Lfalive
|
7fd167fd25
|
tanh avx512 mask optimization (#6096)
|
1 year ago |
Yexuan Wu
|
73d8500988
|
sigmoid avx512 mask optimization (#6093)
|
1 year ago |
tpoisonooo
|
dddfc50282
|
raise max gpu count to 32 (#6084)
|
1 year ago |
nihui
|
a8e4db713b
|
initialize layer featmask zero (#6078)
|
1 year ago |
nihui
|
6510fe6125
|
discover VK_KHR_shader_float_controls2 (#6068)
|
1 year ago |
nihui
|
ca045ac579
|
gpu info query cooperative matrix properties (#6067)
|
1 year ago |
nihui
|
7f899f2e94
|
update glslang, discover VK_KHR_shader_integer_dot_product and VK_KHR_shader_bfloat16 (#6066)
|
1 year ago |
nihui
|
6fa649f103
|
clip x86 avx512 mask optimization (#6059)
|
1 year ago |
nihui
|
572b6198ad
|
fix subgroup min max size without subgroup size control extension (#6047)
|
1 year ago |
nihui
|
f371ecd16e
|
fix x86 conv int8 avx512vnni crash (#6038)
|
1 year ago |
nihui
|
90b84cd50e
|
simplevk support nvidia headless vulkan (#6036)
|
1 year ago |
nihui
|
305837fd4a
|
fix vulkan driver blacklist (#6034)
|
1 year ago |
nihui
|
205ca50e52
|
fix shader compile failure on nvidia (#6029)
|
1 year ago |
AtomAlpaca
|
cdaa54a752
|
risc-v bias optimization (#6025)
|
1 year ago |
nihui
|
243883a782
|
add apple a18 and m4 family types (#6026)
|
1 year ago |
nihui
|
c9caafdafa
|
fix build for mips loongarch risc-v with simplemath (#6024)
|
1 year ago |
nihui
|
409f9aef67
|
workaround vulkan padding stall with new nvidia driver (#5953)
|
1 year ago |
AtomAlpaca
|
71ed023c71
|
risc-v celu optimization (#6013)
|
1 year ago |
nihui
|
81c7df7056
|
set openmp env to disable openmp affinity capability and allow multiple openmp runtime (#6009)
|
1 year ago |
AtomAlpaca
|
a4ca4401dc
|
risc-v bnll optimization (#6005)
|
1 year ago |
nihui
|
80da741307
|
glsl define ncnn_glsl_version macro (#6003)
|
1 year ago |
nihui
|
1d9eddff8f
|
reset extractor local vulkan allocator (#6002)
resolve reclaim_blob_allocator get wild allocator warning when clear() called multple times
|
1 year ago |
nihui
|
8363040cb4
|
pnnx ncnn gelu fast mode, fix interp 2d resize (#5999)
|
1 year ago |
nihui
|
ef0b0e631c
|
interp output size expression (#5994)
|
1 year ago |
nihui
|
f38768a462
|
fix convolution int8 im2col gemm on tiny l2 cache size, update ci native linux aarch64 and windows arm64 (#5995)
|
1 year ago |
xfan1024
|
b1870434f3
|
eltwise riscv rvv optimization (#5992)
|
1 year ago |
nihui
|
39c055d7f2
|
crop axes starts ends expression (#5976)
* skip dynamic tensor index
* handle clone oom
|
1 year ago |
nihui
|
02fcca2a1e
|
guard VK_USE_PLATFORM_ANDROID_KHR with ifdef (#5986)
|
1 year ago |
nihui
|
33d03e625c
|
drop ncnn glsl string literal macros (#5975)
glsl has no string type anyway
|
1 year ago |
nihui
|
8dbcfee5ec
|
option owns vulkan device index (#5973)
|
1 year ago |
nihui
|
84970eed4d
|
vulkan validation layer enables NCNN_LOGE in shader source (#5963)
* NCNN_LOGE in glsl
* Update glsl-extension.md
|
1 year ago |
nihui
|
b284dbd0f4
|
discover VK_KHR_shader_non_semantic_info, checked convolution imagestore (#5955)
|
1 year ago |
nihui
|
eed257df1f
|
ci update llvmpipe (#5954)
* check image fp16
|
1 year ago |
nihui
|
ad30c7f6fb
|
clean vulkan shader common extension (#5952)
* clean vulkan shader common extension
* macro suffix makes glslang unhappy
|
1 year ago |
nihui
|
bf13c30210
|
define device feature macros for glslang, discover VK_EXT_shader_atomic_float and VK_EXT_shader_atomic_float2 (#5949)
|
1 year ago |
peerless2012
|
1d1ad06459
|
Fix get_elf_hwcap always return 0 on HarmonyOS NEXT. (#5951)
|
1 year ago |
nihui
|
8211930a6f
|
discover VK_KHR_shader_subgroup_rotate (#5948)
|
1 year ago |
nihui
|
1b6485fa17
|
discover VK_KHR_zero_initialize_workgroup_memory (#5947)
|
1 year ago |
nihui
|
40f7b4e527
|
discover all subgroup features and VK_KHR_shader_subgroup_extended_types (#5946)
|
1 year ago |
nihui
|
0b9925cfef
|
intergrate VK_EXT_subgroup_size_control features and properties (#5940)
|
1 year ago |
nihui
|
07267f2618
|
softmax 4d test and vulkan, softmax unified elempack optimization for x86 arm riscv (#5931)
|
1 year ago |
nihui
|
6396a732ef
|
reshape shape expression, drop reshape permute, test reshape oom (#5918)
|
1 year ago |
Yexuan Wu
|
3571d7e8ec
|
Support better API to detect big little core in windows after win7 (#5927)
|
1 year ago |
nihui
|
1e3fcb9dda
|
paramdict value string type, natural array representation (#5915)
|
1 year ago |
nihui
|
23890900c2
|
x86 optimization for convolution int8 gemm (#5874)
* cmake check compiler test cannot be optimized out
* drop requant pack4
|
1 year ago |
nihui
|
4a70be45ed
|
fix requantize pack4to8 (#5893)
|
1 year ago |
nihui
|
ff5b554003
|
restrict one dim quantize scale size, test quantize oom (#5892)
* restrict one dim quantize scale size
* sse2 requantize pack8
|
1 year ago |
nihui
|
956bccd295
|
restrict one dim requantize scale bias size (#5888)
|
1 year ago |
nihui
|
48e1260a6f
|
restrict one dim dequantize scale bias size (#5886)
|
1 year ago |
nihui
|
21a71d3673
|
slim x86 dequantize (#5879)
* remove dequantize pack8 test, seems to be useless
|
1 year ago |