potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-29 15:23:05 +04:00
Code Packages Releases Wiki Activity

Merge pull request #12703 from wzw-intel:vkcom

Browse Source 
* dnn: Add a Vulkan based backend

This commit adds a new backend "DNN_BACKEND_VKCOM" and a
new target "DNN_TARGET_VULKAN". VKCOM means vulkan based
computation library.

This backend uses Vulkan API and SPIR-V shaders to do
the inference computation for layers. The layer types
that implemented in DNN_BACKEND_VKCOM include:
Conv, Concat, ReLU, LRN, PriorBox, Softmax, MaxPooling,
AvePooling, Permute

This is just a beginning work for Vulkan in OpenCV DNN,
more layer types will be supported and performance
tuning is on the way.

Signed-off-by: Wu Zhiwen <zhiwen.wu@intel.com>

* dnn/vulkan: Add FindVulkan.cmake to detect Vulkan SDK

In order to build dnn with Vulkan support, need installing
Vulkan SDK and setting environment variable "VULKAN_SDK" and
add "-DWITH_VULKAN=ON" to cmake command.

You can download Vulkan SDK from:
https://vulkan.lunarg.com/sdk/home#linux

For how to install, see
https://vulkan.lunarg.com/doc/sdk/latest/linux/getting_started.html
https://vulkan.lunarg.com/doc/sdk/latest/windows/getting_started.html
https://vulkan.lunarg.com/doc/sdk/latest/mac/getting_started.html
respectively for linux, windows and mac.

To run the vulkan backend, also need installing mesa driver.
On Ubuntu, use this command 'sudo apt-get install mesa-vulkan-drivers'

To test, use command '$BUILD_DIR/bin/opencv_test_dnn --gtest_filter=*VkCom*'

Signed-off-by: Wu Zhiwen <zhiwen.wu@intel.com>

* dnn/Vulkan: dynamically load Vulkan runtime

No compile-time dependency on Vulkan library.
If Vulkan runtime is unavailable, fallback to CPU path.

Use environment "OPENCL_VULKAN_RUNTIME" to specify path to your
own vulkan runtime library.

Signed-off-by: Wu Zhiwen <zhiwen.wu@intel.com>

* dnn/Vulkan: Add a python script to compile GLSL shaders to SPIR-V shaders

The SPIR-V shaders are in format of text-based 32-bit hexadecimal
numbers, and inserted into .cpp files as unsigned int32 array.

* dnn/Vulkan: Put Vulkan headers into 3rdparty directory and some other fixes

Vulkan header files are copied from
https://github.com/KhronosGroup/Vulkan-Docs/tree/master/include/vulkan
to 3rdparty/include

Fix the Copyright declaration issue.

Refine OpenCVDetectVulkan.cmake

* dnn/Vulkan: Add vulkan backend tests into existing ones.

Also fixed some test failures.

- Don't use bool variable as uniform for shader
- Fix dispathed group number beyond max issue
- Bypass "group > 1" convolution. This should be support in future.

* dnn/Vulkan: Fix multiple initialization in one thread.
This commit is contained in:
WuZhiwen
2018-10-29 14:51:26 +00:00
committed by Alexander Alekhin
parent 220b278575
commit 6e3ea8b49d
89 changed files with 16275 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/dnn/src/layers/convolution_layer.cpp
+71 -1
View File
@@ -44,6 +44,7 @@
#include "layers_common.hpp"
#include "../op_halide.hpp"
#include "../op_inf_engine.hpp"
#include "../op_vkcom.hpp"
#include "opencv2/core/hal/hal.hpp"
#include "opencv2/core/hal/intrin.hpp"
#include <iostream>
@@ -222,7 +223,9 @@ public:
if (backendId == DNN_BACKEND_INFERENCE_ENGINE)
return preferableTarget != DNN_TARGET_MYRIAD || dilation.width == dilation.height;
else
return backendId == DNN_BACKEND_OPENCV || backendId == DNN_BACKEND_HALIDE;
return backendId == DNN_BACKEND_OPENCV ||
backendId == DNN_BACKEND_HALIDE ||
backendId == DNN_BACKEND_VKCOM && haveVulkan();
}
bool getMemoryShapes(const std::vector<MatShape> &inputs,
@@ -384,6 +387,73 @@ public:
biasvec[outCn] = biasvec[outCn+1] = biasvec[outCn-1];
}
virtual Ptr<BackendNode> initVkCom(const std::vector<Ptr<BackendWrapper> > &inputs) CV_OVERRIDE
{
#ifdef HAVE_VULKAN
int out_channel = blobs[0].size[0];
bool has_bias = hasBias() || fusedBias;
int filter_size[2] = {kernel.height, kernel.width};
int pad_size[2] = {pad.height, pad.width};
int stride_size[2] = {stride.height, stride.width};
int dilation_size[2] = {dilation.height, dilation.width};
int activation = 0;
vkcom::Tensor input_tensor = VkComTensor(inputs[0]);
int in_channel = input_tensor.dimSize(1);
int group = in_channel / blobs[0].size[1];
// TODO: support group > 1
if (group != 1)
return Ptr<BackendNode>();
int padding_mode;
if (padMode.empty())
{
padding_mode = vkcom::kPaddingModeCaffe;
}
else if (padMode == "VALID")
{
padding_mode = vkcom::kPaddingModeValid;
}
else if (padMode == "SAME")
{
padding_mode = vkcom::kPaddingModeSame;
}
else
CV_Error(Error::StsError, "Unsupported padding mode " + padMode);
std::shared_ptr<vkcom::OpBase> op(new vkcom::OpConv(out_channel, has_bias,
filter_size, pad_size,
stride_size, dilation_size,
activation, group,
padding_mode));
std::vector<Ptr<BackendWrapper> > blobsWrapper;
if (newWeightAndBias)
{
Mat wm;
weightsMat.copyTo(wm); // to handle the case of isContinuous() == false
wm.reshape(1, blobs[0].dims, blobs[0].size);
blobsWrapper.push_back(Ptr<BackendWrapper>(new VkComBackendWrapper(wm)));
}
else
{
blobsWrapper.push_back(Ptr<BackendWrapper>(new VkComBackendWrapper(blobs[0])));
}
if (has_bias)
{
Mat biasesMat({out_channel}, CV_32F, &biasvec[0]);
blobsWrapper.push_back(Ptr<BackendWrapper>(new VkComBackendWrapper(biasesMat)));
}
return Ptr<BackendNode>(new VkComBackendNode(inputs, op, blobsWrapper));
#endif // HAVE_VULKAN
return Ptr<BackendNode>();
}
virtual Ptr<BackendNode> initHalide(const std::vector<Ptr<BackendWrapper> > &inputs) CV_OVERRIDE
{
#ifdef HAVE_HALIDE