_S_16_sm86.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	8327	-
fmha_v2_flash_attention_fp16_64_32_S_16_sm86.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	7676	-
fmha_v2_flash_attention_fp16_64_32_S_16_sm89.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	7676	-
fmha_v2_flash_attention_fp16_64_32_S_16_sm80.cubin.cpp in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/src	7676	-
httplib.h in transformers/llm/engine/src	7138	370
AllShaderRender.cpp in source/backend/vulkan/buffer/render/compiler	6529	-
proto NeuralNetwork.proto in source/backend/coreml/mlmodel/proto	5706	-
CommonOptFunction.cpp in source/backend/cpu/compute	3509	110
__init__.py in pymnn/pip_package/MNN/expr	2951	139
GemmInt8_VNNI.cpp in source/backend/cpu/x86_x64/avx512	2817	3
GemmFunction.hpp in source/backend/cpu/x86_x64/avx	2745	30
protobuf-c.c in source/backend/coreml/mlmodel/src	2655	3
inl Matmul_4_4_64.inl in source/backend/cpu/x86_x64/avx512	2639	2
MNN.cc in pymnn/src	2474	21
AllShader.cpp in source/backend/metal	2305	-
Int8FunctionsOpt.cpp in source/backend/cpu/compute	2151	30
structural.cpp in tools/cv/source/imgproc	2006	23
expr.h in pymnn/src	1892	12
proto caffe.proto in tools/converter/source/caffe	1887	-
constants.py in transformers/llm/export/gguf	1741	1
GemmInt8.cpp in source/backend/cpu/x86_x64/avx	1566	12
NeuralNetWorkOp.cpp in express	1536	103
conv_2d_buf_mnn_cl.cpp in source/backend/opencl/execution/cl	1499	-
GemmBatchedTensorCoreFloat16Tune.cu in source/backend/cuda/execution/cutlass_common/tune	1487	-
WinogradOptFunction.cpp in source/backend/cpu/compute	1469	39
CutlassGemmParamTune.hpp in source/backend/cuda/execution/cutlass_common/tune	1435	-
OpenCLBackend.cpp in source/backend/opencl/core	1417	50
doxygen.css in doc/API/html	1400	-
cl conv_2d_buf.cl in source/backend/opencl/execution/cl	1398	-
matmul_params_buf_mnn_cl.cpp in source/backend/opencl/execution/cl	1393	-
AttentionBufExecution.cpp in source/backend/opencl/execution/buffer	1392	12
WinogradFunctions.cpp in source/backend/cpu/x86_x64/avx512	1388	23
cl matmul_params_buf.cl in source/backend/opencl/execution/cl	1364	-
conv_2d_mnn_cl.cpp in source/backend/opencl/execution/cl	1357	-
AllRenderShader.cpp in source/backend/metal/render	1356	-
cl conv_2d.cl in source/backend/opencl/execution/cl	1355	-
FeatureTypes.pb-c.c in source/backend/coreml/mlmodel/src	1328	-
predicated_tile_access_iterator_residual_last.h in source/backend/cuda/execution/plugin/FmhaCommon/fused_multi_head_attention/iterators	1327	74
ConvInt8TiledExecutor.cpp in source/backend/cpu/compute	1320	10
predicated_tile_iterator_residual_last.h in source/backend/cuda/execution/plugin/FmhaCommon/fused_multi_head_attention/iterators	1315	92
calibration.cpp in tools/quantization	1302	18
Expr.cpp in express	1296	39

Files With Most Units (Top 50)

File	# lines	# units
httplib.h in transformers/llm/engine/src	7138	370
__init__.py in pymnn/pip_package/MNN/numpy	1251	184
__init__.py in pymnn/pip_package/MNN/expr	2951	139
CPUFixedPoint.hpp in source/backend/cpu	794	128
SkNx_neon.h in source/cv	588	124
CommonOptFunction.cpp in source/backend/cpu/compute	3509	110
MathOp.cpp in express	628	108
NeuralNetWorkOp.cpp in express	1536	103
predicated_tile_iterator_residual_last.h in source/backend/cuda/execution/plugin/FmhaCommon/fused_multi_head_attention/iterators	1315	92
llmconfig.hpp in transformers/llm/engine/src	451	85
cxxopts.hpp in tools/converter/include	1011	75
predicated_tile_access_iterator_residual_last.h in source/backend/cuda/execution/plugin/FmhaCommon/fused_multi_head_attention/iterators	1327	74
SkNx.h in source/cv	343	72
Matrix_CV.cpp in source/cv	860	68
MetalBackend.mm in source/backend/metal	1195	64
OpenCLWrapper.cpp in source/backend/opencl/core/runtime	659	57
Matrix.h in include/MNN	368	53
VulkanDevice.cpp in source/backend/vulkan/component	489	53
OpenCLBackend.cpp in source/backend/opencl/core	1417	50
Vec.hpp in source/math	782	48
BufferAllocator.cpp in source/core	608	47
vision.py in transformers/llm/export/utils	633	46
llmexport.py in transformers/llm/export	963	46
ImageProcessFunction.cpp in source/backend/cpu/compute	897	46
Rect.h in include/MNN	199	42
BinaryUtils.hpp in source/backend/cpu	448	41
util.h in pymnn/src	1292	40
Expr.cpp in express	1296	39
CPURuntime.cpp in source/backend/cpu	1152	39
WinogradOptFunction.cpp in source/backend/cpu/compute	1469	39
TensorUtils.cpp in source/core	899	38
token2wav.py in transformers/llm/export/utils	412	37
llm.cpp in transformers/llm/engine/src	599	37
makeshader.py in source/backend/vulkan/image/compiler	387	35
makeshader.py in source/backend/vulkan/buffer/render/compiler	392	35
makeshader.py in source/backend/vulkan/buffer/compiler	390	35
MergeHelpers.cpp in tools/converter/source/optimizer/merge	197	34
sampler.cpp in transformers/llm/engine/src	409	34
ModelConfigManager.swift in apps/iOS/MNNLLMChat/MNNLLMiOS/Chat/Models	172	32
OpenCLRuntime.cpp in source/backend/opencl/core/runtime	900	32
Executor.cpp in express	593	31
VecHalf.hpp in backupcode/cpubackend/bf16	469	31
PackedFunction.cpp in source/backend/cpu/x86_x64/avx512	1010	31
VecHalf.hpp in source/backend/cpu/bf16	469	31
VulkanBackend.cpp in source/backend/vulkan/buffer/backend	448	31
NN.cpp in tools/train/source/nn	1135	30
tokenizer.cpp in transformers/llm/engine/src	633	30
UnaryUtils.hpp in source/backend/cpu	228	30
CPUBackend.cpp in source/backend/cpu	726	30
GemmFunction.hpp in source/backend/cpu/x86_x64/avx	2745	30

Files With Long Lines (Top 50)

There are 1189 files with lines longer than 120 characters. In total, there are 8126 long lines.

File	# lines	# units	# long lines
GemmTensorCoreFloat16TuneInfer.cu in source/backend/cuda/execution/cutlass_common/tune	1230	-	174
Int8FunctionsOpt.cpp in source/backend/cpu/compute	2151	30	170
GemmBatchedTensorCoreFloat16Tune.cu in source/backend/cuda/execution/cutlass_common/tune	1487	-	148
AttentionBufExecution.cpp in source/backend/opencl/execution/buffer	1392	12	144
cl matmul_params_buf.cl in source/backend/opencl/execution/cl	1364	-	132
VulkanShaderMap.cpp in source/backend/vulkan/image/compiler	158	1	126
CutlassGemmBatchedTensorFloat16TuneInfer.cu in source/backend/cuda/execution/cutlass_common/tune	982	-	120
VulkanShaderMap.cpp in source/backend/vulkan/buffer/compiler	174	1	105
fmha_flash_attention.h in source/backend/cuda/execution/plugin/FmhaV2/fmha_flash_attention/include	447	7	101
MatMulExecution.cu in source/backend/cuda/execution	1160	-	85
ConvInt8TiledExecutor.cpp in source/backend/cpu/compute	1320	10	70
structural.cpp in tools/cv/source/imgproc	2006	23	69
CommonOptFunction.h in source/backend/cpu/compute	277	-	57
ConvBufLowMemoryExecution.cpp in source/backend/opencl/execution/buffer	892	11	55
OpenCLBackend.cpp in source/backend/opencl/core	1417	50	53
cl conv_2d_buf.cl in source/backend/opencl/execution/cl	1398	-	53
onnx_model_graph_opt.py in tools/converter/source/onnx	792	26	52
GemmTensorCoreFloat16Tune.cu in source/backend/cuda/execution/cutlass_common/tune	732	-	52
depthwise_conv2d_buf_mnn_cl.cpp in source/backend/opencl/execution/cl	750	-	52
cl depthwise_conv2d_buf.cl in source/backend/opencl/execution/cl	708	-	51
SelfAttentionBufExecution.cpp in source/backend/opencl/execution/buffer	456	6	49
OAQ_quantizer.py in tools/mnncompress/mnncompress/tensorflow	662	25	48
LoopExecution.cpp in source/backend/opencl/execution/image	725	8	44
ConvBufExecution.cpp in source/backend/opencl/execution/buffer	766	5	44
PackedFunction.cpp in source/backend/cpu/x86_x64/avx	658	26	43
ConvolutionPackFreeWinograd.cpp in source/backend/cpu/compute	704	6	43
Raster.cu in source/backend/cuda/execution	1117	-	42
proto NeuralNetwork.proto in source/backend/coreml/mlmodel/proto	5706	-	42
ConvExecution.cpp in source/backend/opencl/execution/image	481	5	41
ConvLowMemoryExecution.cpp in source/backend/opencl/execution/image	509	9	41
OpenCLWrapper.cpp in source/backend/opencl/core/runtime	659	57	40
Arm82Functions.cpp in source/backend/arm82	1132	28	40
conv_2d_buf_mnn_cl.cpp in source/backend/opencl/execution/cl	1499	-	38
GridSampleExecution.cu in source/backend/cuda/execution	319	-	37
cl conv_2d.cl in source/backend/opencl/execution/cl	1355	-	37
MetalConvolution1x1.mm in source/backend/metal	375	4	37
EMA_quantizer.py in tools/mnncompress/mnncompress/tensorflow	583	24	36
Pipeline.cpp in source/core	1204	23	34
PackedFunction.cpp in source/backend/cpu/x86_x64/avx512	1010	31	34
MetalRasterAndInterpolate.mm in source/backend/metal/render	447	10	34
GeometryLSTM.cpp in source/geometry	731	6	33
CommonOptFunction.cpp in source/backend/cpu/compute	3509	110	33
LSQ_quantizer.py in tools/mnncompress/mnncompress/tensorflow	563	24	32
mnn_kleidiai.cpp in source/backend/cpu/arm	380	12	32
Int8FunctionsOpt.h in source/backend/cpu/compute	100	-	32
ConvBufWinograd.cpp in source/backend/opencl/execution/buffer	647	7	32
DenseConvolutionTiledExecutor.cpp in source/backend/cpu/compute	661	13	31
StrassenMatmulComputor.cpp in source/backend/cpu/compute	508	8	31
NN.cpp in tools/train/source/nn	1135	30	30
Transpose.cu in source/backend/cuda/execution	857	-	30