nihui
54e79a62d7
fix crash on non-arm82 build
6 years ago
nihui
c173d51c9b
mish sigmoid swish tanh arm fp16s
6 years ago
nihui
72a27d4776
utility wrapper for neon float32 bfloat16 conversion, deconvolution deconvolutiondepthwise arm fp16s fp16sa bf16s
6 years ago
nihui
e644164873
reshape arm bf16s fp16s, flatten api
6 years ago
nihui
aa68246dc7
more test coverage
6 years ago
nihui
e7abc5fbd7
concat slice arm fp16sa pack8
6 years ago
nihui
aa1a9e90c5
interp shufflechannel arm fp16sa pack8
6 years ago
nihuini
c6d7525367
convolutiondepthwise arm fp16sa pack8
6 years ago
nihuini
bc3822acc3
convolution flatten arm fp16sa pack8
6 years ago
nihuini
dbb761b9a4
binaryop eltwise padding pooling arm fp16sa pack8
6 years ago
nihuini
91d91ba556
hardsigmoid hardswish arm fp16s fp16sa
6 years ago
nihuini
f23122bb3f
since fp16 storage option is on by default, upper-level function may pass fp32 storage with default option, guard with element bits checking
6 years ago
nihuini
a18a9fd8c5
eltwise arm fp16s fp16sa
6 years ago
nihuini
8385d81afa
interp arm fp16s fp16sa
6 years ago
nihuini
5a4243e44e
binaryop arm fp16s
6 years ago
nihuini
301abe657c
relu arm fp16s
6 years ago
nihui
03c9ed11d2
pooling arm fp16s fp16sa
6 years ago
nihui
71f86af8a6
fix non-arm82 ci
6 years ago
nihui
9a2e2a6937
convert fp32 blobs for layers with fp16 storage support
6 years ago
nihui
d4d501a7fe
fix innerproduct fp16sa
6 years ago
nihui
e9c71a1ead
innerproduct arm fp16s fp16sa
6 years ago
nihui
1a57600bd7
fix ci crash
6 years ago
nihuini
11f5033249
convolutiondepthwise arm fp16s fp16sa
6 years ago
nihuini
6ab284bc3a
convolution arm fp16s fp16sa
6 years ago
nihuini
6d2c0e5683
flatten fp16s
6 years ago
nihuini
47ae0c151a
some shared arm bf16s fp16s implementation
6 years ago
zchrissirhcz
b80b84fda5
fix #1542 ; fix avx2 uint16_t including ( #1968 )
* fix #1542 ; fix avx2 uint16_t including
for #1542 , it is for compatibility for opencv 2.x, such as on ubuntu 16.04 apt installed opencv
6 years ago
nihui
1322ae40cb
update engine version
6 years ago
nihui
308145254e
mask bf16 option in layer forward, disable gpu when bf16 enabled, fix #1962
6 years ago
nihui
71dc13625f
disable bf16 storage for int8 inference
6 years ago
nihuini
8700985540
yet another workaround for nexus6p gpu
6 years ago
nihuini
bf279dcf17
workaround corrupted pipeline cache on old qcom adreno
6 years ago
nihuini
4e4f0baa73
set openmp blocktime 20 for reducing power consumption, blocktime option
6 years ago
nihui
21762e09e5
fix dilated convolution ( #1956 )
6 years ago
nihuini
4d2d625432
fix avx2 build, second try, fix #1953
6 years ago
nihuini
8b0890999a
fix avx2 build, fix #1953
6 years ago
nihui
88367f4164
Ci enable mips msa ( #1949 )
6 years ago
nihui
bb5bfe3841
avx2 infrastructure ( #1943 )
6 years ago
nihui
11cffce114
armv8.2 infrastructure ( #1856 )
* runtime cpu dispatch
* force thread one
* disable openmp for coverage
* simplify test layer
* print NCNN_TARGET_ARCH
* less ci build variants
* weight fp16 storage option
* test convdw int8
* apple a12 a13
* ncnn_add_layer ncnn_add_shader cmake macro
6 years ago
nihui
3ff40b0679
Ci rv32imc ( #1940 )
6 years ago
nihui
fe6bc1ed4d
Ci rv64gcv and rv64gc ( #1936 )
6 years ago
nihui
b8f3e1455e
code clean
6 years ago
nihui
6ebb774e37
fix interp arm
6 years ago
nihui
6538b95102
fix interp and lrn test
6 years ago
nihuini
68750a5151
fix data race in innerproduct avx
6 years ago
nihuini
0d6cc01d55
innerproduct handle mish activation, fix naive C testing, fix #1930
6 years ago
nihui
be330e0fc4
test mat pixel
6 years ago
nihui
109e079c51
test deconvolution with output shape and padding
6 years ago
nihuini
00ef566609
implement full permute tag support for reshape
6 years ago
nihui
193e08e834
lazy initialize utility operator, fix #1923
6 years ago