• WuZhiwen's avatar
    Merge pull request #12703 from wzw-intel:vkcom · 6e3ea8b4
    WuZhiwen authored
    * 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: 's avatarWu 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: 's avatarWu 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: 's avatarWu 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.
    6e3ea8b4
op_vkcom.cpp 4.28 KB
// This file is part of OpenCV project.
// It is subject to the license terms in the LICENSE file found in the top-level directory
// of this distribution and at http://opencv.org/license.html.
//
// Copyright (C) 2018, Intel Corporation, all rights reserved.
// Third party copyrights are property of their respective owners.

#include "precomp.hpp"
#include <opencv2/dnn/shape_utils.hpp>
#include "op_vkcom.hpp"

namespace cv
{
namespace dnn
{
#ifdef HAVE_VULKAN
    void copyToTensor(vkcom::Tensor &dst, const Mat &src)
    {
        CV_Assert(src.isContinuous() && src.type() == CV_32F);

        std::vector<int> mat_shape = shape(src);
        dst.reshape((const char*)src.data, mat_shape);
    }

    void copyToMat(Mat &dst, vkcom::Tensor &src)
    {
        CV_Assert(dst.type() == CV_32F);

        std::vector<int> shape = src.getShape();
        void *data = src.map();
        Mat tmp(shape, CV_32F, data);
        tmp.copyTo(dst);
        src.unMap();
    }

    vkcom::Tensor VkComTensor(const Ptr<BackendWrapper>& ptr)
    {
        CV_Assert(!ptr.empty());
        return ptr.dynamicCast<VkComBackendWrapper>()->getTensor();
    }

    void setDirty(std::vector<Ptr<BackendWrapper> >& ptrs)
    {
        for (const Ptr<BackendWrapper>& ptr : ptrs)
        {
            ptr.dynamicCast<VkComBackendWrapper>()->setDeviceDirty();
        }
    }

    std::vector<vkcom::Tensor> VkComTensors(const std::vector<Ptr<BackendWrapper> >& ptrs)
    {
        std::vector<vkcom::Tensor> vec;
        vec.reserve(ptrs.size());
        for (const Ptr<BackendWrapper>& ptr : ptrs)
        {
            vec.push_back(VkComTensor(ptr));
        }
        return vec;
    }

    VkComBackendNode::VkComBackendNode(const std::vector<Ptr<BackendWrapper> >& inputsWrapper,
                                       const std::shared_ptr<vkcom::OpBase>& op,
                                       const std::vector<Ptr<BackendWrapper> >& blobsWrapper)
                                       : BackendNode(DNN_BACKEND_VKCOM)
    {
        operation = op;

        inputsWrapper_ = inputsWrapper;
        ins = VkComTensors(inputsWrapper_);

        if (!blobsWrapper.empty())
        {
            blobs = VkComTensors(blobsWrapper);
        }
    }

    bool VkComBackendNode::forward(std::vector<vkcom::Tensor>& outs)
    {
        for (int i = 0, n = inputsWrapper_.size(); i < n; ++i)
        {
            inputsWrapper_[i].dynamicCast<VkComBackendWrapper>()->copyToDevice();
        }

        return operation->forward(ins, blobs, outs);
    }

    VkComBackendWrapper::VkComBackendWrapper(Mat& m) : BackendWrapper(DNN_BACKEND_VKCOM, DNN_TARGET_VULKAN)
    {
        copyToTensor(tensor, m);
        host = &m;
        hostDirty = false;
        deviceDirty = false;
    }

    VkComBackendWrapper::VkComBackendWrapper(const Ptr<BackendWrapper>& baseBuffer, Mat& m)
        : BackendWrapper(DNN_BACKEND_VKCOM, DNN_TARGET_VULKAN)
    {
        Ptr<VkComBackendWrapper> base = baseBuffer.dynamicCast<VkComBackendWrapper>();
        CV_Assert(!base.empty());

        host = &m;
        tensor = base->tensor;
        CV_Assert(tensor.count() >= m.total());
        tensor.reshape(0, shape(m));
        hostDirty = false;
        deviceDirty = false;
    }

    void VkComBackendWrapper::copyToHost()
    {
        if (deviceDirty)
            copyToMat(*host, tensor);
    }

    void VkComBackendWrapper::setHostDirty()
    {
        hostDirty = true;
    };

    void VkComBackendWrapper::setDeviceDirty()
    {
        deviceDirty = true;
    };

    void VkComBackendWrapper::copyToDevice()
    {
        if (hostDirty)
        {
            copyToTensor(tensor, *host);
            hostDirty = false;
        }
    }

    vkcom::Tensor VkComBackendWrapper::getTensor()
    {
        return tensor;
    }
#endif
    void forwardVkCom(std::vector<Ptr<BackendWrapper> > &outputs,
                      const Ptr<BackendNode>& node)
    {
#ifdef HAVE_VULKAN
        CV_Assert(!node.empty());

        Ptr<VkComBackendNode> node_ = node.dynamicCast<VkComBackendNode>();
        std::vector<vkcom::Tensor> outs = VkComTensors(outputs);
        node_->forward(outs);
        setDirty(outputs);
#endif
    }

    bool haveVulkan()
    {
#ifdef HAVE_VULKAN
        return vkcom::isAvailable();
#else
        return false;
#endif  // HAVE_VULKAN
    }

}  // namespace dnn
}  // namespace cv