nihui
|
9e529354fb
|
arm neon optimization for conv1x1s2 pack4
|
6 years ago |
nihuini
|
f8f3b0b5aa
|
shufflechannel pack4
|
6 years ago |
nihuini
|
50d5896ce7
|
reshape pack4
|
6 years ago |
nihuini
|
624291e2b2
|
use subop optimization for group convolution deconvolution pack4 family
|
6 years ago |
nihui
|
48e3e7d49c
|
move neon activation into a wrapper function
|
6 years ago |
nihui
|
8c1b87b1a2
|
fallback to cpu if no vulkan device found
|
6 years ago |
nihui
|
b37ecab630
|
auto flatten before innerproduct pack4
|
6 years ago |
nihui
|
afd1f08194
|
arm neon assembly optimization for pooling2x2s2 max pack4
|
6 years ago |
nihui
|
e19b7097df
|
arm neon assembly optimization for conv3x3s1 pack1to4
|
6 years ago |
nihui
|
3ac6335ba3
|
hardsigmoid and hardswish pack4
|
6 years ago |
nihui
|
21e74487b4
|
arm neon optimization for convdw5x5 pack4
|
6 years ago |
volvet
|
ecd64fb36b
|
Fixed lots of compile warnings (#1286)
* Fixed lots of compile warnings
* refine the unused warning change
|
6 years ago |
nihui
|
3e1bad4880
|
arm neon assembly optimization for pooling3x3s2 max pack4
|
6 years ago |
nihui
|
08a97c169f
|
arm neon assembly optimization for relu pack4
|
6 years ago |
nihui
|
a1bd88fb4a
|
arm neon assembly optimization for padding constant pack4
|
6 years ago |
nihui
|
17f343e7e4
|
convdw3x3 pack4 arm neon assembly optimization
|
6 years ago |
nihui
|
6703286831
|
the very long ld1, one less load
|
6 years ago |
nihui
|
22a3ade6ce
|
unroll size 12 for conv1x1s1 and conv3x3s1 winograd pack4 on aarch64
|
6 years ago |
nihui
|
3a452f734a
|
arm neon assembly optimization for conv3x3s2 pack1to4
|
6 years ago |
nihui
|
6edd42f566
|
arm neon assembly for conv1x1s1 and conv3x3s1 winograd pack4
|
6 years ago |
nihuini
|
c0a4ffcf66
|
convolution pad_value param
|
6 years ago |
nihuini
|
587a67eb51
|
the noop layer
|
6 years ago |
nihuini
|
b7085ceec0
|
deconvolution apply output adj first, then crop the padding
|
6 years ago |
tpoisonooo
|
8dbafe7764
|
constraint input value to [-127, +127] (#1258)
* constraint input value to [-127, +127]
* keep new line at the end
|
6 years ago |
nihui
|
e56fcc77c5
|
optimize dot memory layout
|
6 years ago |
nihuini
|
8a7b4b035e
|
radv crash with large local group size, workaround
|
6 years ago |
nihuini
|
80f898b079
|
unaryop tanh vulkan
|
6 years ago |
nihuini
|
70f3198703
|
conversion between ncnn Mat and android Bitmap makes life easier :D
|
6 years ago |
nihuini
|
8f72256d1f
|
mat pixel rgb2rgba bgr2rgba gray2rgba
|
6 years ago |
nihuini
|
91ef4eea4f
|
fix unaryop arm, fix #1241
|
6 years ago |
nihuini
|
3e3189736c
|
fix msvc build, fix #1237
|
6 years ago |
Xu Yang
|
31cf7f3c5b
|
fix ConvolutionDepthWise int8_requantize (#1233)
|
6 years ago |
nihuini
|
c4bebc6371
|
x86 conv3x3s1 winograd43 produce wrong result, revert to the good-old winograd23 version
|
6 years ago |
CnybTseng
|
d11c4c1d42
|
修改最新版ncnn/src/layer/vulkan/shader目录下的几个文件,以适配最新版的glslang,本次修改已在大疆Manifold2-G平台上验证通过 (#1231)
|
6 years ago |
nihui
|
46e7ac76ab
|
apply sgemm-like dot in winograd pack4 neon
|
6 years ago |
nihui
|
d6860d93f2
|
fix batchnorm pack4 neon multithreaded
|
6 years ago |
nihui
|
b90a6cf69c
|
split always support packing :)
|
6 years ago |
nihui
|
6b5588ebbe
|
fix batchnorm pack4 out of range write
|
6 years ago |
nihui
|
4dc98ffaab
|
conv1x1s1 and conv3x3s1 winograd pack4 neon optimization, first try
|
6 years ago |
nihui
|
02c811e829
|
fix padding pack4 elempack
|
6 years ago |
nihuini
|
5d1a94826d
|
crop border param, try to make life eaiser
|
6 years ago |
nihui
|
af646b6577
|
workaround for the adjacency same pipeline issue
|
6 years ago |
nihui
|
cb82ccbe20
|
convolution vulkan winograd63 shader compilation is too slow and runs slower in many cases, ok let's drop it
|
6 years ago |
nihuini
|
6db731408c
|
crop vulkan pack1to4 and pack4to1
|
6 years ago |
nihui
|
35990bb5d5
|
instancenorm vulkan
|
6 years ago |
nihuini
|
4b8fb744bc
|
fix potential to_pixels_resize out of range write, fix #1201
|
6 years ago |
nihui
|
c2bc0d1b88
|
padding vulkan reflect mode
|
6 years ago |
nihuini
|
296e0022df
|
deconvolution output adj and output shape
|
6 years ago |
nihuini
|
e4b44d293e
|
more autopad SAME_LOWER
|
6 years ago |
nihuini
|
0e26e3094e
|
autopad SAME_LOWER
|
6 years ago |