D3D11和D3D12共享资源

最新推荐文章于 2023-04-24 15:02:09 发布

sandmangu

最新推荐文章于 2023-04-24 15:02:09 发布

阅读量2.1k

点赞数 4

分类专栏： Direct3D 文章标签： DirectX11 DirectX12

本文链接：https://blog.youkuaiyun.com/sandmangu/article/details/129855826

版权

Direct3D 专栏收录该内容

2 篇文章

订阅专栏

最近碰到个伪需求: 游戏串流。游戏引擎用D3D12渲染，再把游戏画面做视频编码，通过网络发送到远端做解码显示。

第一反应就是走全GPU的流程，不要用CPU把显存里的数据拷来拷去。所以先获取渲染完的D3D12的frame buffer，然后送给Intel MediaSDK去做编码。查了一下MediaSDK文档，只支持D3D11的输入buffer，需要想办法把D3D12 Resource转换成D3D11 Texture2D。可以试试D3D11/D3D12的Texture2D资源共享。查了一下网上的讨论的帖子 Sharing ID3D11Buffer and ID3D12Resource，微软自家的DirectX不同版本间可以用CreateSharedHandle/OpenSharedResource来互相访问。

DX12我不熟，先找个最简单的DX12 Texture的例子。我选了微软官方的例子 https://github.com/microsoft/DirectX-Graphics-Samples/tree/master/Samples/Desktop/D3D12HelloWorld/src/HelloTexture

有关的Texture2D的初始化代码在这里，m_texture就是一块黑白棋盘格样式的Texture2D，所有的初始化和像素填充都是基于D3D12的API来完成。然后在主循环里在WM_PAINT消息OnRender()函数里把m_texture贴到一个三角形上并显示

//创建m_texture, 一块256x256 R8G8B8A8的texture2D的资源, GPU访问
        // Describe and create a Texture2D.
        D3D12_RESOURCE_DESC textureDesc = {};
        textureDesc.MipLevels = 1;
        textureDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
        textureDesc.Width = TextureWidth;
        textureDesc.Height = TextureHeight;
        textureDesc.Flags = D3D12_RESOURCE_FLAG_NONE;
        textureDesc.DepthOrArraySize = 1;
        textureDesc.SampleDesc.Count = 1;
        textureDesc.SampleDesc.Quality = 0;
        textureDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D;

        ThrowIfFailed(m_device->CreateCommittedResource(
            &CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_DEFAULT),
            D3D12_HEAP_FLAG_NONE,
            &textureDesc,
            D3D12_RESOURCE_STATE_COPY_DEST,
            nullptr,
            IID_PPV_ARGS(&m_texture)));




        const UINT64 uploadBufferSize = GetRequiredIntermediateSize(m_texture.Get(), 0, 1);

//创建textureUploadHeap, 一块CPU可以读写的staging buffer
        // Create the GPU upload buffer.
        ThrowIfFailed(m_device->CreateCommittedResource(
            &CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_UPLOAD),
            D3D12_HEAP_FLAG_NONE,
            &CD3DX12_RESOURCE_DESC::Buffer(uploadBufferSize),
            D3D12_RESOURCE_STATE_GENERIC_READ,
            nullptr,
            IID_PPV_ARGS(&textureUploadHeap)));

//生成一块黑白相间的棋盘格样式的图像
        // Copy data to the intermediate upload heap and then schedule a copy 
        // from the upload heap to the Texture2D.
        std::vector<UINT8> texture = GenerateTextureData();

        D3D12_SUBRESOURCE_DATA textureData = {};
        textureData.pData = &texture[0];
        textureData.RowPitch = TextureWidth * TexturePixelSize;
        textureData.SlicePitch = textureData.RowPitch * TextureHeight;

//将棋盘格式的图像填充进textureUploadHeap里，再拷贝进m_texture内
        UpdateSubresources(m_commandList.Get(), m_texture.Get(), textureUploadHeap.Get(), 0, 0, 1, &textureData);
        m_commandList->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(m_texture.Get(), D3D12_RESOURCE_STATE_COPY_DEST, D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE));

运行的输出，可以看到三角形上贴的纹理是黑白棋盘格

接下来要做的是按照Sharing ID3D11Buffer and ID3D12Resource的流程在m_texture初始化后做一个和D3D11 Texture2D的共享，并且通过D3D11的API函数修改这个m_texture的像素纹理

1. 修改创建mtexture的参数，设为可共享

这里需要修改D3D12_RESOURCE_DESC的Flags属性为D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET | D3D12_RESOURCE_FLAG_ALLOW_SIMULTANEOUS_ACCESS
同时ID3D12Device::CreateCommittedResource()的HeapFlags参数设为D3D12_HEAP_FLAG_SHARED

		// Describe and create a Texture2D.
		D3D12_RESOURCE_DESC textureDesc = {};
		textureDesc.MipLevels = 1;
		textureDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
		textureDesc.Width = TextureWidth;
		textureDesc.Height = TextureHeight;
		textureDesc.Flags = D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET | D3D12_RESOURCE_FLAG_ALLOW_SIMULTANEOUS_ACCESS;
		textureDesc.DepthOrArraySize = 1;
		textureDesc.SampleDesc.Count = 1;
		textureDesc.SampleDesc.Quality = 0;
		textureDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D;

		ThrowIfFailed(m_device->CreateCommittedResource(
			&CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_DEFAULT),
			D3D12_HEAP_FLAG_SHARED,
			&textureDesc,
			D3D12_RESOURCE_STATE_COPY_DEST, // D3D12_RESOURCE_STATE_COPY_DEST,
			nullptr,
			IID_PPV_ARGS(&m_texture)));

2. 为m_texture创建一个共享句柄m_sharedTextureHandle

		HRESULT hr;
		HANDLE m_sharedTextureHandle;
		ThrowIfFailed(hr = m_device->CreateSharedHandle(
			m_texture.Get(),
			nullptr,
			GENERIC_ALL,
			nullptr,
			&m_sharedTextureHandle));

3. 创建一个新的d3d11 device和context，

		static D3D_FEATURE_LEVEL FeatureLevels[] = {
			D3D_FEATURE_LEVEL_11_1,
			D3D_FEATURE_LEVEL_11_0,
			D3D_FEATURE_LEVEL_10_1,
			D3D_FEATURE_LEVEL_10_0
		};

		IDXGIFactory2* DxgiFactory;
		IDXGIAdapter* DxgiAdapter;
		DXGI_ADAPTER_DESC DxgiAdapterDesc;
		ID3D11Texture2D* Texture;

		ID3D11Device* D3d11Device;
		D3D_FEATURE_LEVEL FeatureLevel;
		ID3D11DeviceContext* D3d11DeviceContext;

		hr = CreateDXGIFactory2(0, IID_PPV_ARGS(&DxgiFactory));
		hr = DxgiFactory->EnumAdapters(0, &DxgiAdapter);
		hr = DxgiAdapter->GetDesc(&DxgiAdapterDesc);
		hr = D3D11CreateDevice(DxgiAdapter, D3D_DRIVER_TYPE_UNKNOWN, nullptr, D3D11_CREATE_DEVICE_DEBUG, FeatureLevels, 4, D3D11_SDK_VERSION, &D3d11Device, &FeatureLevel, &D3d11DeviceContext);

创建一个和m_texture一样大小的D3D11 Texture2D对象，填充成灰红蓝绿色，等下用ID3D11DeviceContext::CopyResource()来覆盖d3d12的m_texture,用来验证共享的texture2D资源是否成功

//这里需要创建一个和DX12纹理资源相同尺寸的Texture2D
//图省事，没有提取前面DX12的纹理资源的尺寸，直接定义个相同的尺寸
#define SHARED_TEXTURE_WIDTH   256
#define SHARED_TEXTURE_HEIGHT  256

		CD3D11_TEXTURE2D_DESC TextureDesc(DXGI_FORMAT_R8G8B8A8_UNORM, SHARED_TEXTURE_WIDTH, SHARED_TEXTURE_HEIGHT, 1, 1);
		//TextureDesc.MiscFlags = D3D11_RESOURCE_MISC_SHARED | D3D11_RESOURCE_MISC_SHARED_NTHANDLE;

		unsigned long texture_data_1[SHARED_TEXTURE_WIDTH * SHARED_TEXTURE_HEIGHT];

		D3D11_SUBRESOURCE_DATA Data = {};
		//将texture的左上，右上，左下，右下分别填充成灰红绿蓝
		for (int i = 0; i < SHARED_TEXTURE_HEIGHT; i++)
		{
			for (int j = 0; j < SHARED_TEXTURE_WIDTH; j++)
			{
				if (i >= SHARED_TEXTURE_HEIGHT / 2)
				{
					if (j >= SHARED_TEXTURE_WIDTH / 2)
					{
						texture_data_1[i * SHARED_TEXTURE_WIDTH + j] = 0xFF00FF00;		//G	Formet:0xAaBbGgRr
					}
					else
					{
						texture_data_1[i * SHARED_TEXTURE_WIDTH + j] = 0xFFFF0000;		//B
					}
				}
				else
				{
					if (j >= SHARED_TEXTURE_WIDTH / 2)
					{
						texture_data_1[i * SHARED_TEXTURE_WIDTH + j] = 0xFF0000FF;		//R
					}
					else
					{
						texture_data_1[i * SHARED_TEXTURE_WIDTH + j] = 0xFF808080;		//Gray
					}
				}
			}
		}
		Data.pSysMem = texture_data_1;
		Data.SysMemPitch = SHARED_TEXTURE_WIDTH * sizeof(unsigned long);		//п
		Data.SysMemSlicePitch = 0;

		hr = D3d11Device->CreateTexture2D(&TextureDesc, &Data, &Texture);

4. 在d3d11下用m_texture的共享句柄创建一个共享设备, 对应的是D3D11Texture2D类型的SharedTexture

		ID3D11Device1* pDevice1;

		hr = D3d11Device->QueryInterface(__uuidof(ID3D11Device1), (void**)&pDevice1);

		ComPtr<ID3D11Texture2D> SharedTexture;
		hr = pDevice1->OpenSharedResource1(m_sharedTextureHandle, IID_PPV_ARGS(SharedTexture.GetAddressOf()));

5. 修改SharedTexture的内容, 对应的D3D12的m_texture的内容也同步变了。

		D3d11DeviceContext->CopyResource(SharedTexture.Get(), Texture);
		D3d11DeviceContext->Flush();

最后运行一下修改过的程序, 可以看到三角形上的纹理变成了我们在D3D11下面创建的灰红蓝绿的纹理，看来共享成功了

搞定收工

最后老规矩，代码奉上，仅供参考

https://gitee.com/tisandman/dx12-hello-texture/tree/master

PS:

D3D12和D3D11下面的command queue应该是2个独立的queue, 所以操作D3D11的命令一定等D3D12的CommandList执行完了以后再执行，否则会互相影响 (我开始的时候把D3D11的所有操作放到了 D3D12 commandQueue的ExecuteCommandLists()前面了，结果所有的D3D11下的操作先执行，D3D12的操作后执行，导致对D3D11 Texture2D的所有像素改动操作都被覆盖了，查了好久代码)。话说也没有自带个靠谱的同步信号量
微软自家的框架之间的共享真是好弄，毕竟是同一家的东西。
D3D12的编程思路一点不像以前版本的DirectX，反倒是和前面的Vulkan的操作顺序很像...