nihui
241524ffce
discard weight memory for x86 arm vulkan ( #3865 )
...
* discard weight memory for x86 and vulkan
* drop arm innerproduct weight
* drop arm convolution weight
* drop arm convolutiondepthwise weight
* drop x86 vulkan deconvolution deconvolutiondepthwise weight
* drop arm deconvolution deconvolutiondepthwise weight
* arm neon assembly optimization for innerproduct pack4
2022-06-01 10:16:35 +08:00
f10cc6dd93
initial data structure changes for 3dcnn, conv3d, pooling3d ( #3378 )
...
Co-authored-by: ElvisYu <elvisyuovo@gmail.com >
Co-authored-by: 余浩文 <m18107220188@163.com >
Co-authored-by: Zr2223 <67497651+Zr2223@users.noreply.github.com >
2021-12-02 16:50:55 +08:00
nihui
24fbb6e8cb
honor thread setting on load and vulkan command, ci avx512 t4 ( #3391 )
2021-12-01 16:36:37 +08:00
nihui
052b2a1653
apply code-format changes
2021-07-17 23:37:32 +00:00
chenxiemin
9bd8a50b96
fix vulkan memory leak issue ( #3088 )
2021-07-18 07:36:08 +08:00
nihui
9fd4d371ae
bridge image for adreno image upload and download ( #2658 )
...
* add bridge image for adreno image storage upload and download
* enable sbn1, print bugbilz flag
* blacklist old adreno
* let user choose use_image_storage option even when bug_storage_buffer_no_l1
2021-02-10 11:05:45 +08:00
nihuini
77db369cae
fix upload data may be invalidated by previous pipelines with the same memory on integrated device, fix #2640
2021-02-02 16:47:08 +08:00
nihui
54c0a13b9f
build shared library ( #2525 )
...
* build shared lib and enable lto
* reserved for layer and option
* allocator pimpl
* datareader pimpl
* paramdict pimpl, disable copy assign for allocator and datareader
* modelbin pimpl
* net extractor pimpl
* gpu pimple
* disable copy assign vulkandevice, code format
* command pimpl, dummy image readonly
* pipeline pipelinecache pimpl, export platform class
* code format, export simple family
* update ci
* disable lto on android armv7, merge webassembly ci
* link libgcc, fix macos dylib version
* pipeline pimpl, gpu info pimpl
* destroy gpu info after vulkan device
* ignore msvc stl class warning
* fix ncnn_paramdict_get_float return type
* fix vktransfer upload fp16 without flatten, add command test
2021-01-01 22:55:10 +08:00
nihui
79efe33fdc
cmake option for platform api uses ( #2502 )
...
* cmake option for platform api uses
* adroid gpu ci does not rely on glslangvalidator, add android termux ci
2020-12-25 11:24:56 +08:00
nihui
4463c3b455
disable image shader on adreno until a better workaround figured out
2020-10-10 00:09:45 +08:00
nihuini
b766c8cd9e
fix potential divide by zero fault when bf16s / fp16s enabled, fix #2125
2020-09-16 16:28:59 +08:00
Leo
5afd318b86
Support remove libstdc++ denpendency ( #2030 )
...
* [build] add toolchain file w/o stdcxx dependency
* [build] link m and gcc lib explicitly
* [ncnn] complete simple stl impl
* [ncnn] adapt for ncnn simplestl
* [test] adapt for ncnn simplestl
* [ncnn] fix missing algorithm and list when simplestl disabled
* [ncnn] fix guard for operator new and delete
* [style] fix the code style
* [build] fix build failed on darwin and emscripten
* [ci] do not import cxx to avoid operator conflict
* [ncnn] add temporary partial_sort impl using bubble sort
heap sort should be used for better perf.
* [ncnn] add std greater and less function
* [ncnn] fix placement new operator overload
* [ncnn] add operator delete with size info
* [build] disable exception, rtti, example and tools when simplestl on
* [build] add toolchain for arm simplestl
* [build] add toolchain for aarch64 simplestl
* [ncnn] move initializer to constructor
* [ncnn] use deteiled type instead of auto
* [ncnn] use plain lib name in target_link_libraries
2020-08-28 09:20:56 +08:00
ncnnnnn
e2557c1678
fix UNIT64_MAX not declared #2009 ( #2010 )
2020-08-07 16:04:18 +08:00
nihui
3ef995ed1e
format code style and setup restyled.io ( #1840 )
2020-06-13 22:17:51 +08:00
nihui
478e15d7dc
fix memorydata test on nvidia gpu
2020-06-03 00:31:58 +08:00
nihuini
f87f21779f
resolve cast from type properly, no more fp16p to/from fp16s conversion
2020-06-02 19:54:54 +08:00
nihuini
bb56b5439f
fix vkmat download on integrated gpu, workaround priorbox fp16s with online spirv, fix #1700 fix #1805
2020-06-02 15:12:17 +08:00
nihui
2133121e6d
destroy descriptorsets properly
2020-05-27 23:39:20 +08:00
nihui
85fb76a6b9
clear all internal state on command reset
2020-05-27 22:09:10 +08:00
Naiyang Lin
ceef2470a5
Add logger.h ( #1753 )
2020-05-12 19:33:21 +08:00
nihui
9a9a618229
image storage is mandatory, less options makes life easier
2020-05-08 00:09:35 +08:00
nihuini
041437ef48
seperate packing cast type to more shaders
2020-05-07 17:27:23 +08:00
nihui
e8688b042f
fuse packing cast storage, binaryop image shader, dummy buffer and image, device-wide utility packing converter operators, fix multi-blob layer test
2020-05-05 21:31:07 +08:00
nihui
62da1228e1
adreno image shader + fp16 + fp16a ( #1714 )
...
* wip
* wip
* fix
* image and imageview can not be destroyed until command execution ends
* fast copy path for tightly packed data
* wip
* texture load works
* 1d 3d image
* record clone image, multiple commands share one image reference
* upload download image
* layer forward accept vkimagemat
* vkimagemat graph works
* staging vkimagemat for passing dynamic parameters, macro for fp32+image shader, padding image shader
* vkimagemat elemsize
* convolution test pass
* conv1x1s1 image shader
* fast staging image allocator from host memory, pooling image shader
* convolutiondepthwise image shader
* innerproduct image shader
* packing image shader
* crop deconvolution image shader
* resolve spirv binding types
* image fp16 and fp16a, cast image shader
* eltwise image shader
* wip
* absval image shader
* deconvolutiondepthwise image shader
* concat image shader, squeezenet works
* noop split image shader
* uniform precision hint
* layer support_image_storage
* wip
* vulkan device utility operator
* command is storage and packing option aware
* fallback to cpu on image allocation failed, mobilenetssd works
* flatten image shader, enable more test
* ci test
* check imgfp32 imgfp16 imgfp16a features
* fix ci test
* fix ci test
* upgrade swiftshader
* wip
* opt aggressive
* imgfp16p
* opt none
* convolution winograd image shader
* fix flush range, fast copy path for continous buffer
* minor fix
* fix innerproduct
* wip ...
* wip
* cast fix
* packing test
* wip
* image fp16p is fp16p
* wip
* silence
* more line info
* code clean
* softmax image shader
2020-04-30 17:08:30 +08:00
nihui
7365bb80a2
vkmat and command api breaks ( #1689 )
...
* vkmat and command api breaks
* always use compute queue for compute buffer transfer
* no barrier for readonly weight buffer
* record clone, drop queue_owner
* bring back layer forward
* fix validation errors
* lifecycle inside command makes life easier
* update doc
* record_import_android_hardware_buffer
2020-04-12 13:14:20 +08:00
nihuini
1469bc8b19
reclaim immediately after submitting, so we don't block the queue access during the long-time waiting fence, fix #1682
2020-04-09 13:06:42 +08:00
nihuini
64ce9bc847
staging buffer has its offset
2020-01-07 19:24:49 +08:00
nihuini
b361b24832
do not enforce coherent memory type, queue transfer after uploading model weight
2020-01-02 16:51:33 +08:00
nihui
038666e049
the initial auto test ( #1464 )
...
* cpu test
* wip
* ci run test
* travis ci for arm64
* arm64 ctest
* copy vulkan loader
* wip
* run
* Update ccpp.yml
* gpu test
* swiftshader
* cache macos swiftshader
* try MoltenVK
* try vulkaninfo
* give swiftshader another try
* disable failed macos gpu test
* more conv test, fix conv3x3s1 gpu test fail
* fix deconvolution test
* dilation test
* cmake option to build tests
* ncnn_add_layer_test macro
* host barrier before upload and after download, handle packing layout option
* test packing layout
* wip
* wip
* merge deconvolution packing and non-packing code
* merge convolution packing and non-packing code
* pass top_blob_count param
* fix build
* take care of non-coherent mappable memory
2019-12-31 21:12:24 +08:00
nihuini
a86c2f44c3
vkimagemat, vkimageallocator, convenient construct from android hardware buffer
2019-12-25 14:34:34 +08:00
nihui
4a3c70f5b8
import android hardware buffer, fix #1329
2019-12-08 22:36:48 +08:00
nihuini
c4f23ae8ad
rename Mat packing to elempack
2019-07-22 21:19:01 +08:00
nihuini
73911492d7
fix validation warning on querypool destruction, enable fp16p by default
2019-06-10 19:51:27 +08:00
nihuini
e09607bc22
add option to upload model function, pipeline creation honors option use flags, setting allocator per extractor do not make much sense
2019-06-05 19:03:39 +08:00
nihuini
cd7559c639
more fix for fp16p, still disabled by default
2019-05-29 11:13:16 +08:00
nihui
25b9736f82
shader fp16 packed
2019-05-26 19:33:41 +08:00
nihuini
37e150162a
do not retrieve timestamp availabitliy bits
2019-05-22 14:45:00 +08:00
nihuini
738fb6bb14
print gpu per layer benchmark
2019-05-22 11:26:28 +08:00
nihuini
c9a9486307
merge command submit and wait, expose queue_count, concurrent queue submission shall work
2019-05-20 19:51:30 +08:00
nihuini
5d86014d9c
add missing barrier for transfer dst, fix softmax pack4, fix #932
2019-04-24 16:16:26 +08:00
nihuini
4729ea3505
bottom blob memory never alias, reuse blob memory more elegantly relying on refcount
2019-04-23 14:45:56 +08:00
nihui
d753fe2589
upload fp16 weight, enable fp16 storage and arithmetic
2019-03-23 10:34:55 +08:00
nihuini
37573aeeb5
remove unused record download
2019-03-22 11:16:51 +08:00
nihuini
d3a11eb6c9
one codepath for unified and discrete device
2019-03-13 19:14:16 +08:00
nihuini
433a92401a
auto barrier in pipeline and copy command
2019-03-13 17:33:36 +08:00
nihuini
85a28959e4
fix binaryop shader binding, use shared buffer state, fix blob copy in non-light mode, fix #817
2019-03-07 15:18:48 +08:00
nihui
b49cb56ad9
constify vulkan device handle, use default local vulkan device if not specified
2019-03-03 11:27:04 +08:00
nihui
68afd1fa17
reset fence
2019-03-02 15:51:29 +08:00
nihuini
bdd305638d
command reset
2019-03-01 20:49:28 +08:00
nihuini
7312887671
transfer command hold data context
2019-02-20 14:52:53 +08:00
nihui