René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Request: WARP support and fix WARP rendering bugs..","articleBody":"Hi,\nI tested sample on warp 1.0.15 which supports cooperative vectors and is rendering incorrectly..\nnote I have sample correctly rendering on NVidia..\n\nsee screenshot of rendering incorrectly on WARP:\n![Image](https://github.com/user-attachments/assets/0a3db909-4e49-4ad4-b7a4-eba1962b91d2)\n\nfor testing although sample says you can select adapter it isn't used..\n\n\u003e--adapter-id Integer that allows to pick the desired GPU.\n\nI launch:\n\n\u003edino_danger.exe --data-dir ../../.. --adapter-id 1\n\nand still selects GPU (adapter 0)..\n\nto fix I changed code in two places (to allow software renderer) and to pass adapterIndex:\n\nfix1 (allow software renderer):\n```\n--- a/sdk/src/dx12/dx12_backend_graphics_device.cpp\n+++ b/sdk/src/dx12/dx12_backend_graphics_device.cpp\n@@ -133,7 +133,7 @@ namespace d3d12\n                 dxgiAdapter1-\u003eGetDesc1(\u0026dxgiAdapterDesc1);\n \n                 // First is it usable?\n-                bool usableGPU = (dxgiAdapterDesc1.Flags \u0026 DXGI_ADAPTER_FLAG_SOFTWARE) == 0 \u0026\u0026 SUCCEEDED(D3D12CreateDevice(dxgiAdapter1, D3D_FEATURE_LEVEL_12_1, __uuidof(ID3D12Device), nullptr));\n+                bool usableGPU = /*(dxgiAdapterDesc1.Flags \u0026 DXGI_ADAPTER_FLAG_SOFTWARE) == 0 \u0026\u0026*/ SUCCEEDED(D3D12CreateDevice(dxgiAdapter1, D3D_FEATURE_LEVEL_12_1, __uuidof(ID3D12Device), nullptr));\n \n                 // Is it the right vendor?\n                 if (usableGPU \u0026\u0026 adapterIdx == adapterID)\n```\nfix2 (pass adapterIndex):\n```\n--- a/sdk/src/render_pipeline/dino_renderer.cpp\n+++ b/sdk/src/render_pipeline/dino_renderer.cpp\n@@ -56,7 +56,10 @@ void DinoRenderer::initialize(uint64_t hInstance, const CommandLineOptions\u0026 opti\n     graphics::setup_graphics_api(GraphicsAPI::DX12);\n     // graphics::device::enable_debug_layer();\n     graphics::device::enable_experimental_features();\n-    m_Device = graphics::device::create_graphics_device();\n+    if(options.adapterIndex==0)\n+        m_Device = graphics::device::create_graphics_device();\n+    else\n+        m_Device = graphics::device::create_graphics_device(DevicePickStrategy::AdapterID, options.adapterIndex);\n     m_Window = graphics::window::create_window(m_Device, (uint64_t)hInstance, 1920, 1080, \"Intel TSNC (DX12)\");\n     m_CmdQueue = graphics::command_queue::create_command_queue(m_Device);\n     m_SwapChain = graphics::swap_chain::create_swap_chain(m_Window, m_Device, m_CmdQueue, FRAME_BUFFER_FORMAT);\n```\nwith that I was getting also a crash with WARP and was fixed with that:\n\n```\n--- a/sdk/src/dx12/dx12_backend_window.cpp\n+++ b/sdk/src/dx12/dx12_backend_window.cpp\n@@ -113,12 +113,12 @@ namespace d3d12\n \t\t\t\twindowClass.cbClsExtra = 0;\n \t\t\t\twindowClass.cbWndExtra = 0;\n \t\t\t\twindowClass.hInstance = hInst;\n-\t\t\t\twindowClass.hIcon = LoadIcon(hInst, NULL);\n+\t\t\t\twindowClass.hIcon = NULL;// LoadIcon(hInst, NULL);\n \t\t\t\twindowClass.hCursor = LoadCursor(NULL, IDC_ARROW);\n \t\t\t\twindowClass.hbrBackground = (HBRUSH)(COLOR_WINDOW + 1);\n \t\t\t\twindowClass.lpszMenuName = NULL;\n \t\t\t\twindowClass.lpszClassName = windowClassName;\n-\t\t\t\twindowClass.hIconSm = LoadIcon(hInst, NULL);\n+\t\t\t\twindowClass.hIconSm = NULL;// LoadIcon(hInst, NULL);\n \t\t\t\tassert_msg(RegisterClassExA(\u0026windowClass) \u003e 0, \"RegisterClassExA failed.\");\n \t\t\t\tclassRegistered = true;\n \t\t\t}\n```\n\nfinally launches but renders incorrectly don't know if WARP bug or not..\nalso perhaps you can \"massage\" code to make it work on WARP similar to have fixed recently for other GPUs in( for Nvidia ?):\nhttps://github.com/GameTechDev/TextureSetNeuralCompressionSample/commit/ec738e6d2536472b633a915b5d28c488e756db8e\n\nEDIT: also note WARP without coop vector hangs:\n\u003edino_danger.exe --data-dir ../../.. --adapter-id 1 --disable-coop\n\nit's strange but as said using --disable-coop on WARP seems to hang.. or maybe is much slower and didn't wait for long as needed..\n\n\nscreenshot of rendering correctly on NV:\n\n![Image](https://github.com/user-attachments/assets/b42a77a4-f312-453d-bf76-56183b66f4d2)","author":{"url":"https://github.com/oscarbg","@type":"Person","name":"oscarbg"},"datePublished":"2025-06-19T00:38:07.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/2/TextureSetNeuralCompressionSample/issues/2"}