Natsu
6d1944f2c3
CMake improvement ( #1115 )
* CMake improvement
* Fix bugs
* Fix typo
* Propagate vulkan dependency
* import vulkan
* add config files, now exported target cmake should be able to find packages
* Propagate no-rtti and no-exception
* Provide a option to control rtti and exception in mobile platform
* Make cmake clean
* Resolve conflicts
* Update CMake
PIE is propagated by INTERFACE_POSITION_INDEPENDENT_CODE
* Remove bad things
7 years ago
nihuini
eced9c81c6
fix crash on x86 conv7x7s1 dilation > 1, fix #1110 , fix #1117
7 years ago
tpoisonooo
1ca4387c9c
Auto choose conv implementation ( #1085 )
* add relative README_CN.md;
* obtain time cost with op->forward().
7 years ago
nihuini
e9c890ad8a
respect winograd top blob allocator when out size is exactly the one before cut border, fix #1118
7 years ago
wincle
fb9309e460
implement reflect padding.
7 years ago
wincle
bb6b93568d
Tanh Support
7 years ago
nihui
a7490fd71a
fix #1107
7 years ago
nihui
8c53706987
net vkdev getter api
7 years ago
nihui
5ce6ae90be
a portable thread class ( #1081 )
* a portable thread class
* Update platform.h.in
* Update platform.h.in
7 years ago
nihuini
c769437533
fix fp16p deconvolution and convolution-typed innerproduct
7 years ago
xue
cb14d1bbf3
x86模式下不使用sse时卷积越界的bug
7 years ago
hanson.young
698acd28f6
修复在window x64下AVX2编译的问题 ( #1076 )
* fix windows avx2 error in convolution_3x3.h and convolution_sgemm.h
* fix compatibility of avx2 under Linux and Windows
* Update CMakeLists.txt
Space instead of tab
* Update convolution_3x3.h
Space instead of tab
* Update convolution_sgemm.h
Space instead of tab
7 years ago
nihuini
6b933c6dc4
a portable condition variable class
7 years ago
BUG1989
bcfe9f453f
initial the ncnn post training quantization tools ( #1067 )
* initial the ncnn post training quantization tools
* clear some comments of tools
* fix the Travis ci compiler error
7 years ago
nihuini
b25f76833a
restore per extractor allocator setters, patially revert e09607bc22
7 years ago
nihuini
21b5508c96
shared locked vkallocator cannot prevent concurrent accessing during actual gpu inference, use seperated vkallocator for each queue
7 years ago
BUG1989
ae1b69ec53
update some comments of int8 contributors ( #1054 )
7 years ago
Jingdong Li
f7e4155ee3
implemented selu activation ( #1051 )
7 years ago
nihuini
082aa2489d
packing padding support packing
7 years ago
nihuini
4d5c6a00fb
absval batchnorm clip relu sigmoid support packing
7 years ago
nihuini
6935b78926
new layer attribute support_packing
7 years ago
nihuini
51ef60e71b
fix hardsigmoid
7 years ago
Howave
fed22ca4ed
fix neon intrinsic compile warnings ( #1045 )
7 years ago
PENGUINLIONG
084053fed8
Implemented hard sigmoid ( #1046 )
* Implementation for hard sigmoid
Not yet tested.
* Resolve requests
7 years ago
Howave
123ca35e00
fix compile warnings ( #1042 )
7 years ago
nihuini
bade132589
comment++
7 years ago
nihuini
81be8c86ae
fix bus error in resize_bilinear_c2 on armv7
7 years ago
nihuini
17d63a1491
fix bus error in resize_bilinear_c3 on armv7
7 years ago
nihuini
e9ffdb5bdd
16bit storage on arm mali is buggy
7 years ago
nihuini
73911492d7
fix validation warning on querypool destruction, enable fp16p by default
7 years ago
nihuini
040a8d2427
set vulkan device by gpu index
7 years ago
nihui
21f79b8546
prefer cpu fp16 casting to reduce upload/download overhead on discrete gpu
7 years ago
nihui
721abe91a8
packed mat is handy
7 years ago
nihui
afcfe0936f
fix false warnings
7 years ago
nihuini
e56f0d47cc
fix out of range load and store in bilinear resize c2/c3 neon block
7 years ago
BUG1989
c2022f4501
optimize conv sgemm with sse on intel platform ( #1035 )
* optimize conv sgemm with sse
* Update convolution_x86.cpp
7 years ago
nihuini
e09607bc22
add option to upload model function, pipeline creation honors option use flags, setting allocator per extractor do not make much sense
7 years ago
nihuini
e09d11f936
rough fix build without arm neon
7 years ago
nihuini
5fdffbcaac
destroy_gpu_instance is not threadsafe anyway, fix deadlock on exit
7 years ago
BUG1989
d9f269fa3d
use sgemm fp32 on arm platform,optimize conv1x1s2 ( #1031 )
7 years ago
nihuini
838c5df839
option api changes
7 years ago
nihuini
7f7bbf12e5
new api for getting the default gpu device
7 years ago
nihuini
4de4078779
move platform includes out of namespace
7 years ago
BUG1989
b53541e8f9
fix arm winograd int8,optimize winograd x86 ( #1025 )
7 years ago
BUG1989
01b3804828
optimization the x86 convolution layer with avx2 ( #1019 )
* add the "Tu Fa" conv sgemm fp32 with avx2 for x86
* add avx2 cmake option
* fix some bugs of avx2 pull request
7 years ago
nihui
fe4b00f7a2
unroll outh 4 for winograd gemm
7 years ago
nihuini
74276314bb
unroll size 4 for conv1x1s1 pack4
7 years ago
nihuini
cd7559c639
more fix for fp16p, still disabled by default
7 years ago
nihuini
4b6bffa560
Mat row should be elemsize-aware
7 years ago
harhar539
5e317b98c5
fix illegal memory access at conv layer of vulkan ( #1011 )
* 1.fix pad tail bug in commit d1ea2a3 at pooling layer
* fix illegal memory access at conv layer of vulkan
fix illegal memory access at conv layer of vulkan when bias term is 0
7 years ago