You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
nihuini 838c5df839 option api changes 7 years ago
..
avx_mathfun.h add ncnn source qwq 9 years ago
convolution_1x1.h innerproduce layer with int8 impl,the type of top_blob shoud be integer. (#578) 7 years ago
convolution_1x1_int8.h merge de/requantize op, optimize some int8 conv layer on arm64-v8a (#867) 7 years ago
convolution_3x3.h fix arm winograd int8,optimize winograd x86 (#1025) 7 years ago
convolution_3x3_int8.h add int8 winograd F(4,3) with neon assembly optimization (#891) 7 years ago
convolution_5x5.h optimization the x86 convolution layer with avx2 (#1019) 7 years ago
convolution_5x5_int8.h merge de/requantize op, optimize some int8 conv layer on arm64-v8a (#867) 7 years ago
convolution_7x7.h optimization the x86 convolution layer with avx2 (#1019) 7 years ago
convolution_7x7_int8.h merge de/requantize op, optimize some int8 conv layer on arm64-v8a (#867) 7 years ago
convolution_sgemm.h move platform includes out of namespace 7 years ago
convolution_sgemm_int8.h merge de/requantize op, optimize some int8 conv layer on arm64-v8a (#867) 7 years ago
convolution_x86.cpp option api changes 7 years ago
convolution_x86.h fix arm winograd int8,optimize winograd x86 (#1025) 7 years ago
convolutiondepthwise_3x3.h implement ncnn blob/workspace allocator, fine-grained per-layer openmp threads control, fix #469 7 years ago
convolutiondepthwise_3x3_int8.h merge de/requantize op, optimize some int8 conv layer on arm64-v8a (#867) 7 years ago
convolutiondepthwise_x86.cpp option api changes 7 years ago
convolutiondepthwise_x86.h move vulkan code to subdir, new layer interface create_pipeline and destroy_pipeline for post-loading works 7 years ago
sse_mathfun.h add ncnn source qwq 9 years ago