ShuriZma
/
duckstation
mirror of https://github.com/stenzek/duckstation
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

D3D12

This commit is contained in:
Stenzek 2023-08-13 01:16:00 +10:00
parent 623317f9b6
commit dda2c85859
52 changed files with 5223 additions and 7288 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/common/common.props
View File

@ -11,7 +11,7 @@
<ItemDefinitionGroup>
<Link>
<AdditionalDependencies Condition="'$(Platform)'!='ARM64'">$(RootBuildDir)glad\glad.lib;$(RootBuildDir)glslang\glslang.lib;%(AdditionalDependencies)</AdditionalDependencies>
<AdditionalDependencies>$(RootBuildDir)zstd\zstd.lib;$(RootBuildDir)fmt\fmt.lib;$(RootBuildDir)zlib\zlib.lib;$(RootBuildDir)minizip\minizip.lib;$(RootBuildDir)lzma\lzma.lib;d3dcompiler.lib;d3d11.lib;%(AdditionalDependencies)</AdditionalDependencies>
<AdditionalDependencies>$(RootBuildDir)zstd\zstd.lib;$(RootBuildDir)fmt\fmt.lib;$(RootBuildDir)zlib\zlib.lib;$(RootBuildDir)minizip\minizip.lib;$(RootBuildDir)lzma\lzma.lib;Comctl32.lib;winhttp.lib;%(AdditionalDependencies)</AdditionalDependencies>
</Link>
</ItemDefinitionGroup>
</Project>

2
src/common/http_downloader_winhttp.cpp
View File

@ -10,8 +10,6 @@
#include <algorithm>
Log_SetChannel(HTTPDownloaderWinHttp);
#pragma comment(lib, "winhttp.lib")
namespace Common {
HTTPDownloaderWinHttp::HTTPDownloaderWinHttp() : HTTPDownloader() {}

1
src/common/win32_progress_callback.cpp
View File

@ -4,7 +4,6 @@
#include "win32_progress_callback.h"
#include "common/log.h"
#include <CommCtrl.h>
#pragma comment(lib, "Comctl32.lib")
Log_SetChannel(Win32ProgressCallback);
Win32ProgressCallback::Win32ProgressCallback() : BaseProgressCallback()

8
src/core/core.props
View File

@ -9,7 +9,7 @@
<PreprocessorDefinitions Condition="('$(Platform)'=='x64' Or '$(Platform)'=='ARM' Or '$(Platform)'=='ARM64')">WITH_RECOMPILER=1;%(PreprocessorDefinitions)</PreprocessorDefinitions>
<PreprocessorDefinitions Condition="('$(Platform)'=='x64' Or '$(Platform)'=='ARM64')">WITH_MMAP_FASTMEM=1;%(PreprocessorDefinitions)</PreprocessorDefinitions>
<AdditionalIncludeDirectories>$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xxhash\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\rcheevos\include;$(SolutionDir)dep\rapidjson\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
<AdditionalIncludeDirectories>$(SolutionDir)dep\tinyxml2\include;$(SolutionDir)dep\glad\include;$(SolutionDir)dep\stb\include;$(SolutionDir)dep\imgui\include;$(SolutionDir)dep\xxhash\include;$(SolutionDir)dep\zlib\include;$(SolutionDir)dep\d3d12ma\include;$(SolutionDir)dep\rcheevos\include;$(SolutionDir)dep\rapidjson\include;$(SolutionDir)src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
<AdditionalIncludeDirectories Condition="'$(Platform)'!='ARM64'">$(SolutionDir)dep\rainterface;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
<AdditionalIncludeDirectories Condition="'$(Platform)'=='x64'">$(SolutionDir)dep\xbyak\xbyak;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
@ -19,9 +19,11 @@
<ItemDefinitionGroup>
<Lib>
<AdditionalDependencies>$(RootBuildDir)tinyxml2\tinyxml2.lib;$(RootBuildDir)rcheevos\rcheevos.lib;$(RootBuildDir)imgui\imgui.lib;$(RootBuildDir)stb\stb.lib;$(RootBuildDir)xxhash\xxhash.lib;$(RootBuildDir)zlib\zlib.lib;$(RootBuildDir)util\util.lib;$(RootBuildDir)common\common.lib;%(AdditionalDependencies)</AdditionalDependencies>
<AdditionalDependencies Condition="'$(Platform)'!='ARM64'">$(RootBuildDir)rainterface\rainterface.lib;%(AdditionalDependencies)</AdditionalDependencies>
<AdditionalDependencies>$(RootBuildDir)tinyxml2\tinyxml2.lib;$(RootBuildDir)rcheevos\rcheevos.lib;$(RootBuildDir)imgui\imgui.lib;$(RootBuildDir)stb\stb.lib;$(RootBuildDir)xxhash\xxhash.lib;$(RootBuildDir)zlib\zlib.lib;$(RootBuildDir)d3d12ma\d3d12ma.lib;$(RootBuildDir)util\util.lib;$(RootBuildDir)common\common.lib;d3d11.lib;d3d12.lib;d3dcompiler.lib;dxgi.lib;Dwmapi.lib;%(AdditionalDependencies)</AdditionalDependencies>
<AdditionalDependencies Condition="'$(Platform)'!='ARM64'">$(RootBuildDir)rainterface\rainterface.lib;opengl32.lib;%(AdditionalDependencies)</AdditionalDependencies>
<AdditionalDependencies Condition="'$(Platform)'=='ARM64'">$(RootBuildDir)vixl\vixl.lib;%(AdditionalDependencies)</AdditionalDependencies>
</Lib>
</ItemDefinitionGroup>
<Import Project="..\..\dep\msvc\winpixeventruntime\WinPixEventRuntime.props" Condition="$(Configuration.Contains(Debug))" />
</Project>

35
src/core/core.vcxproj
View File

@ -34,14 +34,13 @@
<ClCompile Include="digital_controller.cpp" />
<ClCompile Include="game_database.cpp" />
<ClCompile Include="gpu\d3d11_device.cpp" />
<ClCompile Include="gpu\d3d12\context.cpp" />
<ClCompile Include="gpu\d3d12\descriptor_heap_manager.cpp" />
<ClCompile Include="gpu\d3d12\shader_cache.cpp" />
<ClCompile Include="gpu\d3d12\staging_texture.cpp" />
<ClCompile Include="gpu\d3d12\stream_buffer.cpp" />
<ClCompile Include="gpu\d3d12\texture.cpp" />
<ClCompile Include="gpu\d3d12\util.cpp" />
<ClCompile Include="gpu\d3d12_gpu_device.cpp" />
<ClCompile Include="gpu\d3d12_builders.cpp" />
<ClCompile Include="gpu\d3d12_descriptor_heap_manager.cpp" />
<ClCompile Include="gpu\d3d12_device.cpp" />
<ClCompile Include="gpu\d3d12_pipeline.cpp" />
<ClCompile Include="gpu\d3d12_stream_buffer.cpp" />
<ClCompile Include="gpu\d3d12_texture.cpp" />
<ClCompile Include="gpu\d3d_common.cpp" />
<ClCompile Include="gpu\gl\context.cpp" />
<ClCompile Include="gpu\gl\context_egl.cpp">
<ExcludedFromBuild>true</ExcludedFromBuild>
@ -61,7 +60,6 @@
</ClCompile>
<ClCompile Include="gpu\gpu_device.cpp" />
<ClCompile Include="gpu\gpu_texture.cpp" />
<ClCompile Include="gpu\imgui_impl_dx12.cpp" />
<ClCompile Include="gpu\opengl_device.cpp" />
<ClCompile Include="gpu\opengl_pipeline.cpp" />
<ClCompile Include="gpu\opengl_stream_buffer.cpp" />
@ -81,7 +79,6 @@
<ClCompile Include="gpu\window_info.cpp" />
<ClCompile Include="gpu_backend.cpp" />
<ClCompile Include="gpu_commands.cpp" />
<ClCompile Include="gpu_hw_d3d12.cpp" />
<ClCompile Include="gpu_hw_shadergen.cpp" />
<ClCompile Include="gpu_sw.cpp" />
<ClCompile Include="gpu_sw_backend.cpp" />
@ -144,15 +141,13 @@
<ClInclude Include="digital_controller.h" />
<ClInclude Include="game_database.h" />
<ClInclude Include="gpu\d3d11_device.h" />
<ClInclude Include="gpu\d3d12\texture.h" />
<ClInclude Include="gpu\d3d_shaders.h" />
<ClInclude Include="gpu\d3d12\context.h" />
<ClInclude Include="gpu\d3d12\descriptor_heap_manager.h" />
<ClInclude Include="gpu\d3d12\shader_cache.h" />
<ClInclude Include="gpu\d3d12\staging_texture.h" />
<ClInclude Include="gpu\d3d12\stream_buffer.h" />
<ClInclude Include="gpu\d3d12\util.h" />
<ClInclude Include="gpu\d3d12_gpu_device.h" />
<ClInclude Include="gpu\d3d12_builders.h" />
<ClInclude Include="gpu\d3d12_descriptor_heap_manager.h" />
<ClInclude Include="gpu\d3d12_device.h" />
<ClInclude Include="gpu\d3d12_pipeline.h" />
<ClInclude Include="gpu\d3d12_stream_buffer.h" />
<ClInclude Include="gpu\d3d12_texture.h" />
<ClInclude Include="gpu\d3d_common.h" />
<ClInclude Include="gpu\gl\context.h" />
<ClInclude Include="gpu\gl\context_agl.h">
<ExcludedFromBuild>true</ExcludedFromBuild>
@ -175,7 +170,6 @@
</ClInclude>
<ClInclude Include="gpu\gpu_device.h" />
<ClInclude Include="gpu\gpu_texture.h" />
<ClInclude Include="gpu\imgui_impl_dx12.h" />
<ClInclude Include="gpu\opengl_device.h" />
<ClInclude Include="gpu\opengl_loader.h" />
<ClInclude Include="gpu\opengl_pipeline.h" />
@ -196,7 +190,6 @@
<ClInclude Include="gpu\vulkan_texture.h" />
<ClInclude Include="gpu\window_info.h" />
<ClInclude Include="gpu_backend.h" />
<ClInclude Include="gpu_hw_d3d12.h" />
<ClInclude Include="gpu_hw_shadergen.h" />
<ClInclude Include="gpu_sw.h" />
<ClInclude Include="gpu_sw_backend.h" />

104
src/core/core.vcxproj.filters
View File

@ -52,40 +52,18 @@
<ClCompile Include="libcrypt_serials.cpp" />
<ClCompile Include="texture_replacements.cpp" />
<ClCompile Include="multitap.cpp" />
<ClCompile Include="gpu_hw_d3d12.cpp" />
<ClCompile Include="host.cpp" />
<ClCompile Include="game_database.cpp" />
<ClCompile Include="pcdrv.cpp" />
<ClCompile Include="gpu\d3d11_device.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12_gpu_device.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\gpu_device.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\opengl_device.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12\shader_cache.cpp">
<Filter>gpu\d3d12</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12\staging_texture.cpp">
<Filter>gpu\d3d12</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12\stream_buffer.cpp">
<Filter>gpu\d3d12</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12\util.cpp">
<Filter>gpu\d3d12</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12\context.cpp">
<Filter>gpu\d3d12</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12\descriptor_heap_manager.cpp">
<Filter>gpu\d3d12</Filter>
</ClCompile>
<ClCompile Include="gpu\gl\context_egl_wayland.cpp">
<Filter>gpu\gl</Filter>
</ClCompile>
@ -116,9 +94,6 @@
<ClCompile Include="gpu\postprocessing_shadergen.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\imgui_impl_dx12.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\gpu_texture.cpp">
<Filter>gpu</Filter>
</ClCompile>
@ -134,9 +109,6 @@
<ClCompile Include="gpu\opengl_pipeline.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12\texture.cpp">
<Filter>gpu\d3d12</Filter>
</ClCompile>
<ClCompile Include="gpu\vulkan_loader.cpp">
<Filter>gpu</Filter>
</ClCompile>
@ -164,6 +136,27 @@
<ClCompile Include="gpu\window_info.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12_pipeline.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12_texture.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12_device.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12_builders.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12_descriptor_heap_manager.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d12_stream_buffer.cpp">
<Filter>gpu</Filter>
</ClCompile>
<ClCompile Include="gpu\d3d_common.cpp">
<Filter>gpu</Filter>
</ClCompile>
</ItemGroup>
<ItemGroup>
<ClInclude Include="types.h" />
@ -219,7 +212,6 @@
<ClInclude Include="texture_replacements.h" />
<ClInclude Include="shader_cache_version.h" />
<ClInclude Include="multitap.h" />
<ClInclude Include="gpu_hw_d3d12.h" />
<ClInclude Include="gdb_protocol.h" />
<ClInclude Include="host.h" />
<ClInclude Include="host_settings.h" />
@ -230,33 +222,12 @@
<ClInclude Include="gpu\d3d11_device.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12_gpu_device.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\gpu_device.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\opengl_device.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12\shader_cache.h">
<Filter>gpu\d3d12</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12\staging_texture.h">
<Filter>gpu\d3d12</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12\stream_buffer.h">
<Filter>gpu\d3d12</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12\util.h">
<Filter>gpu\d3d12</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12\context.h">
<Filter>gpu\d3d12</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12\descriptor_heap_manager.h">
<Filter>gpu\d3d12</Filter>
</ClInclude>
<ClInclude Include="gpu\gl\context_egl_wayland.h">
<Filter>gpu\gl</Filter>
</ClInclude>
@ -290,15 +261,9 @@
<ClInclude Include="gpu\postprocessing_shadergen.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\imgui_impl_dx12.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\gpu_texture.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d_shaders.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\gpu_shader_cache.h">
<Filter>gpu</Filter>
</ClInclude>
@ -314,9 +279,6 @@
<ClInclude Include="gpu\opengl_pipeline.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12\texture.h">
<Filter>gpu\d3d12</Filter>
</ClInclude>
<ClInclude Include="gpu\vulkan_loader.h">
<Filter>gpu</Filter>
</ClInclude>
@ -347,14 +309,32 @@
<ClInclude Include="gpu\window_info.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12_device.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12_pipeline.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12_texture.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12_stream_buffer.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12_builders.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d12_descriptor_heap_manager.h">
<Filter>gpu</Filter>
</ClInclude>
<ClInclude Include="gpu\d3d_common.h">
<Filter>gpu</Filter>
</ClInclude>
</ItemGroup>
<ItemGroup>
<Filter Include="gpu">
<UniqueIdentifier>{7b18fac1-ee2b-4816-8537-eb8f32d40ada}</UniqueIdentifier>
</Filter>
<Filter Include="gpu\d3d12">
<UniqueIdentifier>{17c2f89b-5cf4-4d3a-9343-a03c1c668b85}</UniqueIdentifier>
</Filter>
<Filter Include="gpu\gl">
<UniqueIdentifier>{e2aeacc6-65aa-4110-bef9-5f3e1fc0470e}</UniqueIdentifier>
</Filter>

406
src/core/gpu/d3d11_device.cpp
View File

@ -4,6 +4,7 @@
#include "d3d11_device.h"
#include "../host_settings.h"
#include "../shader_cache_version.h"
#include "d3d_common.h"
#include "postprocessing_chain.h" // TODO: Remove me
#include "common/align.h"
@ -22,16 +23,15 @@
Log_SetChannel(D3D11Device);
#pragma comment(lib, "d3d11.lib")
#pragma comment(lib, "dxgi.lib")
// We need to synchronize instance creation because of adapter enumeration from the UI thread.
static std::mutex s_instance_mutex;
static constexpr std::array<DXGI_FORMAT, static_cast<u32>(GPUTexture::Format::MaxCount)> s_dxgi_mapping = {
{DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B5G6R5_UNORM,
DXGI_FORMAT_B5G5R5A1_UNORM, DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_D16_UNORM}};
static constexpr std::array<float, 4> s_clear_color = {};
static unsigned s_next_bad_shader_id = 1;
static constexpr GPUTexture::Format s_swap_chain_format = GPUTexture::Format::RGBA8;
static void SetD3DDebugObjectName(ID3D11DeviceChild* obj, const std::string_view& name)
{
@ -367,70 +367,44 @@ void D3D11Device::SetVSync(bool enabled)
bool D3D11Device::CreateDevice(const std::string_view& adapter)
{
std::unique_lock lock(s_instance_mutex);
UINT create_flags = 0;
if (m_debug_device)
create_flags |= D3D11_CREATE_DEVICE_DEBUG;
ComPtr<IDXGIFactory> temp_dxgi_factory;
HRESULT hr = CreateDXGIFactory(IID_PPV_ARGS(temp_dxgi_factory.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("Failed to create DXGI factory: 0x%08X", hr);
m_dxgi_factory = D3DCommon::CreateFactory(m_debug_device);
if (!m_dxgi_factory)
return false;
}
u32 adapter_index;
if (!adapter.empty())
{
AdapterAndModeList adapter_info(GetAdapterAndModeList(temp_dxgi_factory.Get()));
for (adapter_index = 0; adapter_index < static_cast<u32>(adapter_info.adapter_names.size()); adapter_index++)
{
if (adapter == adapter_info.adapter_names[adapter_index])
break;
}
if (adapter_index == static_cast<u32>(adapter_info.adapter_names.size()))
{
// TODO: Log_Fmt
Log_WarningPrintf(
fmt::format("Could not find adapter '{}', using first ({})", adapter, adapter_info.adapter_names[0]).c_str());
adapter_index = 0;
}
}
else
{
Log_InfoPrintf("No adapter selected, using first.");
adapter_index = 0;
}
ComPtr<IDXGIAdapter> dxgi_adapter;
hr = temp_dxgi_factory->EnumAdapters(adapter_index, dxgi_adapter.GetAddressOf());
if (FAILED(hr))
Log_WarningPrintf("Failed to enumerate adapter %u, using default", adapter_index);
ComPtr<IDXGIAdapter1> dxgi_adapter = D3DCommon::GetAdapterByName(m_dxgi_factory.Get(), adapter);
static constexpr std::array<D3D_FEATURE_LEVEL, 3> requested_feature_levels = {
{D3D_FEATURE_LEVEL_11_0, D3D_FEATURE_LEVEL_10_1, D3D_FEATURE_LEVEL_10_0}};
ComPtr<ID3D11Device> device;
ComPtr<ID3D11DeviceContext> context;
hr =
ComPtr<ID3D11Device> temp_device;
ComPtr<ID3D11DeviceContext> temp_context;
HRESULT hr =
D3D11CreateDevice(dxgi_adapter.Get(), dxgi_adapter ? D3D_DRIVER_TYPE_UNKNOWN : D3D_DRIVER_TYPE_HARDWARE, nullptr,
create_flags, requested_feature_levels.data(), static_cast<UINT>(requested_feature_levels.size()),
D3D11_SDK_VERSION, device.GetAddressOf(), nullptr, context.GetAddressOf());
// we re-grab these later, see below
dxgi_adapter.Reset();
temp_dxgi_factory.Reset();
D3D11_SDK_VERSION, temp_device.GetAddressOf(), nullptr, temp_context.GetAddressOf());
if (FAILED(hr))
{
Log_ErrorPrintf("Failed to create D3D device: 0x%08X", hr);
return false;
}
else if (FAILED(hr = device.As(&m_device)) || FAILED(hr = context.As(&m_context)))
else if (FAILED(hr = temp_device.As(&m_device)) || FAILED(hr = temp_context.As(&m_context)))
{
Log_ErrorPrintf("Failed to get D3D11.1 device: 0x%08X", hr);
return false;
}
// we re-grab these later, see below
dxgi_adapter.Reset();
temp_context.Reset();
temp_device.Reset();
if (m_debug_device && IsDebuggerPresent())
{
ComPtr<ID3D11InfoQueue> info;
@ -447,31 +421,17 @@ bool D3D11Device::CreateDevice(const std::string_view& adapter)
m_context.As(&m_annotation);
#endif
// we need the specific factory for the device, otherwise MakeWindowAssociation() is flaky.
ComPtr<IDXGIDevice> dxgi_device;
if (FAILED(m_device.As(&dxgi_device)) || FAILED(dxgi_device->GetParent(IID_PPV_ARGS(dxgi_adapter.GetAddressOf()))) ||
FAILED(dxgi_adapter->GetParent(IID_PPV_ARGS(m_dxgi_factory.GetAddressOf()))))
{
Log_WarningPrint("Failed to get parent adapter/device/factory");
return false;
}
ComPtr<IDXGIDevice1> dxgi_device1;
if (SUCCEEDED(dxgi_device.As(&dxgi_device1)))
dxgi_device1->SetMaximumFrameLatency(1);
if (SUCCEEDED(m_device.As(&dxgi_device)) &&
SUCCEEDED(dxgi_device->GetParent(IID_PPV_ARGS(dxgi_adapter.GetAddressOf()))))
Log_InfoPrintf("D3D Adapter: %s", D3DCommon::GetAdapterName(dxgi_adapter.Get()).c_str());
else
Log_ErrorPrint("Failed to obtain D3D adapter name.");
DXGI_ADAPTER_DESC adapter_desc;
if (SUCCEEDED(dxgi_adapter->GetDesc(&adapter_desc)))
{
char adapter_name_buffer[128];
const int name_length =
WideCharToMultiByte(CP_UTF8, 0, adapter_desc.Description, static_cast<int>(std::wcslen(adapter_desc.Description)),
adapter_name_buffer, countof(adapter_name_buffer), 0, nullptr);
if (name_length >= 0)
{
adapter_name_buffer[name_length] = 0;
Log_InfoPrintf("D3D Adapter: %s", adapter_name_buffer);
}
}
BOOL allow_tearing_supported = false;
hr = m_dxgi_factory->CheckFeatureSupport(DXGI_FEATURE_PRESENT_ALLOW_TEARING, &allow_tearing_supported,
sizeof(allow_tearing_supported));
m_allow_tearing_supported = (SUCCEEDED(hr) && allow_tearing_supported == TRUE);
SetFeatures();
@ -486,6 +446,8 @@ bool D3D11Device::CreateDevice(const std::string_view& adapter)
void D3D11Device::DestroyDevice()
{
std::unique_lock lock(s_instance_mutex);
DestroyStagingBuffer();
DestroyBuffers();
m_context.Reset();
@ -515,103 +477,15 @@ void D3D11Device::SetFeatures()
m_features.supports_texture_buffers = true;
m_features.texture_buffers_emulated_with_ssbo = false;
m_features.gpu_timing = true;
m_allow_tearing_supported = false;
ComPtr<IDXGIFactory5> dxgi_factory5;
HRESULT hr = m_dxgi_factory.As(&dxgi_factory5);
if (SUCCEEDED(hr))
{
BOOL allow_tearing_supported = false;
hr = dxgi_factory5->CheckFeatureSupport(DXGI_FEATURE_PRESENT_ALLOW_TEARING, &allow_tearing_supported,
sizeof(allow_tearing_supported));
if (SUCCEEDED(hr))
m_allow_tearing_supported = (allow_tearing_supported == TRUE);
}
}
bool D3D11Device::GetRequestedExclusiveFullscreenModeDesc(IDXGIFactory5* factory, const RECT& window_rect, u32 width,
u32 height, float refresh_rate, DXGI_FORMAT format,
DXGI_MODE_DESC* fullscreen_mode, IDXGIOutput** output)
{
// We need to find which monitor the window is located on.
const Common::Rectangle<s32> client_rc_vec(window_rect.left, window_rect.top, window_rect.right, window_rect.bottom);
// The window might be on a different adapter to which we are rendering.. so we have to enumerate them all.
HRESULT hr;
ComPtr<IDXGIOutput> first_output, intersecting_output;
for (u32 adapter_index = 0; !intersecting_output; adapter_index++)
{
ComPtr<IDXGIAdapter1> adapter;
hr = factory->EnumAdapters1(adapter_index, adapter.GetAddressOf());
if (hr == DXGI_ERROR_NOT_FOUND)
break;
else if (FAILED(hr))
continue;
for (u32 output_index = 0;; output_index++)
{
ComPtr<IDXGIOutput> this_output;
DXGI_OUTPUT_DESC output_desc;
hr = adapter->EnumOutputs(output_index, this_output.GetAddressOf());
if (hr == DXGI_ERROR_NOT_FOUND)
break;
else if (FAILED(hr) || FAILED(this_output->GetDesc(&output_desc)))
continue;
const Common::Rectangle<s32> output_rc(output_desc.DesktopCoordinates.left, output_desc.DesktopCoordinates.top,
output_desc.DesktopCoordinates.right,
output_desc.DesktopCoordinates.bottom);
if (!client_rc_vec.Intersects(output_rc))
{
intersecting_output = std::move(this_output);
break;
}
// Fallback to the first monitor.
if (!first_output)
first_output = std::move(this_output);
}
}
if (!intersecting_output)
{
if (!first_output)
{
Log_ErrorPrintf("No DXGI output found. Can't use exclusive fullscreen.");
return false;
}
Log_WarningPrint("No DXGI output found for window, using first.");
intersecting_output = std::move(first_output);
}
DXGI_MODE_DESC request_mode = {};
request_mode.Width = width;
request_mode.Height = height;
request_mode.Format = format;
request_mode.RefreshRate.Numerator = static_cast<UINT>(std::floor(refresh_rate * 1000.0f));
request_mode.RefreshRate.Denominator = 1000u;
if (FAILED(hr = intersecting_output->FindClosestMatchingMode(&request_mode, fullscreen_mode, nullptr)) ||
request_mode.Format != format)
{
Log_ErrorPrintf("Failed to find closest matching mode, hr=%08X", hr);
return false;
}
*output = intersecting_output.Get();
intersecting_output->AddRef();
return true;
}
bool D3D11Device::CreateSwapChain()
{
constexpr DXGI_FORMAT swap_chain_format = DXGI_FORMAT_R8G8B8A8_UNORM;
if (m_window_info.type != WindowInfo::Type::Win32)
return false;
const DXGI_FORMAT dxgi_format = s_dxgi_mapping[static_cast<u8>(s_swap_chain_format)];
const HWND window_hwnd = reinterpret_cast<HWND>(m_window_info.window_handle);
RECT client_rc{};
GetClientRect(window_hwnd, &client_rc);
@ -624,9 +498,9 @@ bool D3D11Device::CreateSwapChain()
float fullscreen_refresh_rate;
m_is_exclusive_fullscreen =
GetRequestedExclusiveFullscreenMode(&fullscreen_width, &fullscreen_height, &fullscreen_refresh_rate) &&
GetRequestedExclusiveFullscreenModeDesc(m_dxgi_factory.Get(), client_rc, fullscreen_width, fullscreen_height,
fullscreen_refresh_rate, swap_chain_format, &fullscreen_mode,
fullscreen_output.GetAddressOf());
D3DCommon::GetRequestedExclusiveFullscreenModeDesc(m_dxgi_factory.Get(), client_rc, fullscreen_width,
fullscreen_height, fullscreen_refresh_rate, dxgi_format,
&fullscreen_mode, fullscreen_output.GetAddressOf());
}
else
{
@ -639,7 +513,7 @@ bool D3D11Device::CreateSwapChain()
DXGI_SWAP_CHAIN_DESC1 swap_chain_desc = {};
swap_chain_desc.Width = static_cast<u32>(client_rc.right - client_rc.left);
swap_chain_desc.Height = static_cast<u32>(client_rc.bottom - client_rc.top);
swap_chain_desc.Format = swap_chain_format;
swap_chain_desc.Format = dxgi_format;
swap_chain_desc.SampleDesc.Count = 1;
swap_chain_desc.BufferCount = 3;
swap_chain_desc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT;
@ -741,6 +615,7 @@ bool D3D11Device::CreateSwapChainRTV()
m_window_info.surface_width = backbuffer_desc.Width;
m_window_info.surface_height = backbuffer_desc.Height;
m_window_info.surface_format = s_swap_chain_format;
Log_VerbosePrintf("Swap chain buffer size: %ux%u", m_window_info.surface_width, m_window_info.surface_height);
if (m_window_info.type == WindowInfo::Type::Win32)
@ -912,85 +787,35 @@ void D3D11Device::EndPresent()
GPUDevice::AdapterAndModeList D3D11Device::StaticGetAdapterAndModeList()
{
ComPtr<IDXGIFactory> dxgi_factory;
HRESULT hr = CreateDXGIFactory(IID_PPV_ARGS(dxgi_factory.GetAddressOf()));
if (FAILED(hr))
return {};
AdapterAndModeList ret;
std::unique_lock lock(s_instance_mutex);
return GetAdapterAndModeList(dxgi_factory.Get());
// Device shouldn't be torn down since we have the lock.
if (g_gpu_device && g_gpu_device->GetRenderAPI() == RenderAPI::D3D12)
{
GetAdapterAndModeList(&ret, D3D11Device::GetInstance().m_dxgi_factory.Get());
}
else
{
ComPtr<IDXGIFactory5> factory = D3DCommon::CreateFactory(false);
if (factory)
GetAdapterAndModeList(&ret, factory.Get());
}
return ret;
}
GPUDevice::AdapterAndModeList D3D11Device::GetAdapterAndModeList(IDXGIFactory* dxgi_factory)
void D3D11Device::GetAdapterAndModeList(AdapterAndModeList* ret, IDXGIFactory5* factory)
{
AdapterAndModeList adapter_info;
ComPtr<IDXGIAdapter> current_adapter;
while (SUCCEEDED(dxgi_factory->EnumAdapters(static_cast<UINT>(adapter_info.adapter_names.size()),
current_adapter.ReleaseAndGetAddressOf())))
{
DXGI_ADAPTER_DESC adapter_desc;
std::string adapter_name;
if (SUCCEEDED(current_adapter->GetDesc(&adapter_desc)))
{
char adapter_name_buffer[128];
const int name_length = WideCharToMultiByte(CP_UTF8, 0, adapter_desc.Description,
static_cast<int>(std::wcslen(adapter_desc.Description)),
adapter_name_buffer, countof(adapter_name_buffer), 0, nullptr);
if (name_length >= 0)
adapter_name.assign(adapter_name_buffer, static_cast<size_t>(name_length));
else
adapter_name.assign("(Unknown)");
}
else
{
adapter_name.assign("(Unknown)");
}
if (adapter_info.fullscreen_modes.empty())
{
ComPtr<IDXGIOutput> output;
if (SUCCEEDED(current_adapter->EnumOutputs(0, &output)))
{
UINT num_modes = 0;
if (SUCCEEDED(output->GetDisplayModeList(DXGI_FORMAT_R8G8B8A8_UNORM, 0, &num_modes, nullptr)))
{
std::vector<DXGI_MODE_DESC> modes(num_modes);
if (SUCCEEDED(output->GetDisplayModeList(DXGI_FORMAT_R8G8B8A8_UNORM, 0, &num_modes, modes.data())))
{
for (const DXGI_MODE_DESC& mode : modes)
{
adapter_info.fullscreen_modes.push_back(GetFullscreenModeString(
mode.Width, mode.Height,
static_cast<float>(mode.RefreshRate.Numerator) / static_cast<float>(mode.RefreshRate.Denominator)));
}
}
}
}
}
// handle duplicate adapter names
if (std::any_of(adapter_info.adapter_names.begin(), adapter_info.adapter_names.end(),
[&adapter_name](const std::string& other) { return (adapter_name == other); }))
{
std::string original_adapter_name = std::move(adapter_name);
u32 current_extra = 2;
do
{
adapter_name = StringUtil::StdStringFromFormat("%s (%u)", original_adapter_name.c_str(), current_extra);
current_extra++;
} while (std::any_of(adapter_info.adapter_names.begin(), adapter_info.adapter_names.end(),
[&adapter_name](const std::string& other) { return (adapter_name == other); }));
}
adapter_info.adapter_names.push_back(std::move(adapter_name));
}
return adapter_info;
ret->adapter_names = D3DCommon::GetAdapterNames(factory);
ret->fullscreen_modes = D3DCommon::GetFullscreenModes(factory, {});
}
GPUDevice::AdapterAndModeList D3D11Device::GetAdapterAndModeList()
{
return GetAdapterAndModeList(m_dxgi_factory.Get());
AdapterAndModeList ret;
GetAdapterAndModeList(&ret, m_dxgi_factory.Get());
return ret;
}
bool D3D11Device::CreateTimestampQueries()
@ -1197,25 +1022,6 @@ std::unique_ptr<GPUSampler> D3D11Device::CreateSampler(const GPUSampler::Config&
D3D11_TEXTURE_ADDRESS_CLAMP, // ClampToEdge
D3D11_TEXTURE_ADDRESS_BORDER, // ClampToBorder
}};
D3D11_SAMPLER_DESC desc = {};
desc.AddressU = ta[static_cast<u8>(config.address_u.GetValue())];
desc.AddressV = ta[static_cast<u8>(config.address_v.GetValue())];
desc.AddressW = ta[static_cast<u8>(config.address_w.GetValue())];
desc.BorderColor[0] = static_cast<float>(config.border_color & 0xFF) / 255.0f;
desc.BorderColor[1] = static_cast<float>((config.border_color >> 8) & 0xFF) / 255.0f;
desc.BorderColor[2] = static_cast<float>((config.border_color >> 16) & 0xFF) / 255.0f;
desc.BorderColor[3] = static_cast<float>((config.border_color >> 24) & 0xFF) / 255.0f;
desc.MinLOD = static_cast<float>(config.min_lod);
desc.MaxLOD = static_cast<float>(config.max_lod);
if (config.anisotropy > 0)
{
desc.Filter = D3D11_FILTER_ANISOTROPIC;
desc.MaxAnisotropy = config.anisotropy;
}
else
{
static constexpr u8 filter_count = static_cast<u8>(GPUSampler::Filter::MaxCount);
static constexpr D3D11_FILTER filters[filter_count][filter_count][filter_count] = {
{
@ -1227,6 +1033,21 @@ std::unique_ptr<GPUSampler> D3D11Device::CreateSampler(const GPUSampler::Config&
{D3D11_FILTER_MIN_LINEAR_MAG_POINT_MIP_LINEAR, D3D11_FILTER_MIN_MAG_MIP_LINEAR},
}};
D3D11_SAMPLER_DESC desc = {};
desc.AddressU = ta[static_cast<u8>(config.address_u.GetValue())];
desc.AddressV = ta[static_cast<u8>(config.address_v.GetValue())];
desc.AddressW = ta[static_cast<u8>(config.address_w.GetValue())];
std::memcpy(desc.BorderColor, RGBA8ToFloat(config.border_color).data(), sizeof(desc.BorderColor));
desc.MinLOD = static_cast<float>(config.min_lod);
desc.MaxLOD = static_cast<float>(config.max_lod);
if (config.anisotropy > 0)
{
desc.Filter = D3D11_FILTER_ANISOTROPIC;
desc.MaxAnisotropy = config.anisotropy;
}
else
{
desc.Filter = filters[static_cast<u8>(config.mip_filter.GetValue())][static_cast<u8>(config.min_filter.GetValue())]
[static_cast<u8>(config.mag_filter.GetValue())];
desc.MaxAnisotropy = 1;
@ -1313,83 +1134,16 @@ std::unique_ptr<GPUShader> D3D11Device::CreateShaderFromBinary(GPUShaderStage st
std::unique_ptr<GPUShader> D3D11Device::CreateShaderFromSource(GPUShaderStage stage, const std::string_view& source,
std::vector<u8>* out_binary /* = nullptr */)
{
const char* target;
switch (m_device->GetFeatureLevel())
{
case D3D_FEATURE_LEVEL_10_0:
{
static constexpr std::array<const char*, 4> targets = {{"vs_4_0", "ps_4_0", "cs_4_0"}};
target = targets[static_cast<int>(stage)];
}
break;
case D3D_FEATURE_LEVEL_10_1:
{
static constexpr std::array<const char*, 4> targets = {{"vs_4_1", "ps_4_1", "cs_4_1"}};
target = targets[static_cast<int>(stage)];
}
break;
case D3D_FEATURE_LEVEL_11_0:
{
static constexpr std::array<const char*, 4> targets = {{"vs_5_0", "ps_5_0", "cs_5_0"}};
target = targets[static_cast<int>(stage)];
}
break;
case D3D_FEATURE_LEVEL_11_1:
default:
{
static constexpr std::array<const char*, 4> targets = {{"vs_5_1", "ps_5_1", "cs_5_1"}};
target = targets[static_cast<int>(stage)];
}
break;
}
static constexpr UINT flags_non_debug = D3DCOMPILE_OPTIMIZATION_LEVEL3;
static constexpr UINT flags_debug = D3DCOMPILE_SKIP_OPTIMIZATION | D3DCOMPILE_DEBUG;
ComPtr<ID3DBlob> blob;
ComPtr<ID3DBlob> error_blob;
const HRESULT hr =
D3DCompile(source.data(), source.size(), "0", nullptr, nullptr, "main", target,
m_debug_device ? flags_debug : flags_non_debug, 0, blob.GetAddressOf(), error_blob.GetAddressOf());
std::string error_string;
if (error_blob)
{
error_string.append(static_cast<const char*>(error_blob->GetBufferPointer()), error_blob->GetBufferSize());
error_blob.Reset();
}
if (FAILED(hr))
{
Log_ErrorPrintf("Failed to compile '%s':\n%s", target, error_string.c_str());
auto fp = FileSystem::OpenManagedCFile(
GetShaderDumpPath(fmt::format("bad_shader_{}.txt", s_next_bad_shader_id++)).c_str(), "wb");
if (fp)
{
std::fwrite(source.data(), source.size(), 1, fp.get());
std::fprintf(fp.get(), "\n\nCompile as %s failed: %08X\n", target, hr);
std::fwrite(error_string.c_str(), error_string.size(), 1, fp.get());
}
std::optional<std::vector<u8>> bytecode =
D3DCommon::CompileShader(m_device->GetFeatureLevel(), m_debug_device, stage, source);
if (!bytecode.has_value())
return {};
}
if (!error_string.empty())
Log_WarningPrintf("'%s' compiled with warnings:\n%s", target, error_string.c_str());
std::unique_ptr<GPUShader> ret = CreateShaderFromBinary(stage, bytecode.value());
if (ret && out_binary)
*out_binary = std::move(bytecode.value());
if (out_binary)
{
const size_t size = blob->GetBufferSize();
out_binary->resize(size);
std::memcpy(out_binary->data(), blob->GetBufferPointer(), size);
}
return CreateShaderFromBinary(
stage, gsl::span<const u8>(static_cast<const u8*>(blob->GetBufferPointer()), blob->GetBufferSize()));
return ret;
}
D3D11Pipeline::D3D11Pipeline(ComPtr<ID3D11RasterizerState> rs, ComPtr<ID3D11DepthStencilState> ds,

7
src/core/gpu/d3d11_device.h
View File

@ -268,11 +268,6 @@ public:
ALWAYS_INLINE static ID3D11Device* GetD3DDevice() { return GetInstance().m_device.Get(); }
ALWAYS_INLINE static ID3D11DeviceContext1* GetD3DContext() { return GetInstance().m_context.Get(); }
// returns the fullscreen mode to use for the specified dimensions
static bool GetRequestedExclusiveFullscreenModeDesc(IDXGIFactory5* factory, const RECT& window_rect, u32 width,
u32 height, float refresh_rate, DXGI_FORMAT format,
DXGI_MODE_DESC* fullscreen_mode, IDXGIOutput** output);
RenderAPI GetRenderAPI() const override;
bool HasSurface() const override;
@ -364,7 +359,7 @@ private:
static constexpr u32 UNIFORM_BUFFER_ALIGNMENT = 256;
static constexpr u8 NUM_TIMESTAMP_QUERIES = 3;
static AdapterAndModeList GetAdapterAndModeList(IDXGIFactory* dxgi_factory);
static void GetAdapterAndModeList(AdapterAndModeList* ret, IDXGIFactory5* factory);
void SetFeatures();

556
src/core/gpu/d3d12/context.cpp
View File

@ -1,556 +0,0 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
// Parts originally from Dolphin Emulator, also written by myself.
#include "context.h"
#include "common/assert.h"
#include "common/log.h"
#include "common/scoped_guard.h"
#include <algorithm>
#include <array>
#include <dxgi1_2.h>
#include <queue>
#include <vector>
Log_SetChannel(D3D12::Context);
std::unique_ptr<D3D12::Context> g_d3d12_context;
namespace D3D12 {
#if WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP)
// Private D3D12 state
static HMODULE s_d3d12_library;
static PFN_D3D12_CREATE_DEVICE s_d3d12_create_device;
static PFN_D3D12_GET_DEBUG_INTERFACE s_d3d12_get_debug_interface;
static PFN_D3D12_SERIALIZE_ROOT_SIGNATURE s_d3d12_serialize_root_signature;
static bool LoadD3D12Library()
{
if ((s_d3d12_library = LoadLibrary("d3d12.dll")) == nullptr ||
(s_d3d12_create_device =
reinterpret_cast<PFN_D3D12_CREATE_DEVICE>(GetProcAddress(s_d3d12_library, "D3D12CreateDevice"))) == nullptr ||
(s_d3d12_get_debug_interface = reinterpret_cast<PFN_D3D12_GET_DEBUG_INTERFACE>(
GetProcAddress(s_d3d12_library, "D3D12GetDebugInterface"))) == nullptr ||
(s_d3d12_serialize_root_signature = reinterpret_cast<PFN_D3D12_SERIALIZE_ROOT_SIGNATURE>(
GetProcAddress(s_d3d12_library, "D3D12SerializeRootSignature"))) == nullptr)
{
Log_ErrorPrintf("d3d12.dll could not be loaded.");
s_d3d12_create_device = nullptr;
s_d3d12_get_debug_interface = nullptr;
s_d3d12_serialize_root_signature = nullptr;
if (s_d3d12_library)
FreeLibrary(s_d3d12_library);
s_d3d12_library = nullptr;
return false;
}
return true;
}
static void UnloadD3D12Library()
{
s_d3d12_serialize_root_signature = nullptr;
s_d3d12_get_debug_interface = nullptr;
s_d3d12_create_device = nullptr;
if (s_d3d12_library)
{
FreeLibrary(s_d3d12_library);
s_d3d12_library = nullptr;
}
}
#else
static const PFN_D3D12_CREATE_DEVICE s_d3d12_create_device = D3D12CreateDevice;
static const PFN_D3D12_GET_DEBUG_INTERFACE s_d3d12_get_debug_interface = D3D12GetDebugInterface;
static const PFN_D3D12_SERIALIZE_ROOT_SIGNATURE s_d3d12_serialize_root_signature = D3D12SerializeRootSignature;
static bool LoadD3D12Library()
{
return true;
}
static void UnloadD3D12Library() {}
#endif
Context::Context() = default;
Context::~Context()
{
DestroyResources();
}
Context::ComPtr<ID3DBlob> Context::SerializeRootSignature(const D3D12_ROOT_SIGNATURE_DESC* desc)
{
ComPtr<ID3DBlob> blob;
ComPtr<ID3DBlob> error_blob;
const HRESULT hr = s_d3d12_serialize_root_signature(desc, D3D_ROOT_SIGNATURE_VERSION_1, blob.GetAddressOf(),
error_blob.GetAddressOf());
if (FAILED(hr))
{
Log_ErrorPrintf("D3D12SerializeRootSignature() failed: %08X", hr);
if (error_blob)
Log_ErrorPrintf("%s", error_blob->GetBufferPointer());
return {};
}
return blob;
}
D3D12::Context::ComPtr<ID3D12RootSignature> Context::CreateRootSignature(const D3D12_ROOT_SIGNATURE_DESC* desc)
{
ComPtr<ID3DBlob> blob = SerializeRootSignature(desc);
if (!blob)
return {};
ComPtr<ID3D12RootSignature> rs;
const HRESULT hr =
m_device->CreateRootSignature(0, blob->GetBufferPointer(), blob->GetBufferSize(), IID_PPV_ARGS(rs.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateRootSignature() failed: %08X", hr);
return {};
}
return rs;
}
bool Context::SupportsTextureFormat(DXGI_FORMAT format)
{
constexpr u32 required = D3D12_FORMAT_SUPPORT1_TEXTURE2D | D3D12_FORMAT_SUPPORT1_SHADER_SAMPLE;
D3D12_FEATURE_DATA_FORMAT_SUPPORT support = {format};
return SUCCEEDED(m_device->CheckFeatureSupport(D3D12_FEATURE_FORMAT_SUPPORT, &support, sizeof(support))) &&
(support.Support1 & required) == required;
}
bool Context::Create(IDXGIFactory* dxgi_factory, u32 adapter_index, bool enable_debug_layer)
{
Assert(!g_d3d12_context);
if (!LoadD3D12Library())
return false;
g_d3d12_context.reset(new Context());
if (!g_d3d12_context->CreateDevice(dxgi_factory, adapter_index, enable_debug_layer) ||
!g_d3d12_context->CreateCommandQueue() || !g_d3d12_context->CreateFence() ||
!g_d3d12_context->CreateDescriptorHeaps() || !g_d3d12_context->CreateCommandLists() ||
!g_d3d12_context->CreateTimestampQuery() || !g_d3d12_context->CreateTextureStreamBuffer())
{
Destroy();
return false;
}
return true;
}
void Context::Destroy()
{
if (g_d3d12_context)
g_d3d12_context.reset();
UnloadD3D12Library();
}
bool Context::CreateDevice(IDXGIFactory* dxgi_factory, u32 adapter_index, bool enable_debug_layer)
{
ComPtr<IDXGIAdapter> adapter;
HRESULT hr = dxgi_factory->EnumAdapters(adapter_index, &adapter);
if (FAILED(hr))
{
Log_ErrorPrintf("Adapter %u not found, using default", adapter_index);
adapter = nullptr;
}
else
{
DXGI_ADAPTER_DESC adapter_desc;
if (SUCCEEDED(adapter->GetDesc(&adapter_desc)))
{
char adapter_name_buffer[128];
const int name_length = WideCharToMultiByte(CP_UTF8, 0, adapter_desc.Description,
static_cast<int>(std::wcslen(adapter_desc.Description)),
adapter_name_buffer, countof(adapter_name_buffer), 0, nullptr);
if (name_length >= 0)
{
adapter_name_buffer[name_length] = 0;
Log_InfoPrintf("D3D Adapter: %s", adapter_name_buffer);
}
}
}
// Enabling the debug layer will fail if the Graphics Tools feature is not installed.
if (enable_debug_layer)
{
hr = s_d3d12_get_debug_interface(IID_PPV_ARGS(&m_debug_interface));
if (SUCCEEDED(hr))
{
m_debug_interface->EnableDebugLayer();
}
else
{
Log_ErrorPrintf("Debug layer requested but not available.");
enable_debug_layer = false;
}
}
// Create the actual device.
hr = s_d3d12_create_device(adapter.Get(), D3D_FEATURE_LEVEL_11_0, IID_PPV_ARGS(&m_device));
AssertMsg(SUCCEEDED(hr), "Create D3D12 device");
if (FAILED(hr))
return false;
if (enable_debug_layer)
{
ComPtr<ID3D12InfoQueue> info_queue;
if (SUCCEEDED(m_device.As(&info_queue)))
{
info_queue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_ERROR, TRUE);
info_queue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_WARNING, TRUE);
D3D12_INFO_QUEUE_FILTER filter = {};
std::array<D3D12_MESSAGE_ID, 5> id_list{
D3D12_MESSAGE_ID_CLEARRENDERTARGETVIEW_MISMATCHINGCLEARVALUE,
D3D12_MESSAGE_ID_CLEARDEPTHSTENCILVIEW_MISMATCHINGCLEARVALUE,
D3D12_MESSAGE_ID_CREATEGRAPHICSPIPELINESTATE_RENDERTARGETVIEW_NOT_SET,
D3D12_MESSAGE_ID_CREATEINPUTLAYOUT_TYPE_MISMATCH,
D3D12_MESSAGE_ID_DRAW_EMPTY_SCISSOR_RECTANGLE,
};
filter.DenyList.NumIDs = static_cast<UINT>(id_list.size());
filter.DenyList.pIDList = id_list.data();
info_queue->PushStorageFilter(&filter);
}
}
return true;
}
bool Context::CreateCommandQueue()
{
const D3D12_COMMAND_QUEUE_DESC queue_desc = {D3D12_COMMAND_LIST_TYPE_DIRECT, D3D12_COMMAND_QUEUE_PRIORITY_NORMAL,
D3D12_COMMAND_QUEUE_FLAG_NONE};
HRESULT hr = m_device->CreateCommandQueue(&queue_desc, IID_PPV_ARGS(&m_command_queue));
AssertMsg(SUCCEEDED(hr), "Create command queue");
return SUCCEEDED(hr);
}
bool Context::CreateFence()
{
HRESULT hr = m_device->CreateFence(m_completed_fence_value, D3D12_FENCE_FLAG_NONE, IID_PPV_ARGS(&m_fence));
AssertMsg(SUCCEEDED(hr), "Create fence");
if (FAILED(hr))
return false;
m_fence_event = CreateEvent(nullptr, FALSE, FALSE, nullptr);
AssertMsg(m_fence_event != NULL, "Create fence event");
if (!m_fence_event)
return false;
return true;
}
bool Context::CreateDescriptorHeaps()
{
static constexpr size_t MAX_SRVS = 16384;
static constexpr size_t MAX_RTVS = 8192;
static constexpr size_t MAX_DSVS = 128;
static constexpr size_t MAX_SAMPLERS = 128;
if (!m_descriptor_heap_manager.Create(m_device.Get(), D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, MAX_SRVS, true) ||
!m_rtv_heap_manager.Create(m_device.Get(), D3D12_DESCRIPTOR_HEAP_TYPE_RTV, MAX_RTVS, false) ||
!m_dsv_heap_manager.Create(m_device.Get(), D3D12_DESCRIPTOR_HEAP_TYPE_DSV, MAX_DSVS, false) ||
!m_sampler_heap_manager.Create(m_device.Get(), D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER, MAX_SAMPLERS, true))
{
return false;
}
m_gpu_descriptor_heaps[0] = m_descriptor_heap_manager.GetDescriptorHeap();
m_gpu_descriptor_heaps[1] = m_sampler_heap_manager.GetDescriptorHeap();
// Allocate null SRV descriptor for unbound textures.
constexpr D3D12_SHADER_RESOURCE_VIEW_DESC null_srv_desc = {DXGI_FORMAT_R8G8B8A8_UNORM, D3D12_SRV_DIMENSION_TEXTURE2D,
D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING};
if (!m_descriptor_heap_manager.Allocate(&m_null_srv_descriptor))
{
Panic("Failed to allocate null descriptor");
return false;
}
m_device->CreateShaderResourceView(nullptr, &null_srv_desc, m_null_srv_descriptor.cpu_handle);
return true;
}
bool Context::CreateCommandLists()
{
for (u32 i = 0; i < NUM_COMMAND_LISTS; i++)
{
CommandListResources& res = m_command_lists[i];
HRESULT hr = m_device->CreateCommandAllocator(D3D12_COMMAND_LIST_TYPE_DIRECT,
IID_PPV_ARGS(res.command_allocator.GetAddressOf()));
AssertMsg(SUCCEEDED(hr), "Create command allocator");
if (FAILED(hr))
return false;
hr = m_device->CreateCommandList(1, D3D12_COMMAND_LIST_TYPE_DIRECT, res.command_allocator.Get(), nullptr,
IID_PPV_ARGS(res.command_list.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("Failed to create command list: %08X", hr);
return false;
}
// Close the command list, since the first thing we do is reset them.
hr = res.command_list->Close();
AssertMsg(SUCCEEDED(hr), "Closing new command list failed");
if (FAILED(hr))
return false;
}
MoveToNextCommandList();
return true;
}
bool Context::CreateTextureStreamBuffer()
{
return m_texture_stream_buffer.Create(TEXTURE_UPLOAD_BUFFER_SIZE);
}
void Context::MoveToNextCommandList()
{
m_current_command_list = (m_current_command_list + 1) % NUM_COMMAND_LISTS;
m_current_fence_value++;
// We may have to wait if this command list hasn't finished on the GPU.
CommandListResources& res = m_command_lists[m_current_command_list];
WaitForFence(res.ready_fence_value);
res.ready_fence_value = m_current_fence_value;
// Begin command list.
res.command_allocator->Reset();
res.command_list->Reset(res.command_allocator.Get(), nullptr);
if (res.has_timestamp_query)
{
// readback timestamp from the last time this cmdlist was used.
// we don't need to worry about disjoint in dx12, the frequency is reliable within a single cmdlist.
const u32 offset = (m_current_command_list * (sizeof(u64) * NUM_TIMESTAMP_QUERIES_PER_CMDLIST));
const D3D12_RANGE read_range = {offset, offset + (sizeof(u64) * NUM_TIMESTAMP_QUERIES_PER_CMDLIST)};
void* map;
HRESULT hr = m_timestamp_query_buffer->Map(0, &read_range, &map);
if (SUCCEEDED(hr))
{
u64 timestamps[2];
std::memcpy(timestamps, static_cast<const u8*>(map) + offset, sizeof(timestamps));
m_accumulated_gpu_time +=
static_cast<float>(static_cast<double>(timestamps[1] - timestamps[0]) / m_timestamp_frequency);
const D3D12_RANGE write_range = {};
m_timestamp_query_buffer->Unmap(0, &write_range);
}
else
{
Log_WarningPrintf("Map() for timestamp query failed: %08X", hr);
}
}
res.has_timestamp_query = m_gpu_timing_enabled;
if (m_gpu_timing_enabled)
{
res.command_list->EndQuery(m_timestamp_query_heap.Get(), D3D12_QUERY_TYPE_TIMESTAMP,
m_current_command_list * NUM_TIMESTAMP_QUERIES_PER_CMDLIST);
}
res.command_list->SetDescriptorHeaps(static_cast<UINT>(m_gpu_descriptor_heaps.size()), m_gpu_descriptor_heaps.data());
}
void Context::ExecuteCommandList(bool wait_for_completion)
{
CommandListResources& res = m_command_lists[m_current_command_list];
HRESULT hr;
if (res.has_timestamp_query)
{
// write the timestamp back at the end of the cmdlist
res.command_list->EndQuery(m_timestamp_query_heap.Get(), D3D12_QUERY_TYPE_TIMESTAMP,
(m_current_command_list * NUM_TIMESTAMP_QUERIES_PER_CMDLIST) + 1);
res.command_list->ResolveQueryData(m_timestamp_query_heap.Get(), D3D12_QUERY_TYPE_TIMESTAMP,
m_current_command_list * NUM_TIMESTAMP_QUERIES_PER_CMDLIST,
NUM_TIMESTAMP_QUERIES_PER_CMDLIST, m_timestamp_query_buffer.Get(),
m_current_command_list * (sizeof(u64) * NUM_TIMESTAMP_QUERIES_PER_CMDLIST));
}
// Close and queue command list.
hr = res.command_list->Close();
AssertMsg(SUCCEEDED(hr), "Close command list");
const std::array<ID3D12CommandList*, 1> execute_lists{res.command_list.Get()};
m_command_queue->ExecuteCommandLists(static_cast<UINT>(execute_lists.size()), execute_lists.data());
// Update fence when GPU has completed.
hr = m_command_queue->Signal(m_fence.Get(), m_current_fence_value);
AssertMsg(SUCCEEDED(hr), "Signal fence");
MoveToNextCommandList();
if (wait_for_completion)
WaitForFence(res.ready_fence_value);
}
void Context::DeferResourceDestruction(ID3D12Resource* resource)
{
if (!resource)
return;
resource->AddRef();
m_command_lists[m_current_command_list].pending_resources.push_back(resource);
}
void Context::DeferDescriptorDestruction(DescriptorHeapManager& manager, u32 index)
{
m_command_lists[m_current_command_list].pending_descriptors.emplace_back(manager, index);
}
void Context::DeferDescriptorDestruction(DescriptorHeapManager& manager, DescriptorHandle* handle)
{
if (handle->index == DescriptorHandle::INVALID_INDEX)
return;
m_command_lists[m_current_command_list].pending_descriptors.emplace_back(manager, handle->index);
handle->Clear();
}
void Context::DestroyPendingResources(CommandListResources& cmdlist)
{
for (const auto& dd : cmdlist.pending_descriptors)
dd.first.Free(dd.second);
cmdlist.pending_descriptors.clear();
for (ID3D12Resource* res : cmdlist.pending_resources)
res->Release();
cmdlist.pending_resources.clear();
}
void Context::DestroyResources()
{
ExecuteCommandList(true);
m_timestamp_query_buffer.Reset();
m_timestamp_query_heap.Reset();
m_texture_stream_buffer.Destroy(false);
m_descriptor_heap_manager.Free(&m_null_srv_descriptor);
m_sampler_heap_manager.Destroy();
m_dsv_heap_manager.Destroy();
m_rtv_heap_manager.Destroy();
m_descriptor_heap_manager.Destroy();
m_command_lists = {};
m_current_command_list = 0;
m_completed_fence_value = 0;
m_current_fence_value = 0;
if (m_fence_event)
{
CloseHandle(m_fence_event);
m_fence_event = {};
}
m_command_queue.Reset();
m_debug_interface.Reset();
m_device.Reset();
}
void Context::WaitForFence(u64 fence)
{
if (m_completed_fence_value >= fence)
return;
// Try non-blocking check.
m_completed_fence_value = m_fence->GetCompletedValue();
if (m_completed_fence_value < fence)
{
// Fall back to event.
HRESULT hr = m_fence->SetEventOnCompletion(fence, m_fence_event);
AssertMsg(SUCCEEDED(hr), "Set fence event on completion");
WaitForSingleObject(m_fence_event, INFINITE);
m_completed_fence_value = m_fence->GetCompletedValue();
}
// Release resources for as many command lists which have completed.
u32 index = (m_current_command_list + 1) % NUM_COMMAND_LISTS;
for (u32 i = 0; i < NUM_COMMAND_LISTS; i++)
{
CommandListResources& res = m_command_lists[index];
if (m_completed_fence_value < res.ready_fence_value)
break;
DestroyPendingResources(res);
index = (index + 1) % NUM_COMMAND_LISTS;
}
}
void Context::WaitForGPUIdle()
{
u32 index = (m_current_command_list + 1) % NUM_COMMAND_LISTS;
for (u32 i = 0; i < (NUM_COMMAND_LISTS - 1); i++)
{
WaitForFence(m_command_lists[index].ready_fence_value);
index = (index + 1) % NUM_COMMAND_LISTS;
}
}
bool Context::CreateTimestampQuery()
{
constexpr u32 QUERY_COUNT = NUM_TIMESTAMP_QUERIES_PER_CMDLIST * NUM_COMMAND_LISTS;
constexpr u32 BUFFER_SIZE = sizeof(u64) * QUERY_COUNT;
const D3D12_QUERY_HEAP_DESC desc = {D3D12_QUERY_HEAP_TYPE_TIMESTAMP, QUERY_COUNT};
HRESULT hr = m_device->CreateQueryHeap(&desc, IID_PPV_ARGS(m_timestamp_query_heap.ReleaseAndGetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateQueryHeap() for timestamp failed with %08X", hr);
return false;
}
const D3D12_HEAP_PROPERTIES heap_properties = {D3D12_HEAP_TYPE_READBACK};
const D3D12_RESOURCE_DESC resource_desc = {D3D12_RESOURCE_DIMENSION_BUFFER,
0,
BUFFER_SIZE,
1,
1,
1,
DXGI_FORMAT_UNKNOWN,
{1, 0},
D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
D3D12_RESOURCE_FLAG_NONE};
hr = m_device->CreateCommittedResource(&heap_properties, D3D12_HEAP_FLAG_NONE, &resource_desc,
D3D12_RESOURCE_STATE_COPY_DEST, nullptr,
IID_PPV_ARGS(m_timestamp_query_buffer.ReleaseAndGetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateResource() for timestamp failed with %08X", hr);
return false;
}
u64 frequency;
hr = m_command_queue->GetTimestampFrequency(&frequency);
if (FAILED(hr))
{
Log_ErrorPrintf("GetTimestampFrequency() failed: %08X", hr);
return false;
}
m_timestamp_frequency = static_cast<double>(frequency) / 1000.0;
return true;
}
float Context::GetAndResetAccumulatedGPUTime()
{
const float time = m_accumulated_gpu_time;
m_accumulated_gpu_time = 0.0f;
return time;
}
void Context::SetEnableGPUTiming(bool enabled)
{
m_gpu_timing_enabled = enabled;
}
} // namespace D3D12

156
src/core/gpu/d3d12/context.h
View File

@ -1,156 +0,0 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
// Parts originally from Dolphin Emulator, also written by myself.
#pragma once
#include "common/types.h"
#include "common/windows_headers.h"
#include "descriptor_heap_manager.h"
#include "stream_buffer.h"
#include <array>
#include <d3d12.h>
#include <memory>
#include <vector>
#include <wrl/client.h>
struct IDXGIFactory;
namespace D3D12 {
class Context
{
public:
template<typename T>
using ComPtr = Microsoft::WRL::ComPtr<T>;
enum : u32
{
// Number of command lists. One is being built while the other(s) are executed.
NUM_COMMAND_LISTS = 3,
// Textures that don't fit into this buffer will be uploaded with a staging buffer.
TEXTURE_UPLOAD_BUFFER_SIZE = 16 * 1024 * 1024,
/// Start/End timestamp queries.
NUM_TIMESTAMP_QUERIES_PER_CMDLIST = 2,
};
~Context();
// Creates new device and context.
static bool Create(IDXGIFactory* dxgi_factory, u32 adapter_index, bool enable_debug_layer);
// Destroys active context.
static void Destroy();
ID3D12Device* GetDevice() const { return m_device.Get(); }
ID3D12CommandQueue* GetCommandQueue() const { return m_command_queue.Get(); }
// Returns the current command list, commands can be recorded directly.
ID3D12GraphicsCommandList* GetCommandList() const
{
return m_command_lists[m_current_command_list].command_list.Get();
}
// Descriptor manager access.
DescriptorHeapManager& GetDescriptorHeapManager() { return m_descriptor_heap_manager; }
DescriptorHeapManager& GetRTVHeapManager() { return m_rtv_heap_manager; }
DescriptorHeapManager& GetDSVHeapManager() { return m_dsv_heap_manager; }
DescriptorHeapManager& GetSamplerHeapManager() { return m_sampler_heap_manager; }
ID3D12DescriptorHeap* const* GetGPUDescriptorHeaps() const { return m_gpu_descriptor_heaps.data(); }
u32 GetGPUDescriptorHeapCount() const { return static_cast<u32>(m_gpu_descriptor_heaps.size()); }
const DescriptorHandle& GetNullSRVDescriptor() const { return m_null_srv_descriptor; }
StreamBuffer& GetTextureStreamBuffer() { return m_texture_stream_buffer; }
// Root signature access.
ComPtr<ID3DBlob> SerializeRootSignature(const D3D12_ROOT_SIGNATURE_DESC* desc);
ComPtr<ID3D12RootSignature> CreateRootSignature(const D3D12_ROOT_SIGNATURE_DESC* desc);
// Fence value for current command list.
u64 GetCurrentFenceValue() const { return m_current_fence_value; }
// Last "completed" fence.
u64 GetCompletedFenceValue() const { return m_completed_fence_value; }
// Feature level to use when compiling shaders.
D3D_FEATURE_LEVEL GetFeatureLevel() const { return m_feature_level; }
// Test for support for the specified texture format.
bool SupportsTextureFormat(DXGI_FORMAT format);
// Executes the current command list.
void ExecuteCommandList(bool wait_for_completion);
// Waits for a specific fence.
void WaitForFence(u64 fence);
// Waits for any in-flight command buffers to complete.
void WaitForGPUIdle();
// Defers destruction of a D3D resource (associates it with the current list).
void DeferResourceDestruction(ID3D12Resource* resource);
// Defers destruction of a descriptor handle (associates it with the current list).
void DeferDescriptorDestruction(DescriptorHeapManager& manager, u32 index);
void DeferDescriptorDestruction(DescriptorHeapManager& manager, DescriptorHandle* handle);
float GetAndResetAccumulatedGPUTime();
void SetEnableGPUTiming(bool enabled);
private:
struct CommandListResources
{
ComPtr<ID3D12CommandAllocator> command_allocator;
ComPtr<ID3D12GraphicsCommandList> command_list;
std::vector<ID3D12Resource*> pending_resources;
std::vector<std::pair<DescriptorHeapManager&, u32>> pending_descriptors;
u64 ready_fence_value = 0;
bool has_timestamp_query = false;
};
Context();
bool CreateDevice(IDXGIFactory* dxgi_factory, u32 adapter_index, bool enable_debug_layer);
bool CreateCommandQueue();
bool CreateFence();
bool CreateDescriptorHeaps();
bool CreateCommandLists();
bool CreateTextureStreamBuffer();
bool CreateTimestampQuery();
void MoveToNextCommandList();
void DestroyPendingResources(CommandListResources& cmdlist);
void DestroyResources();
ComPtr<ID3D12Debug> m_debug_interface;
ComPtr<ID3D12Device> m_device;
ComPtr<ID3D12CommandQueue> m_command_queue;
ComPtr<ID3D12Fence> m_fence = nullptr;
HANDLE m_fence_event = {};
u32 m_current_fence_value = 0;
u64 m_completed_fence_value = 0;
std::array<CommandListResources, NUM_COMMAND_LISTS> m_command_lists;
u32 m_current_command_list = NUM_COMMAND_LISTS - 1;
ComPtr<ID3D12QueryHeap> m_timestamp_query_heap;
ComPtr<ID3D12Resource> m_timestamp_query_buffer;
double m_timestamp_frequency = 0.0;
float m_accumulated_gpu_time = 0.0f;
bool m_gpu_timing_enabled = false;
DescriptorHeapManager m_descriptor_heap_manager;
DescriptorHeapManager m_rtv_heap_manager;
DescriptorHeapManager m_dsv_heap_manager;
DescriptorHeapManager m_sampler_heap_manager;
std::array<ID3D12DescriptorHeap*, 2> m_gpu_descriptor_heaps = {};
DescriptorHandle m_null_srv_descriptor;
StreamBuffer m_texture_stream_buffer;
D3D_FEATURE_LEVEL m_feature_level = D3D_FEATURE_LEVEL_11_0;
};
} // namespace D3D12
extern std::unique_ptr<D3D12::Context> g_d3d12_context;

116
src/core/gpu/d3d12/descriptor_heap_manager.cpp
View File

@ -1,116 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
// Parts originally from Dolphin Emulator, also written by myself.
#include "descriptor_heap_manager.h"
#include "common/assert.h"
#include "common/log.h"
#include "context.h"
Log_SetChannel(DescriptorHeapManager);
namespace D3D12 {
DescriptorHeapManager::DescriptorHeapManager() = default;
DescriptorHeapManager::~DescriptorHeapManager() = default;
bool DescriptorHeapManager::Create(ID3D12Device* device, D3D12_DESCRIPTOR_HEAP_TYPE type, u32 num_descriptors,
bool shader_visible)
{
D3D12_DESCRIPTOR_HEAP_DESC desc = {type, static_cast<UINT>(num_descriptors),
shader_visible ? D3D12_DESCRIPTOR_HEAP_FLAG_SHADER_VISIBLE :
D3D12_DESCRIPTOR_HEAP_FLAG_NONE};
HRESULT hr = device->CreateDescriptorHeap(&desc, IID_PPV_ARGS(&m_descriptor_heap));
AssertMsg(SUCCEEDED(hr), "Create descriptor heap");
if (FAILED(hr))
return false;
m_heap_base_cpu = m_descriptor_heap->GetCPUDescriptorHandleForHeapStart();
if (shader_visible)
m_heap_base_gpu = m_descriptor_heap->GetGPUDescriptorHandleForHeapStart();
m_num_descriptors = num_descriptors;
m_descriptor_increment_size = device->GetDescriptorHandleIncrementSize(type);
// Set all slots to unallocated (1)
const u32 bitset_count = num_descriptors / BITSET_SIZE + (((num_descriptors % BITSET_SIZE) != 0) ? 1 : 0);
m_free_slots.resize(bitset_count);
for (BitSetType& bs : m_free_slots)
bs.flip();
return true;
}
void DescriptorHeapManager::Destroy()
{
for (BitSetType& bs : m_free_slots)
Assert(bs.all());
m_num_descriptors = 0;
m_descriptor_increment_size = 0;
m_heap_base_cpu = {};
m_heap_base_gpu = {};
m_descriptor_heap.Reset();
m_free_slots.clear();
}
bool DescriptorHeapManager::Allocate(DescriptorHandle* handle, u32 count /* = 1 */)
{
// Start past the temporary slots, no point in searching those.
for (u32 group = 0; group < m_free_slots.size(); group++)
{
BitSetType& bs = m_free_slots[group];
if (bs.none())
continue;
u32 bit = 0;
for (; bit < BITSET_SIZE; bit++)
{
if (bs[bit])
{
u32 offset;
for (offset = 0; offset < count; offset++)
{
if (!bs[bit + offset])
break;
}
if (offset == count)
break;
}
}
u32 index = group * BITSET_SIZE + bit;
for (u32 offset = 0; offset < count; offset++)
bs[bit + offset] = false;
handle->index = index;
handle->cpu_handle.ptr = m_heap_base_cpu.ptr + index * m_descriptor_increment_size;
handle->gpu_handle.ptr = m_heap_base_gpu.ptr + index * m_descriptor_increment_size;
return true;
}
Panic("Out of fixed descriptors");
return false;
}
void DescriptorHeapManager::Free(u32 index, u32 count /* = 1 */)
{
Assert(index < m_num_descriptors);
for (u32 i = 0; i < count; i++, index++)
{
u32 group = index / BITSET_SIZE;
u32 bit = index % BITSET_SIZE;
m_free_slots[group][bit] = true;
}
}
void DescriptorHeapManager::Free(DescriptorHandle* handle, u32 count /* = 1 */)
{
if (handle->index == DescriptorHandle::INVALID_INDEX)
return;
Free(handle->index, count);
handle->Clear();
}
} // namespace D3D12

84
src/core/gpu/d3d12/descriptor_heap_manager.h
View File

@ -1,84 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
// Parts originally from Dolphin Emulator, also written by myself.
#pragma once
#include "common/types.h"
#include "common/windows_headers.h"
#include <bitset>
#include <d3d12.h>
#include <map>
#include <vector>
#include <wrl/client.h>
namespace D3D12 {
// This class provides an abstraction for D3D12 descriptor heaps.
struct DescriptorHandle final
{
enum : u32
{
INVALID_INDEX = 0xFFFFFFFF
};
D3D12_CPU_DESCRIPTOR_HANDLE cpu_handle{};
D3D12_GPU_DESCRIPTOR_HANDLE gpu_handle{};
u32 index = INVALID_INDEX;
ALWAYS_INLINE operator bool() const { return index != INVALID_INDEX; }
ALWAYS_INLINE operator D3D12_CPU_DESCRIPTOR_HANDLE() const { return cpu_handle; }
ALWAYS_INLINE operator D3D12_GPU_DESCRIPTOR_HANDLE() const { return gpu_handle; }
ALWAYS_INLINE void Clear()
{
cpu_handle = {};
gpu_handle = {};
index = INVALID_INDEX;
}
};
class DescriptorHeapManager final
{
public:
DescriptorHeapManager();
~DescriptorHeapManager();
ALWAYS_INLINE ID3D12DescriptorHeap* GetDescriptorHeap() const { return m_descriptor_heap.Get(); }
ALWAYS_INLINE u32 GetDescriptorIncrementSize() const { return m_descriptor_increment_size; }
ALWAYS_INLINE D3D12_CPU_DESCRIPTOR_HANDLE OffsetCPUHandle(D3D12_CPU_DESCRIPTOR_HANDLE handle, u32 count) const
{
D3D12_CPU_DESCRIPTOR_HANDLE ret;
ret.ptr = handle.ptr + m_descriptor_increment_size * count;
return ret;
}
ALWAYS_INLINE D3D12_GPU_DESCRIPTOR_HANDLE OffsetGPUHandle(D3D12_GPU_DESCRIPTOR_HANDLE handle, u32 count) const
{
D3D12_GPU_DESCRIPTOR_HANDLE ret;
ret.ptr = handle.ptr + m_descriptor_increment_size * count;
return ret;
}
bool Create(ID3D12Device* device, D3D12_DESCRIPTOR_HEAP_TYPE type, u32 num_descriptors, bool shader_visible);
void Destroy();
bool Allocate(DescriptorHandle* handle, u32 count = 1);
void Free(DescriptorHandle* handle, u32 count = 1);
void Free(u32 index, u32 count = 1);
private:
Microsoft::WRL::ComPtr<ID3D12DescriptorHeap> m_descriptor_heap;
u32 m_num_descriptors = 0;
u32 m_descriptor_increment_size = 0;
D3D12_CPU_DESCRIPTOR_HANDLE m_heap_base_cpu = {};
D3D12_GPU_DESCRIPTOR_HANDLE m_heap_base_gpu = {};
static constexpr u32 BITSET_SIZE = 1024;
using BitSetType = std::bitset<BITSET_SIZE>;
std::vector<BitSetType> m_free_slots = {};
};
} // namespace D3D12

520
src/core/gpu/d3d12/shader_cache.cpp
View File

@ -1,520 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "shader_cache.h"
#include "common/file_system.h"
#include "common/log.h"
#include "common/md5_digest.h"
#include "common/string_util.h"
#include <array>
#include <d3dcompiler.h>
#include <fstream>
Log_SetChannel(D3D12::ShaderCache);
namespace D3D12 {
#pragma pack(push, 1)
struct CacheIndexEntry
{
u64 source_hash_low;
u64 source_hash_high;
u32 source_length;
u32 shader_type;
u32 file_offset;
u32 blob_size;
};
#pragma pack(pop)
ShaderCache::ShaderCache() = default;
ShaderCache::~ShaderCache()
{
if (m_pipeline_index_file)
std::fclose(m_pipeline_index_file);
if (m_pipeline_blob_file)
std::fclose(m_pipeline_blob_file);
if (m_shader_index_file)
std::fclose(m_shader_index_file);
if (m_shader_blob_file)
std::fclose(m_shader_blob_file);
}
bool ShaderCache::CacheIndexKey::operator==(const CacheIndexKey& key) const
{
return (source_hash_low == key.source_hash_low && source_hash_high == key.source_hash_high &&
source_length == key.source_length && type == key.type);
}
bool ShaderCache::CacheIndexKey::operator!=(const CacheIndexKey& key) const
{
return (source_hash_low != key.source_hash_low || source_hash_high != key.source_hash_high ||
source_length != key.source_length || type != key.type);
}
void ShaderCache::Open(std::string_view base_path, D3D_FEATURE_LEVEL feature_level, bool debug)
{
m_base_path = base_path;
m_feature_level = feature_level;
m_debug = debug;
if (!base_path.empty())
{
const std::string base_shader_filename = GetCacheBaseFileName(base_path, "shaders", feature_level, debug);
const std::string shader_index_filename = base_shader_filename + ".idx";
const std::string shader_blob_filename = base_shader_filename + ".bin";
if (!ReadExisting(shader_index_filename, shader_blob_filename, m_shader_index_file, m_shader_blob_file,
m_shader_index))
{
CreateNew(shader_index_filename, shader_blob_filename, m_shader_index_file, m_shader_blob_file);
}
const std::string base_pipelines_filename = GetCacheBaseFileName(base_path, "pipelines", feature_level, debug);
const std::string pipelines_index_filename = base_pipelines_filename + ".idx";
const std::string pipelines_blob_filename = base_pipelines_filename + ".bin";
if (!ReadExisting(pipelines_index_filename, pipelines_blob_filename, m_pipeline_index_file, m_pipeline_blob_file,
m_pipeline_index))
{
CreateNew(pipelines_index_filename, pipelines_blob_filename, m_pipeline_index_file, m_pipeline_blob_file);
}
}
}
void ShaderCache::InvalidatePipelineCache()
{
m_pipeline_index.clear();
if (m_pipeline_blob_file)
{
std::fclose(m_pipeline_blob_file);
m_pipeline_blob_file = nullptr;
}
if (m_pipeline_index_file)
{
std::fclose(m_pipeline_index_file);
m_pipeline_index_file = nullptr;
}
const std::string base_pipelines_filename = GetCacheBaseFileName(m_base_path, "pipelines", m_feature_level, m_debug);
const std::string pipelines_index_filename = base_pipelines_filename + ".idx";
const std::string pipelines_blob_filename = base_pipelines_filename + ".bin";
CreateNew(pipelines_index_filename, pipelines_blob_filename, m_pipeline_index_file, m_pipeline_blob_file);
}
bool ShaderCache::CreateNew(const std::string& index_filename, const std::string& blob_filename, std::FILE*& index_file,
std::FILE*& blob_file)
{
if (FileSystem::FileExists(index_filename.c_str()))
{
Log_WarningPrintf("Removing existing index file '%s'", index_filename.c_str());
FileSystem::DeleteFile(index_filename.c_str());
}
if (FileSystem::FileExists(blob_filename.c_str()))
{
Log_WarningPrintf("Removing existing blob file '%s'", blob_filename.c_str());
FileSystem::DeleteFile(blob_filename.c_str());
}
index_file = FileSystem::OpenCFile(index_filename.c_str(), "wb");
if (!index_file)
{
Log_ErrorPrintf("Failed to open index file '%s' for writing", index_filename.c_str());
return false;
}
const u32 index_version = FILE_VERSION;
if (std::fwrite(&index_version, sizeof(index_version), 1, index_file) != 1)
{
Log_ErrorPrintf("Failed to write version to index file '%s'", index_filename.c_str());
std::fclose(index_file);
index_file = nullptr;
FileSystem::DeleteFile(index_filename.c_str());
return false;
}
blob_file = FileSystem::OpenCFile(blob_filename.c_str(), "w+b");
if (!blob_file)
{
Log_ErrorPrintf("Failed to open blob file '%s' for writing", blob_filename.c_str());
std::fclose(blob_file);
blob_file = nullptr;
FileSystem::DeleteFile(index_filename.c_str());
return false;
}
return true;
}
bool ShaderCache::ReadExisting(const std::string& index_filename, const std::string& blob_filename,
std::FILE*& index_file, std::FILE*& blob_file, CacheIndex& index)
{
index_file = FileSystem::OpenCFile(index_filename.c_str(), "r+b");
if (!index_file)
return false;
u32 file_version;
if (std::fread(&file_version, sizeof(file_version), 1, index_file) != 1 || file_version != FILE_VERSION)
{
Log_ErrorPrintf("Bad file version in '%s'", index_filename.c_str());
std::fclose(index_file);
index_file = nullptr;
return false;
}
blob_file = FileSystem::OpenCFile(blob_filename.c_str(), "a+b");
if (!blob_file)
{
Log_ErrorPrintf("Blob file '%s' is missing", blob_filename.c_str());
std::fclose(index_file);
index_file = nullptr;
return false;
}
std::fseek(blob_file, 0, SEEK_END);
const u32 blob_file_size = static_cast<u32>(std::ftell(blob_file));
for (;;)
{
CacheIndexEntry entry;
if (std::fread(&entry, sizeof(entry), 1, index_file) != 1 || (entry.file_offset + entry.blob_size) > blob_file_size)
{
if (std::feof(index_file))
break;
Log_ErrorPrintf("Failed to read entry from '%s', corrupt file?", index_filename.c_str());
index.clear();
std::fclose(blob_file);
blob_file = nullptr;
std::fclose(index_file);
index_file = nullptr;
return false;
}
const CacheIndexKey key{entry.source_hash_low, entry.source_hash_high, entry.source_length,
static_cast<EntryType>(entry.shader_type)};
const CacheIndexData data{entry.file_offset, entry.blob_size};
index.emplace(key, data);
}
// ensure we don't write before seeking
std::fseek(index_file, 0, SEEK_END);
Log_InfoPrintf("Read %zu entries from '%s'", index.size(), index_filename.c_str());
return true;
}
std::string ShaderCache::GetCacheBaseFileName(const std::string_view& base_path, const std::string_view& type,
D3D_FEATURE_LEVEL feature_level, bool debug)
{
std::string base_filename(base_path);
base_filename += FS_OSPATH_SEPARATOR_STR "d3d12_";
base_filename += type;
base_filename += "_";
switch (feature_level)
{
case D3D_FEATURE_LEVEL_10_0:
base_filename += "sm40";
break;
case D3D_FEATURE_LEVEL_10_1:
base_filename += "sm41";
break;
case D3D_FEATURE_LEVEL_11_0:
base_filename += "sm50";
break;
default:
base_filename += "unk";
break;
}
if (debug)
base_filename += "_debug";
return base_filename;
}
union MD5Hash
{
struct
{
u64 low;
u64 high;
};
u8 hash[16];
};
ShaderCache::CacheIndexKey ShaderCache::GetShaderCacheKey(EntryType type, const std::string_view& shader_code)
{
MD5Hash h;
MD5Digest digest;
digest.Update(shader_code.data(), static_cast<u32>(shader_code.length()));
digest.Final(h.hash);
return CacheIndexKey{h.low, h.high, static_cast<u32>(shader_code.length()), type};
}
ShaderCache::CacheIndexKey ShaderCache::GetPipelineCacheKey(const D3D12_GRAPHICS_PIPELINE_STATE_DESC& gpdesc)
{
MD5Digest digest;
u32 length = sizeof(D3D12_GRAPHICS_PIPELINE_STATE_DESC);
if (gpdesc.VS.BytecodeLength > 0)
{
digest.Update(gpdesc.VS.pShaderBytecode, static_cast<u32>(gpdesc.VS.BytecodeLength));
length += static_cast<u32>(gpdesc.VS.BytecodeLength);
}
if (gpdesc.GS.BytecodeLength > 0)
{
digest.Update(gpdesc.GS.pShaderBytecode, static_cast<u32>(gpdesc.GS.BytecodeLength));
length += static_cast<u32>(gpdesc.GS.BytecodeLength);
}
if (gpdesc.PS.BytecodeLength > 0)
{
digest.Update(gpdesc.PS.pShaderBytecode, static_cast<u32>(gpdesc.PS.BytecodeLength));
length += static_cast<u32>(gpdesc.PS.BytecodeLength);
}
digest.Update(&gpdesc.BlendState, sizeof(gpdesc.BlendState));
digest.Update(&gpdesc.SampleMask, sizeof(gpdesc.SampleMask));
digest.Update(&gpdesc.RasterizerState, sizeof(gpdesc.RasterizerState));
digest.Update(&gpdesc.DepthStencilState, sizeof(gpdesc.DepthStencilState));
for (u32 i = 0; i < gpdesc.InputLayout.NumElements; i++)
{
const D3D12_INPUT_ELEMENT_DESC& ie = gpdesc.InputLayout.pInputElementDescs[i];
digest.Update(ie.SemanticName, static_cast<u32>(std::strlen(ie.SemanticName)));
digest.Update(&ie.SemanticIndex, sizeof(ie.SemanticIndex));
digest.Update(&ie.Format, sizeof(ie.Format));
digest.Update(&ie.InputSlot, sizeof(ie.InputSlot));
digest.Update(&ie.AlignedByteOffset, sizeof(ie.AlignedByteOffset));
digest.Update(&ie.InputSlotClass, sizeof(ie.InputSlotClass));
digest.Update(&ie.InstanceDataStepRate, sizeof(ie.InstanceDataStepRate));
length += sizeof(D3D12_INPUT_ELEMENT_DESC);
}
digest.Update(&gpdesc.IBStripCutValue, sizeof(gpdesc.IBStripCutValue));
digest.Update(&gpdesc.PrimitiveTopologyType, sizeof(gpdesc.PrimitiveTopologyType));
digest.Update(&gpdesc.NumRenderTargets, sizeof(gpdesc.NumRenderTargets));
digest.Update(gpdesc.RTVFormats, sizeof(gpdesc.RTVFormats));
digest.Update(&gpdesc.DSVFormat, sizeof(gpdesc.DSVFormat));
digest.Update(&gpdesc.SampleDesc, sizeof(gpdesc.SampleDesc));
digest.Update(&gpdesc.Flags, sizeof(gpdesc.Flags));
MD5Hash h;
digest.Final(h.hash);
return CacheIndexKey{h.low, h.high, length, EntryType::GraphicsPipeline};
}
ShaderCache::ComPtr<ID3DBlob> ShaderCache::GetShaderBlob(EntryType type, std::string_view shader_code)
{
const auto key = GetShaderCacheKey(type, shader_code);
auto iter = m_shader_index.find(key);
if (iter == m_shader_index.end())
return CompileAndAddShaderBlob(key, shader_code);
ComPtr<ID3DBlob> blob;
HRESULT hr = D3DCreateBlob(iter->second.blob_size, blob.GetAddressOf());
if (FAILED(hr) || std::fseek(m_shader_blob_file, iter->second.file_offset, SEEK_SET) != 0 ||
std::fread(blob->GetBufferPointer(), 1, iter->second.blob_size, m_shader_blob_file) != iter->second.blob_size)
{
Log_ErrorPrintf("Read blob from file failed");
return {};
}
return blob;
}
ShaderCache::ComPtr<ID3D12PipelineState> ShaderCache::GetPipelineState(ID3D12Device* device,
const D3D12_GRAPHICS_PIPELINE_STATE_DESC& desc)
{
const auto key = GetPipelineCacheKey(desc);
auto iter = m_pipeline_index.find(key);
if (iter == m_pipeline_index.end())
return CompileAndAddPipeline(device, key, desc);
ComPtr<ID3DBlob> blob;
HRESULT hr = D3DCreateBlob(iter->second.blob_size, blob.GetAddressOf());
if (FAILED(hr) || std::fseek(m_pipeline_blob_file, iter->second.file_offset, SEEK_SET) != 0 ||
std::fread(blob->GetBufferPointer(), 1, iter->second.blob_size, m_pipeline_blob_file) != iter->second.blob_size)
{
Log_ErrorPrintf("Read blob from file failed");
return {};
}
D3D12_GRAPHICS_PIPELINE_STATE_DESC desc_with_blob(desc);
desc_with_blob.CachedPSO.pCachedBlob = blob->GetBufferPointer();
desc_with_blob.CachedPSO.CachedBlobSizeInBytes = blob->GetBufferSize();
ComPtr<ID3D12PipelineState> pso;
hr = device->CreateGraphicsPipelineState(&desc_with_blob, IID_PPV_ARGS(pso.GetAddressOf()));
if (FAILED(hr))
{
Log_WarningPrintf("Creating cached PSO failed: %08X. Invalidating cache.", hr);
InvalidatePipelineCache();
pso = CompileAndAddPipeline(device, key, desc);
}
return pso;
}
static unsigned s_next_bad_shader_id = 1;
ShaderCache::ComPtr<ID3DBlob> ShaderCache::CompileShader(EntryType type, D3D_FEATURE_LEVEL feature_level, std::string_view code, bool debug)
{
const char* target;
switch (feature_level)
{
case D3D_FEATURE_LEVEL_10_0:
{
static constexpr std::array<const char*, 4> targets = {{"vs_4_0", "gs_4_0", "ps_4_0", "cs_4_0"}};
target = targets[static_cast<int>(type)];
}
break;
case D3D_FEATURE_LEVEL_10_1:
{
static constexpr std::array<const char*, 4> targets = {{"vs_4_1", "gs_4_1", "ps_4_1", "cs_4_1"}};
target = targets[static_cast<int>(type)];
}
break;
case D3D_FEATURE_LEVEL_11_0:
{
static constexpr std::array<const char*, 4> targets = {{"vs_5_0", "gs_5_0", "ps_5_0", "cs_5_0"}};
target = targets[static_cast<int>(type)];
}
break;
case D3D_FEATURE_LEVEL_11_1:
default:
{
static constexpr std::array<const char*, 4> targets = {{"vs_5_1", "gs_5_1", "ps_5_1", "cs_5_1"}};
target = targets[static_cast<int>(type)];
}
break;
}
static constexpr UINT flags_non_debug = D3DCOMPILE_OPTIMIZATION_LEVEL3;
static constexpr UINT flags_debug = D3DCOMPILE_SKIP_OPTIMIZATION | D3DCOMPILE_DEBUG;
ComPtr<ID3DBlob> blob;
ComPtr<ID3DBlob> error_blob;
const HRESULT hr =
D3DCompile(code.data(), code.size(), "0", nullptr, nullptr, "main", target, debug ? flags_debug : flags_non_debug,
0, blob.GetAddressOf(), error_blob.GetAddressOf());
std::string error_string;
if (error_blob)
{
error_string.append(static_cast<const char*>(error_blob->GetBufferPointer()), error_blob->GetBufferSize());
error_blob.Reset();
}
if (FAILED(hr))
{
Log_ErrorPrintf("Failed to compile '%s':\n%s", target, error_string.c_str());
std::ofstream ofs(StringUtil::StdStringFromFormat("bad_shader_%u.txt", s_next_bad_shader_id++).c_str(),
std::ofstream::out | std::ofstream::binary);
if (ofs.is_open())
{
ofs << code;
ofs << "\n\nCompile as " << target << " failed: " << hr << "\n";
ofs.write(error_string.c_str(), error_string.size());
ofs.close();
}
return {};
}
if (!error_string.empty())
Log_WarningPrintf("'%s' compiled with warnings:\n%s", target, error_string.c_str());
return blob;
}
ShaderCache::ComPtr<ID3DBlob> ShaderCache::CompileAndAddShaderBlob(const CacheIndexKey& key,
std::string_view shader_code)
{
ComPtr<ID3DBlob> blob = CompileShader(key.type, m_feature_level, shader_code, m_debug);
if (!blob)
return {};
if (!m_shader_blob_file || std::fseek(m_shader_blob_file, 0, SEEK_END) != 0)
return blob;
CacheIndexData data;
data.file_offset = static_cast<u32>(std::ftell(m_shader_blob_file));
data.blob_size = static_cast<u32>(blob->GetBufferSize());
CacheIndexEntry entry = {};
entry.source_hash_low = key.source_hash_low;
entry.source_hash_high = key.source_hash_high;
entry.source_length = key.source_length;
entry.shader_type = static_cast<u32>(key.type);
entry.blob_size = data.blob_size;
entry.file_offset = data.file_offset;
if (std::fwrite(blob->GetBufferPointer(), 1, entry.blob_size, m_shader_blob_file) != entry.blob_size ||
std::fflush(m_shader_blob_file) != 0 || std::fwrite(&entry, sizeof(entry), 1, m_shader_index_file) != 1 ||
std::fflush(m_shader_index_file) != 0)
{
Log_ErrorPrintf("Failed to write shader blob to file");
return blob;
}
m_shader_index.emplace(key, data);
return blob;
}
ShaderCache::ComPtr<ID3D12PipelineState>
ShaderCache::CompileAndAddPipeline(ID3D12Device* device, const CacheIndexKey& key,
const D3D12_GRAPHICS_PIPELINE_STATE_DESC& gpdesc)
{
ComPtr<ID3D12PipelineState> pso;
HRESULT hr = device->CreateGraphicsPipelineState(&gpdesc, IID_PPV_ARGS(pso.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("Creating cached PSO failed: %08X", hr);
return {};
}
if (!m_pipeline_blob_file || std::fseek(m_pipeline_blob_file, 0, SEEK_END) != 0)
return pso;
ComPtr<ID3DBlob> blob;
hr = pso->GetCachedBlob(blob.GetAddressOf());
if (FAILED(hr))
{
Log_WarningPrintf("Failed to get cached PSO data: %08X", hr);
return pso;
}
CacheIndexData data;
data.file_offset = static_cast<u32>(std::ftell(m_pipeline_blob_file));
data.blob_size = static_cast<u32>(blob->GetBufferSize());
CacheIndexEntry entry = {};
entry.source_hash_low = key.source_hash_low;
entry.source_hash_high = key.source_hash_high;
entry.source_length = key.source_length;
entry.shader_type = static_cast<u32>(key.type);
entry.blob_size = data.blob_size;
entry.file_offset = data.file_offset;
if (std::fwrite(blob->GetBufferPointer(), 1, entry.blob_size, m_pipeline_blob_file) != entry.blob_size ||
std::fflush(m_pipeline_blob_file) != 0 || std::fwrite(&entry, sizeof(entry), 1, m_pipeline_index_file) != 1 ||
std::fflush(m_pipeline_index_file) != 0)
{
Log_ErrorPrintf("Failed to write pipeline blob to file");
return pso;
}
m_shader_index.emplace(key, data);
return pso;
}
} // namespace D3D12

122
src/core/gpu/d3d12/shader_cache.h
View File

@ -1,122 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "common/hash_combine.h"
#include "common/types.h"
#include "common/windows_headers.h"
#include <cstdio>
#include <d3d12.h>
#include <string_view>
#include <unordered_map>
#include <vector>
#include <wrl/client.h>
namespace D3D12 {
class ShaderCache
{
public:
template<typename T>
using ComPtr = Microsoft::WRL::ComPtr<T>;
enum class EntryType
{
VertexShader,
GeometryShader,
PixelShader,
ComputeShader,
GraphicsPipeline,
};
ShaderCache();
~ShaderCache();
void Open(std::string_view base_path, D3D_FEATURE_LEVEL feature_level, bool debug);
ALWAYS_INLINE ComPtr<ID3DBlob> GetVertexShader(std::string_view shader_code)
{
return GetShaderBlob(EntryType::VertexShader, shader_code);
}
ALWAYS_INLINE ComPtr<ID3DBlob> GetGeometryShader(std::string_view shader_code)
{
return GetShaderBlob(EntryType::GeometryShader, shader_code);
}
ALWAYS_INLINE ComPtr<ID3DBlob> GetPixelShader(std::string_view shader_code)
{
return GetShaderBlob(EntryType::PixelShader, shader_code);
}
ALWAYS_INLINE ComPtr<ID3DBlob> GetComputeShader(std::string_view shader_code)
{
return GetShaderBlob(EntryType::ComputeShader, shader_code);
}
ComPtr<ID3DBlob> GetShaderBlob(EntryType type, std::string_view shader_code);
ComPtr<ID3D12PipelineState> GetPipelineState(ID3D12Device* device, const D3D12_GRAPHICS_PIPELINE_STATE_DESC& desc);
private:
static constexpr u32 FILE_VERSION = 1;
struct CacheIndexKey
{
u64 source_hash_low;
u64 source_hash_high;
u32 source_length;
EntryType type;
bool operator==(const CacheIndexKey& key) const;
bool operator!=(const CacheIndexKey& key) const;
};
struct CacheIndexEntryHasher
{
std::size_t operator()(const CacheIndexKey& e) const noexcept
{
std::size_t h = 0;
hash_combine(h, e.source_hash_low, e.source_hash_high, e.source_length, e.type);
return h;
}
};
struct CacheIndexData
{
u32 file_offset;
u32 blob_size;
};
using CacheIndex = std::unordered_map<CacheIndexKey, CacheIndexData, CacheIndexEntryHasher>;
static ComPtr<ID3DBlob> CompileShader(EntryType type, D3D_FEATURE_LEVEL feature_level, std::string_view code, bool debug);
static std::string GetCacheBaseFileName(const std::string_view& base_path, const std::string_view& type,
D3D_FEATURE_LEVEL feature_level, bool debug);
static CacheIndexKey GetShaderCacheKey(EntryType type, const std::string_view& shader_code);
static CacheIndexKey GetPipelineCacheKey(const D3D12_GRAPHICS_PIPELINE_STATE_DESC& gpdesc);
bool CreateNew(const std::string& index_filename, const std::string& blob_filename, std::FILE*& index_file,
std::FILE*& blob_file);
bool ReadExisting(const std::string& index_filename, const std::string& blob_filename, std::FILE*& index_file,
std::FILE*& blob_file, CacheIndex& index);
void InvalidatePipelineCache();
void Close();
ComPtr<ID3DBlob> CompileAndAddShaderBlob(const CacheIndexKey& key, std::string_view shader_code);
ComPtr<ID3D12PipelineState> CompileAndAddPipeline(ID3D12Device* device, const CacheIndexKey& key,
const D3D12_GRAPHICS_PIPELINE_STATE_DESC& gpdesc);
std::string m_base_path;
std::FILE* m_shader_index_file = nullptr;
std::FILE* m_shader_blob_file = nullptr;
CacheIndex m_shader_index;
std::FILE* m_pipeline_index_file = nullptr;
std::FILE* m_pipeline_blob_file = nullptr;
CacheIndex m_pipeline_index;
D3D_FEATURE_LEVEL m_feature_level = D3D_FEATURE_LEVEL_11_0;
bool m_debug = false;
};
} // namespace D3D12

239
src/core/gpu/d3d12/staging_texture.cpp
View File

@ -1,239 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "staging_texture.h"
#include "common/align.h"
#include "common/assert.h"
#include "common/log.h"
#include "context.h"
#include "util.h"
Log_SetChannel(D3D12);
namespace D3D12 {
StagingTexture::StagingTexture() : m_width(0), m_height(0) {}
StagingTexture::~StagingTexture()
{
Destroy();
}
bool StagingTexture::Create(u32 width, u32 height, DXGI_FORMAT format, bool for_uploading)
{
const u32 texel_size = GetTexelSize(format);
const u32 row_pitch = Common::AlignUpPow2(width * texel_size, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const u32 buffer_size = height * row_pitch;
const D3D12_HEAP_PROPERTIES heap_properties = {for_uploading ? D3D12_HEAP_TYPE_UPLOAD : D3D12_HEAP_TYPE_READBACK};
D3D12_RESOURCE_DESC desc = {};
desc.Dimension = D3D12_RESOURCE_DIMENSION_BUFFER;
desc.Width = buffer_size;
desc.Height = 1;
desc.DepthOrArraySize = 1;
desc.MipLevels = 1;
desc.Format = DXGI_FORMAT_UNKNOWN;
desc.SampleDesc.Count = 1;
desc.Layout = D3D12_TEXTURE_LAYOUT_ROW_MAJOR;
desc.Flags = D3D12_RESOURCE_FLAG_NONE;
D3D12_RESOURCE_STATES state = for_uploading ? D3D12_RESOURCE_STATE_GENERIC_READ : D3D12_RESOURCE_STATE_COPY_DEST;
ComPtr<ID3D12Resource> resource;
HRESULT hr = g_d3d12_context->GetDevice()->CreateCommittedResource(
&heap_properties, D3D12_HEAP_FLAG_NONE, &desc, state, nullptr, IID_PPV_ARGS(resource.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("Create buffer failed: 0x%08X", hr);
return false;
}
Destroy(true);
m_resource = std::move(resource);
m_width = width;
m_height = height;
m_format = format;
m_buffer_size = buffer_size;
m_row_pitch = row_pitch;
m_texel_size = texel_size;
return true;
}
void StagingTexture::Destroy(bool defer)
{
if (IsMapped())
Unmap();
if (m_resource && defer)
g_d3d12_context->DeferResourceDestruction(m_resource.Get());
m_resource.Reset();
m_width = 0;
m_height = 0;
m_format = DXGI_FORMAT_UNKNOWN;
m_buffer_size = 0;
m_row_pitch = 0;
m_texel_size = 0;
}
bool StagingTexture::Map(bool writing)
{
D3D12_RANGE range{0u, m_buffer_size};
Assert(!IsMapped());
const HRESULT hr = m_resource->Map(0, writing ? nullptr : &range, &m_mapped_pointer);
if (FAILED(hr))
{
Log_ErrorPrintf("Map staging buffer failed: 0x%08X", hr);
return false;
}
m_mapped_for_write = writing;
return true;
}
void StagingTexture::Unmap()
{
Assert(IsMapped());
D3D12_RANGE range{0u, m_buffer_size};
m_resource->Unmap(0, m_mapped_for_write ? &range : nullptr);
m_mapped_pointer = nullptr;
m_mapped_for_write = false;
}
void StagingTexture::Flush()
{
if (!m_needs_flush)
return;
m_needs_flush = false;
// If the completed fence is the same as the current command buffer fence, we need to execute
// the current list and wait for it to complete. This is the slowest path. Otherwise, if the
// command list with the copy has been submitted, we only need to wait for the fence.
if (m_completed_fence == g_d3d12_context->GetCurrentFenceValue())
g_d3d12_context->ExecuteCommandList(true);
else
g_d3d12_context->WaitForFence(m_completed_fence);
}
void StagingTexture::CopyToTexture(u32 src_x, u32 src_y, ID3D12Resource* dst_texture, u32 dst_subresource, u32 dst_x,
u32 dst_y, u32 width, u32 height)
{
DebugAssert((src_x + width) <= m_width && (src_y + height) <= m_height);
D3D12_TEXTURE_COPY_LOCATION dst;
dst.pResource = dst_texture;
dst.SubresourceIndex = 0;
dst.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX;
D3D12_TEXTURE_COPY_LOCATION src;
src.pResource = m_resource.Get();
src.SubresourceIndex = 0;
src.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT;
src.PlacedFootprint.Offset = 0;
src.PlacedFootprint.Footprint.Width = m_width;
src.PlacedFootprint.Footprint.Height = m_height;
src.PlacedFootprint.Footprint.Depth = 1;
src.PlacedFootprint.Footprint.Format = m_format;
src.PlacedFootprint.Footprint.RowPitch = m_row_pitch;
const D3D12_BOX src_box{src_x, src_y, 0u, src_x + width, src_y + height, 1u};
g_d3d12_context->GetCommandList()->CopyTextureRegion(&dst, dst_x, dst_y, 0, &src, &src_box);
}
void StagingTexture::CopyFromTexture(ID3D12Resource* src_texture, u32 src_subresource, u32 src_x, u32 src_y, u32 dst_x,
u32 dst_y, u32 width, u32 height)
{
DebugAssert((dst_x + width) <= m_width && (dst_y + height) <= m_height);
D3D12_TEXTURE_COPY_LOCATION src;
src.pResource = src_texture;
src.SubresourceIndex = 0;
src.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX;
D3D12_TEXTURE_COPY_LOCATION dst;
dst.pResource = m_resource.Get();
dst.SubresourceIndex = 0;
dst.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT;
dst.PlacedFootprint.Offset = 0;
dst.PlacedFootprint.Footprint.Width = m_width;
dst.PlacedFootprint.Footprint.Height = m_height;
dst.PlacedFootprint.Footprint.Depth = 1;
dst.PlacedFootprint.Footprint.Format = m_format;
dst.PlacedFootprint.Footprint.RowPitch = m_row_pitch;
const D3D12_BOX src_box{src_x, src_y, 0u, src_x + width, src_y + height, 1u};
g_d3d12_context->GetCommandList()->CopyTextureRegion(&dst, dst_x, dst_y, 0, &src, &src_box);
m_completed_fence = g_d3d12_context->GetCurrentFenceValue();
m_needs_flush = true;
}
bool StagingTexture::ReadPixels(u32 x, u32 y, u32 width, u32 height, void* data, u32 row_pitch)
{
if (m_needs_flush)
Flush();
const bool was_mapped = IsMapped();
if (!was_mapped && !Map(false))
return false;
const u8* src_ptr = static_cast<u8*>(m_mapped_pointer) + (y * m_row_pitch) + (x * m_texel_size);
u8* dst_ptr = reinterpret_cast<u8*>(data);
if (m_row_pitch != row_pitch || width != m_width || x != 0)
{
const u32 copy_size = m_texel_size * width;
for (u32 row = 0; row < height; row++)
{
std::memcpy(dst_ptr, src_ptr, copy_size);
src_ptr += m_row_pitch;
dst_ptr += row_pitch;
}
}
else
{
std::memcpy(dst_ptr, src_ptr, row_pitch * height);
}
return true;
}
bool StagingTexture::WritePixels(u32 x, u32 y, u32 width, u32 height, const void* data, u32 row_pitch)
{
const bool was_mapped = IsMapped();
if (!was_mapped && !Map(true))
return false;
const u8* src_ptr = reinterpret_cast<const u8*>(data);
u8* dst_ptr = static_cast<u8*>(m_mapped_pointer) + (y * m_row_pitch) + (x * m_texel_size);
if (m_row_pitch != row_pitch || width != m_width || x != 0)
{
const u32 copy_size = m_texel_size * width;
for (u32 row = 0; row < height; row++)
{
std::memcpy(dst_ptr, src_ptr, copy_size);
src_ptr += row_pitch;
dst_ptr += m_row_pitch;
}
}
else
{
std::memcpy(dst_ptr, src_ptr, m_row_pitch * height);
}
if (!was_mapped)
Unmap();
return true;
}
bool StagingTexture::EnsureSize(u32 width, u32 height, DXGI_FORMAT format, bool for_uploading)
{
if (m_resource && m_width >= width && m_height >= height && m_format == format)
return true;
return Create(width, height, format, for_uploading);
}
} // namespace D3D12

64
src/core/gpu/d3d12/staging_texture.h
View File

@ -1,64 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "common/types.h"
#include "common/windows_headers.h"
#include <cstring>
#include <d3d12.h>
#include <wrl/client.h>
namespace D3D12 {
class StagingTexture
{
public:
template<typename T>
using ComPtr = Microsoft::WRL::ComPtr<T>;
StagingTexture();
~StagingTexture();
ALWAYS_INLINE ID3D12Resource* GetD3DResource() const { return m_resource.Get(); }
ALWAYS_INLINE u32 GetWidth() const { return m_width; }
ALWAYS_INLINE u32 GetHeight() const { return m_height; }
ALWAYS_INLINE DXGI_FORMAT GetFormat() const { return m_format; }
ALWAYS_INLINE bool IsMapped() const { return m_mapped_pointer != nullptr; }
ALWAYS_INLINE const void* GetMapPointer() const { return m_mapped_pointer; }
ALWAYS_INLINE operator bool() const { return static_cast<bool>(m_resource); }
bool Create(u32 width, u32 height, DXGI_FORMAT format, bool for_uploading);
void Destroy(bool defer = true);
bool Map(bool writing);
void Unmap();
void Flush();
void CopyToTexture(u32 src_x, u32 src_y, ID3D12Resource* dst_texture, u32 dst_subresource, u32 dst_x, u32 dst_y,
u32 width, u32 height);
void CopyFromTexture(ID3D12Resource* src_texture, u32 src_subresource, u32 src_x, u32 src_y, u32 dst_x, u32 dst_y,
u32 width, u32 height);
bool ReadPixels(u32 x, u32 y, u32 width, u32 height, void* data, u32 row_pitch);
bool WritePixels(u32 x, u32 y, u32 width, u32 height, const void* data, u32 row_pitch);
bool EnsureSize(u32 width, u32 height, DXGI_FORMAT format, bool for_uploading);
protected:
ComPtr<ID3D12Resource> m_resource;
u32 m_width;
u32 m_height;
DXGI_FORMAT m_format;
u32 m_texel_size;
u32 m_row_pitch;
u32 m_buffer_size;
void* m_mapped_pointer = nullptr;
u64 m_completed_fence = 0;
bool m_mapped_for_write = false;
bool m_needs_flush = false;
};
} // namespace D3D12

466
src/core/gpu/d3d12/texture.cpp
View File

@ -1,466 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "texture.h"
#include "common/align.h"
#include "common/assert.h"
#include "common/log.h"
#include "context.h"
#include "staging_texture.h"
#include "stream_buffer.h"
#include "util.h"
Log_SetChannel(D3D12);
static constexpr std::array<DXGI_FORMAT, static_cast<u32>(GPUTexture::Format::MaxCount)> s_dxgi_mapping = {
{DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B5G6R5_UNORM,
DXGI_FORMAT_B5G5R5A1_UNORM, DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_D16_UNORM}};
D3D12::Texture::Texture() = default;
D3D12::Texture::Texture(ID3D12Resource* resource, D3D12_RESOURCE_STATES state) : m_resource(std::move(resource))
{
const D3D12_RESOURCE_DESC desc = GetDesc();
m_width = static_cast<u16>(desc.Width);
m_height = static_cast<u16>(desc.Height);
m_layers = static_cast<u8>(desc.DepthOrArraySize);
m_levels = static_cast<u8>(desc.MipLevels);
m_samples = static_cast<u8>(desc.SampleDesc.Count);
m_format = LookupBaseFormat(desc.Format);
}
D3D12::Texture::Texture(Texture&& texture)
: m_resource(std::move(texture.m_resource)), m_srv_descriptor(texture.m_srv_descriptor),
m_rtv_or_dsv_descriptor(texture.m_rtv_or_dsv_descriptor), m_is_depth_view(texture.m_is_depth_view)
{
m_width = texture.m_width;
m_height = texture.m_height;
m_layers = texture.m_layers;
m_levels = texture.m_levels;
m_samples = texture.m_samples;
texture.m_srv_descriptor = {};
texture.m_rtv_or_dsv_descriptor = {};
texture.m_state = D3D12_RESOURCE_STATE_COMMON;
texture.m_is_depth_view = false;
texture.ClearBaseProperties();
}
DXGI_FORMAT D3D12::Texture::GetDXGIFormat(Format format)
{
return s_dxgi_mapping[static_cast<u8>(format)];
}
GPUTexture::Format D3D12::Texture::LookupBaseFormat(DXGI_FORMAT dformat)
{
for (u32 i = 0; i < static_cast<u32>(s_dxgi_mapping.size()); i++)
{
if (s_dxgi_mapping[i] == dformat)
return static_cast<Format>(i);
}
return GPUTexture::Format::Unknown;
}
D3D12::Texture::~Texture()
{
Destroy();
}
D3D12::Texture& D3D12::Texture::operator=(Texture&& texture)
{
Destroy();
m_width = texture.m_width;
m_height = texture.m_height;
m_layers = texture.m_layers;
m_levels = texture.m_levels;
m_samples = texture.m_samples;
m_resource = std::move(texture.m_resource);
m_srv_descriptor = texture.m_srv_descriptor;
m_rtv_or_dsv_descriptor = texture.m_rtv_or_dsv_descriptor;
m_state = texture.m_state;
m_is_depth_view = texture.m_is_depth_view;
texture.ClearBaseProperties();
texture.m_srv_descriptor = {};
texture.m_rtv_or_dsv_descriptor = {};
texture.m_state = D3D12_RESOURCE_STATE_COMMON;
texture.m_is_depth_view = false;
return *this;
}
D3D12_RESOURCE_DESC D3D12::Texture::GetDesc() const
{
return m_resource->GetDesc();
}
bool D3D12::Texture::IsValid() const
{
return static_cast<bool>(m_resource);
}
bool D3D12::Texture::Update(u32 x, u32 y, u32 width, u32 height, const void* data, u32 pitch, u32 layer /*= 0*/,
u32 level /*= 0*/)
{
UnreachableCode();
return false;
}
bool D3D12::Texture::Map(void** map, u32* map_stride, u32 x, u32 y, u32 width, u32 height, u32 layer /*= 0*/,
u32 level /*= 0*/)
{
UnreachableCode();
return false;
}
void D3D12::Texture::Unmap()
{
UnreachableCode();
}
bool D3D12::Texture::Create(u32 width, u32 height, u32 layers, u32 levels, u32 samples, DXGI_FORMAT format,
DXGI_FORMAT srv_format, DXGI_FORMAT rtv_format, DXGI_FORMAT dsv_format,
D3D12_RESOURCE_FLAGS flags)
{
constexpr D3D12_HEAP_PROPERTIES heap_properties = {D3D12_HEAP_TYPE_DEFAULT};
if (width > MAX_WIDTH || height > MAX_HEIGHT || layers > MAX_LAYERS || levels > MAX_LEVELS || samples > MAX_SAMPLES)
{
Log_ErrorPrintf("Invalid dimensions: %ux%ux%u %u %u", width, height, layers, levels, samples);
return false;
}
D3D12_RESOURCE_DESC desc = {};
desc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D;
desc.Width = width;
desc.Height = static_cast<u16>(height);
desc.DepthOrArraySize = static_cast<u16>(layers);
desc.MipLevels = static_cast<u16>(levels);
desc.Format = format;
desc.SampleDesc.Count = samples;
desc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN;
desc.Flags = flags;
D3D12_CLEAR_VALUE optimized_clear_value = {};
D3D12_RESOURCE_STATES state;
if (rtv_format != DXGI_FORMAT_UNKNOWN)
{
optimized_clear_value.Format = rtv_format;
state = D3D12_RESOURCE_STATE_RENDER_TARGET;
}
else if (dsv_format != DXGI_FORMAT_UNKNOWN)
{
optimized_clear_value.Format = dsv_format;
state = D3D12_RESOURCE_STATE_DEPTH_WRITE;
}
else
{
state = D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE;
}
ComPtr<ID3D12Resource> resource;
HRESULT hr = g_d3d12_context->GetDevice()->CreateCommittedResource(
&heap_properties, D3D12_HEAP_FLAG_NONE, &desc, state,
(rtv_format != DXGI_FORMAT_UNKNOWN || dsv_format != DXGI_FORMAT_UNKNOWN) ? &optimized_clear_value : nullptr,
IID_PPV_ARGS(resource.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("Create texture failed: 0x%08X", hr);
return false;
}
DescriptorHandle srv_descriptor, rtv_descriptor;
bool is_depth_view = false;
if (srv_format != DXGI_FORMAT_UNKNOWN)
{
if (!CreateSRVDescriptor(resource.Get(), srv_format, samples > 1, &srv_descriptor))
return false;
}
if (rtv_format != DXGI_FORMAT_UNKNOWN)
{
Assert(dsv_format == DXGI_FORMAT_UNKNOWN);
if (!CreateRTVDescriptor(resource.Get(), rtv_format, samples > 1, &rtv_descriptor))
{
g_d3d12_context->GetDescriptorHeapManager().Free(&srv_descriptor);
return false;
}
}
else if (dsv_format != DXGI_FORMAT_UNKNOWN)
{
if (!CreateDSVDescriptor(resource.Get(), dsv_format, samples > 1, &rtv_descriptor))
{
g_d3d12_context->GetDescriptorHeapManager().Free(&srv_descriptor);
return false;
}
is_depth_view = true;
}
Destroy(true);
m_resource = std::move(resource);
m_srv_descriptor = std::move(srv_descriptor);
m_rtv_or_dsv_descriptor = std::move(rtv_descriptor);
m_width = static_cast<u16>(width);
m_height = static_cast<u16>(height);
m_layers = static_cast<u8>(layers);
m_levels = static_cast<u8>(levels);
m_samples = static_cast<u8>(samples);
m_format = LookupBaseFormat(format);
m_state = state;
m_is_depth_view = is_depth_view;
return true;
}
bool D3D12::Texture::Adopt(ComPtr<ID3D12Resource> texture, DXGI_FORMAT srv_format, DXGI_FORMAT rtv_format,
DXGI_FORMAT dsv_format, D3D12_RESOURCE_STATES state)
{
const D3D12_RESOURCE_DESC desc(texture->GetDesc());
DescriptorHandle srv_descriptor, rtv_descriptor;
if (srv_format != DXGI_FORMAT_UNKNOWN)
{
if (!CreateSRVDescriptor(texture.Get(), srv_format, desc.SampleDesc.Count > 1, &srv_descriptor))
return false;
}
m_is_depth_view = false;
if (rtv_format != DXGI_FORMAT_UNKNOWN)
{
Assert(dsv_format == DXGI_FORMAT_UNKNOWN);
if (!CreateRTVDescriptor(texture.Get(), rtv_format, desc.SampleDesc.Count > 1, &rtv_descriptor))
{
g_d3d12_context->GetDescriptorHeapManager().Free(&srv_descriptor);
return false;
}
}
else if (dsv_format != DXGI_FORMAT_UNKNOWN)
{
if (!CreateDSVDescriptor(texture.Get(), dsv_format, desc.SampleDesc.Count > 1, &rtv_descriptor))
{
g_d3d12_context->GetDescriptorHeapManager().Free(&srv_descriptor);
return false;
}
m_is_depth_view = true;
}
m_resource = std::move(texture);
m_srv_descriptor = std::move(srv_descriptor);
m_rtv_or_dsv_descriptor = std::move(rtv_descriptor);
m_width = static_cast<u16>(desc.Width);
m_height = static_cast<u16>(desc.Height);
m_layers = static_cast<u8>(desc.DepthOrArraySize);
m_levels = static_cast<u8>(desc.MipLevels);
m_samples = static_cast<u8>(desc.SampleDesc.Count);
m_format = LookupBaseFormat(desc.Format);
m_state = state;
return true;
}
void D3D12::Texture::Destroy(bool defer /* = true */)
{
if (defer)
{
g_d3d12_context->DeferDescriptorDestruction(g_d3d12_context->GetDescriptorHeapManager(), &m_srv_descriptor);
if (m_is_depth_view)
g_d3d12_context->DeferDescriptorDestruction(g_d3d12_context->GetDSVHeapManager(), &m_rtv_or_dsv_descriptor);
else
g_d3d12_context->DeferDescriptorDestruction(g_d3d12_context->GetRTVHeapManager(), &m_rtv_or_dsv_descriptor);
g_d3d12_context->DeferResourceDestruction(m_resource.Get());
m_resource.Reset();
}
else
{
g_d3d12_context->GetDescriptorHeapManager().Free(&m_srv_descriptor);
if (m_is_depth_view)
g_d3d12_context->GetDSVHeapManager().Free(&m_rtv_or_dsv_descriptor);
else
g_d3d12_context->GetRTVHeapManager().Free(&m_rtv_or_dsv_descriptor);
m_resource.Reset();
}
ClearBaseProperties();
m_is_depth_view = false;
}
void D3D12::Texture::TransitionToState(D3D12_RESOURCE_STATES state) const
{
if (m_state == state)
return;
ResourceBarrier(g_d3d12_context->GetCommandList(), m_resource.Get(), m_state, state);
m_state = state;
}
bool D3D12::Texture::BeginStreamUpdate(u32 x, u32 y, u32 width, u32 height, void** out_data, u32* out_data_pitch)
{
const u32 copy_pitch = Common::AlignUpPow2(width * GetPixelSize(), D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const u32 upload_size = copy_pitch * height;
if (!g_d3d12_context->GetTextureStreamBuffer().ReserveMemory(upload_size, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT))
{
Log_PerfPrintf("Executing command buffer while waiting for %u bytes (%ux%u) in upload buffer", upload_size, width,
height);
g_d3d12_context->ExecuteCommandList(false);
if (!g_d3d12_context->GetTextureStreamBuffer().ReserveMemory(upload_size, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT))
{
Log_ErrorPrintf("Failed to reserve %u bytes for %ux%u upload", upload_size, width, height);
return false;
}
}
*out_data = g_d3d12_context->GetTextureStreamBuffer().GetCurrentHostPointer();
*out_data_pitch = copy_pitch;
return true;
}
void D3D12::Texture::EndStreamUpdate(u32 x, u32 y, u32 width, u32 height)
{
const u32 copy_pitch = Common::AlignUpPow2(width * GetPixelSize(), D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const u32 upload_size = copy_pitch * height;
StreamBuffer& sb = g_d3d12_context->GetTextureStreamBuffer();
const u32 sb_offset = sb.GetCurrentOffset();
sb.CommitMemory(upload_size);
CopyFromBuffer(x, y, width, height, copy_pitch, sb.GetBuffer(), sb_offset);
}
void D3D12::Texture::CopyFromBuffer(u32 x, u32 y, u32 width, u32 height, u32 pitch, ID3D12Resource* buffer,
u32 buffer_offset)
{
D3D12_TEXTURE_COPY_LOCATION src;
src.pResource = buffer;
src.SubresourceIndex = 0;
src.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT;
src.PlacedFootprint.Offset = buffer_offset;
src.PlacedFootprint.Footprint.Width = width;
src.PlacedFootprint.Footprint.Height = height;
src.PlacedFootprint.Footprint.Depth = 1;
src.PlacedFootprint.Footprint.RowPitch = pitch;
src.PlacedFootprint.Footprint.Format = GetDXGIFormat();
D3D12_TEXTURE_COPY_LOCATION dst;
dst.pResource = m_resource.Get();
dst.SubresourceIndex = 0;
dst.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX;
const D3D12_BOX src_box{0u, 0u, 0u, width, height, 1u};
const D3D12_RESOURCE_STATES old_state = m_state;
TransitionToState(D3D12_RESOURCE_STATE_COPY_DEST);
g_d3d12_context->GetCommandList()->CopyTextureRegion(&dst, x, y, 0, &src, &src_box);
TransitionToState(old_state);
}
void D3D12::Texture::SetDebugName(const std::string_view& name)
{
UnreachableCode();
}
void D3D12::Texture::MakeReadyForSampling()
{
UnreachableCode();
}
bool D3D12::Texture::LoadData(u32 x, u32 y, u32 width, u32 height, const void* data, u32 pitch)
{
const u32 texel_size = GetPixelSize();
const u32 upload_pitch = Common::AlignUpPow2(width * texel_size, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const u32 upload_size = upload_pitch * height;
if (upload_size >= g_d3d12_context->GetTextureStreamBuffer().GetSize())
{
StagingTexture st;
if (!st.Create(width, height, GetDXGIFormat(), true) || !st.WritePixels(0, 0, width, height, data, pitch))
return false;
D3D12_RESOURCE_STATES old_state = m_state;
TransitionToState(D3D12_RESOURCE_STATE_COPY_DEST);
st.CopyToTexture(0, 0, m_resource.Get(), 0, x, y, width, height);
st.Destroy(true);
TransitionToState(old_state);
return true;
}
void* write_ptr;
u32 write_pitch;
if (!BeginStreamUpdate(x, y, width, height, &write_ptr, &write_pitch))
return false;
CopyToUploadBuffer(data, pitch, height, write_ptr, write_pitch);
EndStreamUpdate(x, y, width, height);
return true;
}
void D3D12::Texture::CopyToUploadBuffer(const void* src_data, u32 src_pitch, u32 height, void* dst_data, u32 dst_pitch)
{
const u8* src_ptr = static_cast<const u8*>(src_data);
u8* dst_ptr = static_cast<u8*>(dst_data);
if (src_pitch == dst_pitch)
{
std::memcpy(dst_ptr, src_ptr, dst_pitch * height);
}
else
{
const u32 copy_size = std::min(src_pitch, dst_pitch);
for (u32 row = 0; row < height; row++)
{
std::memcpy(dst_ptr, src_ptr, copy_size);
src_ptr += src_pitch;
dst_ptr += dst_pitch;
}
}
}
bool D3D12::Texture::CreateSRVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, bool multisampled,
DescriptorHandle* dh)
{
if (!g_d3d12_context->GetDescriptorHeapManager().Allocate(dh))
{
Log_ErrorPrintf("Failed to allocate SRV descriptor");
return false;
}
D3D12_SHADER_RESOURCE_VIEW_DESC desc = {
format, multisampled ? D3D12_SRV_DIMENSION_TEXTURE2DMS : D3D12_SRV_DIMENSION_TEXTURE2D,
D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING};
if (!multisampled)
desc.Texture2D.MipLevels = 1;
g_d3d12_context->GetDevice()->CreateShaderResourceView(resource, &desc, dh->cpu_handle);
return true;
}
bool D3D12::Texture::CreateRTVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, bool multisampled,
DescriptorHandle* dh)
{
if (!g_d3d12_context->GetRTVHeapManager().Allocate(dh))
{
Log_ErrorPrintf("Failed to allocate SRV descriptor");
return false;
}
D3D12_RENDER_TARGET_VIEW_DESC desc = {format,
multisampled ? D3D12_RTV_DIMENSION_TEXTURE2DMS : D3D12_RTV_DIMENSION_TEXTURE2D};
g_d3d12_context->GetDevice()->CreateRenderTargetView(resource, &desc, dh->cpu_handle);
return true;
}
bool D3D12::Texture::CreateDSVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, bool multisampled,
DescriptorHandle* dh)
{
if (!g_d3d12_context->GetDSVHeapManager().Allocate(dh))
{
Log_ErrorPrintf("Failed to allocate SRV descriptor");
return false;
}
D3D12_DEPTH_STENCIL_VIEW_DESC desc = {
format, multisampled ? D3D12_DSV_DIMENSION_TEXTURE2DMS : D3D12_DSV_DIMENSION_TEXTURE2D, D3D12_DSV_FLAG_NONE};
g_d3d12_context->GetDevice()->CreateDepthStencilView(resource, &desc, dh->cpu_handle);
return true;
}

86
src/core/gpu/d3d12/texture.h
View File

@ -1,86 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "../gpu_texture.h"
#include "common/windows_headers.h"
#include "descriptor_heap_manager.h"
#include <d3d12.h>
#include <wrl/client.h>
namespace D3D12 {
class StreamBuffer;
class Texture final : public GPUTexture
{
public:
template<typename T>
using ComPtr = Microsoft::WRL::ComPtr<T>;
Texture();
Texture(ID3D12Resource* resource, D3D12_RESOURCE_STATES state);
Texture(Texture&& texture);
Texture(const Texture&) = delete;
~Texture();
static DXGI_FORMAT GetDXGIFormat(Format format);
static Format LookupBaseFormat(DXGI_FORMAT dformat);
ALWAYS_INLINE ID3D12Resource* GetResource() const { return m_resource.Get(); }
ALWAYS_INLINE const DescriptorHandle& GetSRVDescriptor() const { return m_srv_descriptor; }
ALWAYS_INLINE const DescriptorHandle& GetRTVOrDSVDescriptor() const { return m_rtv_or_dsv_descriptor; }
ALWAYS_INLINE D3D12_RESOURCE_STATES GetState() const { return m_state; }
ALWAYS_INLINE DXGI_FORMAT GetDXGIFormat() const { return GetDXGIFormat(m_format); }
ALWAYS_INLINE operator ID3D12Resource*() const { return m_resource.Get(); }
ALWAYS_INLINE operator bool() const { return static_cast<bool>(m_resource); }
bool IsValid() const override;
bool Update(u32 x, u32 y, u32 width, u32 height, const void* data, u32 pitch, u32 layer = 0, u32 level = 0) override;
bool Map(void** map, u32* map_stride, u32 x, u32 y, u32 width, u32 height, u32 layer = 0, u32 level = 0) override;
void Unmap() override;
bool Create(u32 width, u32 height, u32 layers, u32 levels, u32 samples, DXGI_FORMAT format, DXGI_FORMAT srv_format, DXGI_FORMAT rtv_format,
DXGI_FORMAT dsv_format, D3D12_RESOURCE_FLAGS flags);
bool Adopt(ComPtr<ID3D12Resource> texture, DXGI_FORMAT srv_format, DXGI_FORMAT rtv_format, DXGI_FORMAT dsv_format,
D3D12_RESOURCE_STATES state);
D3D12_RESOURCE_DESC GetDesc() const;
void Destroy(bool defer = true);
void TransitionToState(D3D12_RESOURCE_STATES state) const;
Texture& operator=(const Texture&) = delete;
Texture& operator=(Texture&& texture);
bool BeginStreamUpdate(u32 x, u32 y, u32 width, u32 height, void** out_data, u32* out_data_pitch);
void EndStreamUpdate(u32 x, u32 y, u32 width, u32 height);
bool LoadData(u32 x, u32 y, u32 width, u32 height, const void* data, u32 pitch);
static void CopyToUploadBuffer(const void* src_data, u32 src_pitch, u32 height, void* dst_data, u32 dst_pitch);
void CopyFromBuffer(u32 x, u32 y, u32 width, u32 height, u32 pitch, ID3D12Resource* buffer, u32 buffer_offset);
void SetDebugName(const std::string_view& name) override;
void MakeReadyForSampling() override;
private:
static bool CreateSRVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, bool multisampled,
DescriptorHandle* dh);
static bool CreateRTVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, bool multisampled,
DescriptorHandle* dh);
static bool CreateDSVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, bool multisampled,
DescriptorHandle* dh);
ComPtr<ID3D12Resource> m_resource;
DescriptorHandle m_srv_descriptor = {};
DescriptorHandle m_rtv_or_dsv_descriptor = {};
mutable D3D12_RESOURCE_STATES m_state = D3D12_RESOURCE_STATE_COMMON;
bool m_is_depth_view = false;
};
} // namespace D3D12

392
src/core/gpu/d3d12/util.cpp
View File

@ -1,392 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "util.h"
#include "common/assert.h"
#include "common/log.h"
#include "common/string.h"
#include "common/string_util.h"
#include "context.h"
#include "shader_cache.h"
#include <cstdarg>
#include <limits>
Log_SetChannel(D3D12);
namespace D3D12 {
void ResourceBarrier(ID3D12GraphicsCommandList* cmdlist, ID3D12Resource* resource, D3D12_RESOURCE_STATES from_state,
D3D12_RESOURCE_STATES to_state)
{
const D3D12_RESOURCE_BARRIER barrier = {D3D12_RESOURCE_BARRIER_TYPE_TRANSITION,
D3D12_RESOURCE_BARRIER_FLAG_NONE,
{{resource, D3D12_RESOURCE_BARRIER_ALL_SUBRESOURCES, from_state, to_state}}};
cmdlist->ResourceBarrier(1, &barrier);
}
void SetViewport(ID3D12GraphicsCommandList* cmdlist, int x, int y, int width, int height, float min_depth /*= 0.0f*/,
float max_depth /*= 1.0f*/)
{
const D3D12_VIEWPORT vp{static_cast<float>(x),
static_cast<float>(y),
static_cast<float>(width),
static_cast<float>(height),
min_depth,
max_depth};
cmdlist->RSSetViewports(1, &vp);
}
void SetScissor(ID3D12GraphicsCommandList* cmdlist, int x, int y, int width, int height)
{
const D3D12_RECT r{x, y, x + width, y + height};
cmdlist->RSSetScissorRects(1, &r);
}
void SetViewportAndScissor(ID3D12GraphicsCommandList* cmdlist, int x, int y, int width, int height,
float min_depth /*= 0.0f*/, float max_depth /*= 1.0f*/)
{
SetViewport(cmdlist, x, y, width, height, min_depth, max_depth);
SetScissor(cmdlist, x, y, width, height);
}
void SetViewportAndClampScissor(ID3D12GraphicsCommandList* cmdlist, int x, int y, int width, int height,
float min_depth /*= 0.0f*/, float max_depth /*= 1.0f*/)
{
SetViewport(cmdlist, x, y, width, height, min_depth, max_depth);
const int cx = std::max(x, 0);
const int cy = std::max(y, 0);
const int cwidth = width - (cx - x);
const int cheight = height - (cy - y);
SetScissor(cmdlist, cx, cy, cwidth, cheight);
}
u32 GetTexelSize(DXGI_FORMAT format)
{
switch (format)
{
case DXGI_FORMAT_R8G8B8A8_UNORM:
case DXGI_FORMAT_R8G8B8A8_SNORM:
case DXGI_FORMAT_R8G8B8A8_TYPELESS:
case DXGI_FORMAT_B8G8R8A8_UNORM:
case DXGI_FORMAT_B8G8R8A8_TYPELESS:
return 4;
case DXGI_FORMAT_B5G5R5A1_UNORM:
case DXGI_FORMAT_B5G6R5_UNORM:
return 2;
default:
Panic("Unknown format");
return 1;
}
}
void SetDefaultSampler(D3D12_SAMPLER_DESC* desc)
{
desc->Filter = D3D12_FILTER_MIN_MAG_MIP_LINEAR;
desc->AddressU = D3D12_TEXTURE_ADDRESS_MODE_CLAMP;
desc->AddressV = D3D12_TEXTURE_ADDRESS_MODE_CLAMP;
desc->AddressW = D3D12_TEXTURE_ADDRESS_MODE_CLAMP;
desc->MipLODBias = 0;
desc->MaxAnisotropy = 1;
desc->ComparisonFunc = D3D12_COMPARISON_FUNC_NEVER;
desc->BorderColor[0] = 1.0f;
desc->BorderColor[1] = 1.0f;
desc->BorderColor[2] = 1.0f;
desc->BorderColor[3] = 1.0f;
desc->MinLOD = -3.402823466e+38F; // -FLT_MAX
desc->MaxLOD = 3.402823466e+38F; // FLT_MAX
}
#ifdef _DEBUG
void SetObjectName(ID3D12Object* object, const char* name)
{
object->SetName(StringUtil::UTF8StringToWideString(name).c_str());
}
void SetObjectNameFormatted(ID3D12Object* object, const char* format, ...)
{
std::va_list ap;
va_start(ap, format);
SmallString str;
str.FormatVA(format, ap);
SetObjectName(object, str);
va_end(ap);
}
#endif
GraphicsPipelineBuilder::GraphicsPipelineBuilder()
{
Clear();
}
void GraphicsPipelineBuilder::Clear()
{
std::memset(&m_desc, 0, sizeof(m_desc));
std::memset(m_input_elements.data(), 0, sizeof(D3D12_INPUT_ELEMENT_DESC) * m_input_elements.size());
m_desc.NodeMask = 1;
m_desc.SampleMask = 0xFFFFFFFF;
m_desc.SampleDesc.Count = 1;
}
Microsoft::WRL::ComPtr<ID3D12PipelineState> GraphicsPipelineBuilder::Create(ID3D12Device* device, bool clear /*= true*/)
{
Microsoft::WRL::ComPtr<ID3D12PipelineState> ps;
HRESULT hr = device->CreateGraphicsPipelineState(&m_desc, IID_PPV_ARGS(ps.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateGraphicsPipelineState() failed: %08X", hr);
return {};
}
if (clear)
Clear();
return ps;
}
Microsoft::WRL::ComPtr<ID3D12PipelineState> GraphicsPipelineBuilder::Create(ID3D12Device* device, ShaderCache& cache,
bool clear /*= true*/)
{
Microsoft::WRL::ComPtr<ID3D12PipelineState> pso = cache.GetPipelineState(device, m_desc);
if (!pso)
return {};
if (clear)
Clear();
return pso;
}
void GraphicsPipelineBuilder::SetRootSignature(ID3D12RootSignature* rs)
{
m_desc.pRootSignature = rs;
}
void GraphicsPipelineBuilder::SetVertexShader(ID3DBlob* blob)
{
SetVertexShader(blob->GetBufferPointer(), static_cast<u32>(blob->GetBufferSize()));
}
void GraphicsPipelineBuilder::SetVertexShader(const void* data, u32 data_size)
{
m_desc.VS.pShaderBytecode = data;
m_desc.VS.BytecodeLength = data_size;
}
void GraphicsPipelineBuilder::SetGeometryShader(ID3DBlob* blob)
{
SetGeometryShader(blob->GetBufferPointer(), static_cast<u32>(blob->GetBufferSize()));
}
void GraphicsPipelineBuilder::SetGeometryShader(const void* data, u32 data_size)
{
m_desc.GS.pShaderBytecode = data;
m_desc.GS.BytecodeLength = data_size;
}
void GraphicsPipelineBuilder::SetPixelShader(ID3DBlob* blob)
{
SetPixelShader(blob->GetBufferPointer(), static_cast<u32>(blob->GetBufferSize()));
}
void GraphicsPipelineBuilder::SetPixelShader(const void* data, u32 data_size)
{
m_desc.PS.pShaderBytecode = data;
m_desc.PS.BytecodeLength = data_size;
}
void GraphicsPipelineBuilder::AddVertexAttribute(const char* semantic_name, u32 semantic_index, DXGI_FORMAT format,
u32 buffer, u32 offset)
{
const u32 index = m_desc.InputLayout.NumElements;
m_input_elements[index].SemanticIndex = semantic_index;
m_input_elements[index].SemanticName = semantic_name;
m_input_elements[index].Format = format;
m_input_elements[index].AlignedByteOffset = offset;
m_input_elements[index].InputSlot = buffer;
m_input_elements[index].InputSlotClass = D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA;
m_input_elements[index].InstanceDataStepRate = 0;
m_desc.InputLayout.pInputElementDescs = m_input_elements.data();
m_desc.InputLayout.NumElements++;
}
void GraphicsPipelineBuilder::SetPrimitiveTopologyType(D3D12_PRIMITIVE_TOPOLOGY_TYPE type)
{
m_desc.PrimitiveTopologyType = type;
}
void GraphicsPipelineBuilder::SetRasterizationState(D3D12_FILL_MODE polygon_mode, D3D12_CULL_MODE cull_mode,
bool front_face_ccw)
{
m_desc.RasterizerState.FillMode = polygon_mode;
m_desc.RasterizerState.CullMode = cull_mode;
m_desc.RasterizerState.FrontCounterClockwise = front_face_ccw;
}
void GraphicsPipelineBuilder::SetMultisamples(u32 multisamples)
{
m_desc.RasterizerState.MultisampleEnable = multisamples > 1;
m_desc.SampleDesc.Count = multisamples;
}
void GraphicsPipelineBuilder::SetNoCullRasterizationState()
{
SetRasterizationState(D3D12_FILL_MODE_SOLID, D3D12_CULL_MODE_NONE, false);
}
void GraphicsPipelineBuilder::SetDepthState(bool depth_test, bool depth_write, D3D12_COMPARISON_FUNC compare_op)
{
m_desc.DepthStencilState.DepthEnable = depth_test;
m_desc.DepthStencilState.DepthWriteMask = depth_write ? D3D12_DEPTH_WRITE_MASK_ALL : D3D12_DEPTH_WRITE_MASK_ZERO;
m_desc.DepthStencilState.DepthFunc = compare_op;
}
void GraphicsPipelineBuilder::SetNoDepthTestState()
{
SetDepthState(false, false, D3D12_COMPARISON_FUNC_ALWAYS);
}
void GraphicsPipelineBuilder::SetBlendState(u32 rt, bool blend_enable, D3D12_BLEND src_factor, D3D12_BLEND dst_factor,
D3D12_BLEND_OP op, D3D12_BLEND alpha_src_factor,
D3D12_BLEND alpha_dst_factor, D3D12_BLEND_OP alpha_op,
u8 write_mask /*= 0xFF*/)
{
m_desc.BlendState.RenderTarget[rt].BlendEnable = blend_enable;
m_desc.BlendState.RenderTarget[rt].SrcBlend = src_factor;
m_desc.BlendState.RenderTarget[rt].DestBlend = dst_factor;
m_desc.BlendState.RenderTarget[rt].BlendOp = op;
m_desc.BlendState.RenderTarget[rt].SrcBlendAlpha = alpha_src_factor;
m_desc.BlendState.RenderTarget[rt].DestBlendAlpha = alpha_dst_factor;
m_desc.BlendState.RenderTarget[rt].BlendOpAlpha = alpha_op;
m_desc.BlendState.RenderTarget[rt].RenderTargetWriteMask = write_mask;
if (rt > 0)
m_desc.BlendState.IndependentBlendEnable = TRUE;
}
void GraphicsPipelineBuilder::SetNoBlendingState()
{
SetBlendState(0, false, D3D12_BLEND_ONE, D3D12_BLEND_ZERO, D3D12_BLEND_OP_ADD, D3D12_BLEND_ONE, D3D12_BLEND_ZERO,
D3D12_BLEND_OP_ADD, D3D12_COLOR_WRITE_ENABLE_ALL);
m_desc.BlendState.IndependentBlendEnable = FALSE;
}
void GraphicsPipelineBuilder::ClearRenderTargets()
{
m_desc.NumRenderTargets = 0;
for (u32 i = 0; i < sizeof(m_desc.RTVFormats) / sizeof(m_desc.RTVFormats[0]); i++)
m_desc.RTVFormats[i] = DXGI_FORMAT_UNKNOWN;
}
void GraphicsPipelineBuilder::SetRenderTarget(u32 rt, DXGI_FORMAT format)
{
m_desc.RTVFormats[rt] = format;
if (rt >= m_desc.NumRenderTargets)
m_desc.NumRenderTargets = rt + 1;
}
void GraphicsPipelineBuilder::ClearDepthStencilFormat()
{
m_desc.DSVFormat = DXGI_FORMAT_UNKNOWN;
}
void GraphicsPipelineBuilder::SetDepthStencilFormat(DXGI_FORMAT format)
{
m_desc.DSVFormat = format;
}
RootSignatureBuilder::RootSignatureBuilder()
{
Clear();
}
void RootSignatureBuilder::Clear()
{
m_desc = {};
m_desc.pParameters = m_params.data();
m_params = {};
m_descriptor_ranges = {};
m_num_descriptor_ranges = 0;
}
Microsoft::WRL::ComPtr<ID3D12RootSignature> RootSignatureBuilder::Create(bool clear /*= true*/)
{
Microsoft::WRL::ComPtr<ID3D12RootSignature> rs = g_d3d12_context->CreateRootSignature(&m_desc);
if (!rs)
return {};
if (clear)
Clear();
return rs;
}
void RootSignatureBuilder::SetInputAssemblerFlag()
{
m_desc.Flags |= D3D12_ROOT_SIGNATURE_FLAG_ALLOW_INPUT_ASSEMBLER_INPUT_LAYOUT;
}
u32 RootSignatureBuilder::Add32BitConstants(u32 shader_reg, u32 num_values, D3D12_SHADER_VISIBILITY visibility)
{
const u32 index = m_desc.NumParameters++;
m_params[index].ParameterType = D3D12_ROOT_PARAMETER_TYPE_32BIT_CONSTANTS;
m_params[index].ShaderVisibility = visibility;
m_params[index].Constants.ShaderRegister = shader_reg;
m_params[index].Constants.RegisterSpace = 0;
m_params[index].Constants.Num32BitValues = num_values;
return index;
}
u32 RootSignatureBuilder::AddCBVParameter(u32 shader_reg, D3D12_SHADER_VISIBILITY visibility)
{
const u32 index = m_desc.NumParameters++;
m_params[index].ParameterType = D3D12_ROOT_PARAMETER_TYPE_CBV;
m_params[index].ShaderVisibility = visibility;
m_params[index].Descriptor.ShaderRegister = shader_reg;
m_params[index].Descriptor.RegisterSpace = 0;
return index;
}
u32 RootSignatureBuilder::AddSRVParameter(u32 shader_reg, D3D12_SHADER_VISIBILITY visibility)
{
const u32 index = m_desc.NumParameters++;
m_params[index].ParameterType = D3D12_ROOT_PARAMETER_TYPE_SRV;
m_params[index].ShaderVisibility = visibility;
m_params[index].Descriptor.ShaderRegister = shader_reg;
m_params[index].Descriptor.RegisterSpace = 0;
return index;
}
u32 RootSignatureBuilder::AddDescriptorTable(D3D12_DESCRIPTOR_RANGE_TYPE rt, u32 start_shader_reg, u32 num_shader_regs,
D3D12_SHADER_VISIBILITY visibility)
{
const u32 index = m_desc.NumParameters++;
const u32 dr_index = m_num_descriptor_ranges++;
m_descriptor_ranges[dr_index].RangeType = rt;
m_descriptor_ranges[dr_index].NumDescriptors = num_shader_regs;
m_descriptor_ranges[dr_index].BaseShaderRegister = start_shader_reg;
m_descriptor_ranges[dr_index].RegisterSpace = 0;
m_descriptor_ranges[dr_index].OffsetInDescriptorsFromTableStart = D3D12_DESCRIPTOR_RANGE_OFFSET_APPEND;
m_params[index].ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE;
m_params[index].DescriptorTable.pDescriptorRanges = &m_descriptor_ranges[dr_index];
m_params[index].DescriptorTable.NumDescriptorRanges = 1;
m_params[index].ShaderVisibility = visibility;
return index;
}
} // namespace D3D12

352
src/core/gpu/d3d12_builders.cpp Normal file
View File

@ -0,0 +1,352 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "d3d12_builders.h"
#include "d3d12_device.h"
#include "common/assert.h"
#include "common/log.h"
#include "common/string_util.h"
#include <cstdarg>
#include <limits>
Log_SetChannel(D3D12Device);
D3D12::GraphicsPipelineBuilder::GraphicsPipelineBuilder()
{
Clear();
}
void D3D12::GraphicsPipelineBuilder::Clear()
{
std::memset(&m_desc, 0, sizeof(m_desc));
std::memset(m_input_elements.data(), 0, sizeof(D3D12_INPUT_ELEMENT_DESC) * m_input_elements.size());
m_desc.NodeMask = 1;
m_desc.SampleMask = 0xFFFFFFFF;
m_desc.SampleDesc.Count = 1;
}
Microsoft::WRL::ComPtr<ID3D12PipelineState> D3D12::GraphicsPipelineBuilder::Create(ID3D12Device* device,
bool clear /*= true*/)
{
Microsoft::WRL::ComPtr<ID3D12PipelineState> ps;
HRESULT hr = device->CreateGraphicsPipelineState(&m_desc, IID_PPV_ARGS(ps.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateGraphicsPipelineState() failed: %08X", hr);
return {};
}
if (clear)
Clear();
return ps;
}
void D3D12::GraphicsPipelineBuilder::SetRootSignature(ID3D12RootSignature* rs)
{
m_desc.pRootSignature = rs;
}
void D3D12::GraphicsPipelineBuilder::SetVertexShader(const ID3DBlob* blob)
{
SetVertexShader(const_cast<ID3DBlob*>(blob)->GetBufferPointer(),
static_cast<u32>(const_cast<ID3DBlob*>(blob)->GetBufferSize()));
}
void D3D12::GraphicsPipelineBuilder::SetVertexShader(const void* data, u32 data_size)
{
m_desc.VS.pShaderBytecode = data;
m_desc.VS.BytecodeLength = data_size;
}
void D3D12::GraphicsPipelineBuilder::SetGeometryShader(const ID3DBlob* blob)
{
SetGeometryShader(const_cast<ID3DBlob*>(blob)->GetBufferPointer(),
static_cast<u32>(const_cast<ID3DBlob*>(blob)->GetBufferSize()));
}
void D3D12::GraphicsPipelineBuilder::SetGeometryShader(const void* data, u32 data_size)
{
m_desc.GS.pShaderBytecode = data;
m_desc.GS.BytecodeLength = data_size;
}
void D3D12::GraphicsPipelineBuilder::SetPixelShader(const ID3DBlob* blob)
{
SetPixelShader(const_cast<ID3DBlob*>(blob)->GetBufferPointer(),
static_cast<u32>(const_cast<ID3DBlob*>(blob)->GetBufferSize()));
}
void D3D12::GraphicsPipelineBuilder::SetPixelShader(const void* data, u32 data_size)
{
m_desc.PS.pShaderBytecode = data;
m_desc.PS.BytecodeLength = data_size;
}
void D3D12::GraphicsPipelineBuilder::AddVertexAttribute(const char* semantic_name, u32 semantic_index,
DXGI_FORMAT format, u32 buffer, u32 offset)
{
const u32 index = m_desc.InputLayout.NumElements;
m_input_elements[index].SemanticIndex = semantic_index;
m_input_elements[index].SemanticName = semantic_name;
m_input_elements[index].Format = format;
m_input_elements[index].AlignedByteOffset = offset;
m_input_elements[index].InputSlot = buffer;
m_input_elements[index].InputSlotClass = D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA;
m_input_elements[index].InstanceDataStepRate = 0;
m_desc.InputLayout.pInputElementDescs = m_input_elements.data();
m_desc.InputLayout.NumElements++;
}
void D3D12::GraphicsPipelineBuilder::SetPrimitiveTopologyType(D3D12_PRIMITIVE_TOPOLOGY_TYPE type)
{
m_desc.PrimitiveTopologyType = type;
}
void D3D12::GraphicsPipelineBuilder::SetRasterizationState(D3D12_FILL_MODE polygon_mode, D3D12_CULL_MODE cull_mode,
bool front_face_ccw)
{
m_desc.RasterizerState.FillMode = polygon_mode;
m_desc.RasterizerState.CullMode = cull_mode;
m_desc.RasterizerState.FrontCounterClockwise = front_face_ccw;
}
void D3D12::GraphicsPipelineBuilder::SetMultisamples(u32 multisamples)
{
m_desc.RasterizerState.MultisampleEnable = multisamples > 1;
m_desc.SampleDesc.Count = multisamples;
}
void D3D12::GraphicsPipelineBuilder::SetNoCullRasterizationState()
{
SetRasterizationState(D3D12_FILL_MODE_SOLID, D3D12_CULL_MODE_NONE, false);
}
void D3D12::GraphicsPipelineBuilder::SetDepthState(bool depth_test, bool depth_write, D3D12_COMPARISON_FUNC compare_op)
{
m_desc.DepthStencilState.DepthEnable = depth_test;
m_desc.DepthStencilState.DepthWriteMask = depth_write ? D3D12_DEPTH_WRITE_MASK_ALL : D3D12_DEPTH_WRITE_MASK_ZERO;
m_desc.DepthStencilState.DepthFunc = compare_op;
}
void D3D12::GraphicsPipelineBuilder::SetStencilState(bool stencil_test, u8 read_mask, u8 write_mask,
const D3D12_DEPTH_STENCILOP_DESC& front,
const D3D12_DEPTH_STENCILOP_DESC& back)
{
m_desc.DepthStencilState.StencilEnable = stencil_test;
m_desc.DepthStencilState.StencilReadMask = read_mask;
m_desc.DepthStencilState.StencilWriteMask = write_mask;
m_desc.DepthStencilState.FrontFace = front;
m_desc.DepthStencilState.BackFace = back;
}
void D3D12::GraphicsPipelineBuilder::SetNoDepthTestState()
{
SetDepthState(false, false, D3D12_COMPARISON_FUNC_ALWAYS);
}
void D3D12::GraphicsPipelineBuilder::SetNoStencilState()
{
D3D12_DEPTH_STENCILOP_DESC empty = {};
SetStencilState(false, 0, 0, empty, empty);
}
void D3D12::GraphicsPipelineBuilder::SetBlendState(u32 rt, bool blend_enable, D3D12_BLEND src_factor,
D3D12_BLEND dst_factor, D3D12_BLEND_OP op,
D3D12_BLEND alpha_src_factor, D3D12_BLEND alpha_dst_factor,
D3D12_BLEND_OP alpha_op, u8 write_mask /*= 0xFF*/)
{
m_desc.BlendState.RenderTarget[rt].BlendEnable = blend_enable;
m_desc.BlendState.RenderTarget[rt].SrcBlend = src_factor;
m_desc.BlendState.RenderTarget[rt].DestBlend = dst_factor;
m_desc.BlendState.RenderTarget[rt].BlendOp = op;
m_desc.BlendState.RenderTarget[rt].SrcBlendAlpha = alpha_src_factor;
m_desc.BlendState.RenderTarget[rt].DestBlendAlpha = alpha_dst_factor;
m_desc.BlendState.RenderTarget[rt].BlendOpAlpha = alpha_op;
m_desc.BlendState.RenderTarget[rt].RenderTargetWriteMask = write_mask;
if (rt > 0)
m_desc.BlendState.IndependentBlendEnable = TRUE;
}
void D3D12::GraphicsPipelineBuilder::SetColorWriteMask(u32 rt, u8 write_mask /* = D3D12_COLOR_WRITE_ENABLE_ALL */)
{
m_desc.BlendState.RenderTarget[rt].RenderTargetWriteMask = write_mask;
}
void D3D12::GraphicsPipelineBuilder::SetNoBlendingState()
{
SetBlendState(0, false, D3D12_BLEND_ONE, D3D12_BLEND_ZERO, D3D12_BLEND_OP_ADD, D3D12_BLEND_ONE, D3D12_BLEND_ZERO,
D3D12_BLEND_OP_ADD, D3D12_COLOR_WRITE_ENABLE_ALL);
m_desc.BlendState.IndependentBlendEnable = FALSE;
}
void D3D12::GraphicsPipelineBuilder::ClearRenderTargets()
{
m_desc.NumRenderTargets = 0;
for (u32 i = 0; i < sizeof(m_desc.RTVFormats) / sizeof(m_desc.RTVFormats[0]); i++)
m_desc.RTVFormats[i] = DXGI_FORMAT_UNKNOWN;
}
void D3D12::GraphicsPipelineBuilder::SetRenderTarget(u32 rt, DXGI_FORMAT format)
{
m_desc.RTVFormats[rt] = format;
if (rt >= m_desc.NumRenderTargets)
m_desc.NumRenderTargets = rt + 1;
}
void D3D12::GraphicsPipelineBuilder::ClearDepthStencilFormat()
{
m_desc.DSVFormat = DXGI_FORMAT_UNKNOWN;
}
void D3D12::GraphicsPipelineBuilder::SetDepthStencilFormat(DXGI_FORMAT format)
{
m_desc.DSVFormat = format;
}
D3D12::ComputePipelineBuilder::ComputePipelineBuilder()
{
Clear();
}
void D3D12::ComputePipelineBuilder::Clear()
{
std::memset(&m_desc, 0, sizeof(m_desc));
}
Microsoft::WRL::ComPtr<ID3D12PipelineState> D3D12::ComputePipelineBuilder::Create(ID3D12Device* device,
bool clear /*= true*/)
{
Microsoft::WRL::ComPtr<ID3D12PipelineState> ps;
HRESULT hr = device->CreateComputePipelineState(&m_desc, IID_PPV_ARGS(ps.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateComputePipelineState() failed: %08X", hr);
return {};
}
if (clear)
Clear();
return ps;
}
void D3D12::ComputePipelineBuilder::SetRootSignature(ID3D12RootSignature* rs)
{
m_desc.pRootSignature = rs;
}
void D3D12::ComputePipelineBuilder::SetShader(const void* data, u32 data_size)
{
m_desc.CS.pShaderBytecode = data;
m_desc.CS.BytecodeLength = data_size;
}
D3D12::RootSignatureBuilder::RootSignatureBuilder()
{
Clear();
}
void D3D12::RootSignatureBuilder::Clear()
{
m_desc = {};
m_desc.pParameters = m_params.data();
m_params = {};
m_descriptor_ranges = {};
m_num_descriptor_ranges = 0;
}
Microsoft::WRL::ComPtr<ID3D12RootSignature> D3D12::RootSignatureBuilder::Create(bool clear /*= true*/)
{
Microsoft::WRL::ComPtr<ID3D12RootSignature> rs = D3D12Device::GetInstance().CreateRootSignature(&m_desc);
if (!rs)
return {};
if (clear)
Clear();
return rs;
}
void D3D12::RootSignatureBuilder::SetInputAssemblerFlag()
{
m_desc.Flags |= D3D12_ROOT_SIGNATURE_FLAG_ALLOW_INPUT_ASSEMBLER_INPUT_LAYOUT;
}
u32 D3D12::RootSignatureBuilder::Add32BitConstants(u32 shader_reg, u32 num_values, D3D12_SHADER_VISIBILITY visibility)
{
const u32 index = m_desc.NumParameters++;
m_params[index].ParameterType = D3D12_ROOT_PARAMETER_TYPE_32BIT_CONSTANTS;
m_params[index].ShaderVisibility = visibility;
m_params[index].Constants.ShaderRegister = shader_reg;
m_params[index].Constants.RegisterSpace = 0;
m_params[index].Constants.Num32BitValues = num_values;
return index;
}
u32 D3D12::RootSignatureBuilder::AddCBVParameter(u32 shader_reg, D3D12_SHADER_VISIBILITY visibility)
{
const u32 index = m_desc.NumParameters++;
m_params[index].ParameterType = D3D12_ROOT_PARAMETER_TYPE_CBV;
m_params[index].ShaderVisibility = visibility;
m_params[index].Descriptor.ShaderRegister = shader_reg;
m_params[index].Descriptor.RegisterSpace = 0;
return index;
}
u32 D3D12::RootSignatureBuilder::AddSRVParameter(u32 shader_reg, D3D12_SHADER_VISIBILITY visibility)
{
const u32 index = m_desc.NumParameters++;
m_params[index].ParameterType = D3D12_ROOT_PARAMETER_TYPE_SRV;
m_params[index].ShaderVisibility = visibility;
m_params[index].Descriptor.ShaderRegister = shader_reg;
m_params[index].Descriptor.RegisterSpace = 0;
return index;
}
u32 D3D12::RootSignatureBuilder::AddDescriptorTable(D3D12_DESCRIPTOR_RANGE_TYPE rt, u32 start_shader_reg,
u32 num_shader_regs, D3D12_SHADER_VISIBILITY visibility)
{
const u32 index = m_desc.NumParameters++;
const u32 dr_index = m_num_descriptor_ranges++;
m_descriptor_ranges[dr_index].RangeType = rt;
m_descriptor_ranges[dr_index].NumDescriptors = num_shader_regs;
m_descriptor_ranges[dr_index].BaseShaderRegister = start_shader_reg;
m_descriptor_ranges[dr_index].RegisterSpace = 0;
m_descriptor_ranges[dr_index].OffsetInDescriptorsFromTableStart = D3D12_DESCRIPTOR_RANGE_OFFSET_APPEND;
m_params[index].ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE;
m_params[index].DescriptorTable.pDescriptorRanges = &m_descriptor_ranges[dr_index];
m_params[index].DescriptorTable.NumDescriptorRanges = 1;
m_params[index].ShaderVisibility = visibility;
return index;
}
#ifdef _DEBUG
void D3D12::SetObjectName(ID3D12Object* object, const std::string_view& name)
{
object->SetName(StringUtil::UTF8StringToWideString(name).c_str());
}
void D3D12::SetObjectNameFormatted(ID3D12Object* object, const char* format, ...)
{
std::va_list ap;
va_start(ap, format);
SetObjectName(object, StringUtil::StdStringFromFormatV(format, ap).c_str());
va_end(ap);
}
#endif

84
src/core/gpu/d3d12/util.h → src/core/gpu/d3d12_builders.h
View File

@ -1,47 +1,17 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "common/types.h"
#include "common/windows_headers.h"
#include <array>
#include <d3d12.h>
#include <string_view>
#include <wrl/client.h>
namespace D3D12 {
class ShaderCache;
void ResourceBarrier(ID3D12GraphicsCommandList* cmdlist, ID3D12Resource* resource, D3D12_RESOURCE_STATES from_state,
D3D12_RESOURCE_STATES to_state);
void SetViewport(ID3D12GraphicsCommandList* cmdlist, int x, int y, int width, int height, float min_depth = 0.0f,
float max_depth = 1.0f);
void SetScissor(ID3D12GraphicsCommandList* cmdlist, int x, int y, int width, int height);
void SetViewportAndScissor(ID3D12GraphicsCommandList* cmdlist, int x, int y, int width, int height,
float min_depth = 0.0f, float max_depth = 1.0f);
void SetViewportAndClampScissor(ID3D12GraphicsCommandList* cmdlist, int x, int y, int width, int height,
float min_depth = 0.0f, float max_depth = 1.0f);
u32 GetTexelSize(DXGI_FORMAT format);
void SetDefaultSampler(D3D12_SAMPLER_DESC* desc);
#ifdef _DEBUG
void SetObjectName(ID3D12Object* object, const char* name);
void SetObjectNameFormatted(ID3D12Object* object, const char* format, ...);
#else
static inline void SetObjectName(ID3D12Object* object, const char* name) {}
static inline void SetObjectNameFormatted(ID3D12Object* object, const char* format, ...) {}
#endif
class RootSignatureBuilder
{
public:
@ -86,8 +56,8 @@ public:
void Clear();
// TODO: Pipeline Cache
Microsoft::WRL::ComPtr<ID3D12PipelineState> Create(ID3D12Device* device, bool clear = true);
Microsoft::WRL::ComPtr<ID3D12PipelineState> Create(ID3D12Device* device, ShaderCache& cache, bool clear = true);
void SetRootSignature(ID3D12RootSignature* rs);
@ -95,9 +65,9 @@ public:
void SetGeometryShader(const void* data, u32 data_size);
void SetPixelShader(const void* data, u32 data_size);
void SetVertexShader(ID3DBlob* blob);
void SetGeometryShader(ID3DBlob* blob);
void SetPixelShader(ID3DBlob* blob);
void SetVertexShader(const ID3DBlob* blob);
void SetGeometryShader(const ID3DBlob* blob);
void SetPixelShader(const ID3DBlob* blob);
void AddVertexAttribute(const char* semantic_name, u32 semantic_index, DXGI_FORMAT format, u32 buffer, u32 offset);
@ -110,12 +80,16 @@ public:
void SetNoCullRasterizationState();
void SetDepthState(bool depth_test, bool depth_write, D3D12_COMPARISON_FUNC compare_op);
void SetStencilState(bool stencil_test, u8 read_mask, u8 write_mask, const D3D12_DEPTH_STENCILOP_DESC& front,
const D3D12_DEPTH_STENCILOP_DESC& back);
void SetNoDepthTestState();
void SetNoStencilState();
void SetBlendState(u32 rt, bool blend_enable, D3D12_BLEND src_factor, D3D12_BLEND dst_factor, D3D12_BLEND_OP op,
D3D12_BLEND alpha_src_factor, D3D12_BLEND alpha_dst_factor, D3D12_BLEND_OP alpha_op,
u8 write_mask = D3D12_COLOR_WRITE_ENABLE_ALL);
void SetColorWriteMask(u32 rt, u8 write_mask = D3D12_COLOR_WRITE_ENABLE_ALL);
void SetNoBlendingState();
@ -128,8 +102,38 @@ public:
void SetDepthStencilFormat(DXGI_FORMAT format);
private:
D3D12_GRAPHICS_PIPELINE_STATE_DESC m_desc{};
std::array<D3D12_INPUT_ELEMENT_DESC, MAX_VERTEX_ATTRIBUTES> m_input_elements{};
D3D12_GRAPHICS_PIPELINE_STATE_DESC m_desc;
std::array<D3D12_INPUT_ELEMENT_DESC, MAX_VERTEX_ATTRIBUTES> m_input_elements;
};
class ComputePipelineBuilder
{
public:
ComputePipelineBuilder();
~ComputePipelineBuilder() = default;
void Clear();
// TODO: Pipeline Library
Microsoft::WRL::ComPtr<ID3D12PipelineState> Create(ID3D12Device* device, bool clear = true);
void SetRootSignature(ID3D12RootSignature* rs);
void SetShader(const void* data, u32 data_size);
private:
D3D12_COMPUTE_PIPELINE_STATE_DESC m_desc;
};
#ifdef _DEBUG
void SetObjectName(ID3D12Object* object, const std::string_view& name);
void SetObjectNameFormatted(ID3D12Object* object, const char* format, ...);
#else
static inline void SetObjectName(ID3D12Object* object, const std::string_view& name)
{
}
static inline void SetObjectNameFormatted(ID3D12Object* object, const char* format, ...)
{
}
#endif
} // namespace D3D12

154
src/core/gpu/d3d12_descriptor_heap_manager.cpp Normal file
View File

@ -0,0 +1,154 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "d3d12_descriptor_heap_manager.h"
#include "common/assert.h"
#include "common/log.h"
Log_SetChannel(D3D12Device);
D3D12DescriptorHeapManager::D3D12DescriptorHeapManager() = default;
D3D12DescriptorHeapManager::~D3D12DescriptorHeapManager() = default;
bool D3D12DescriptorHeapManager::Create(ID3D12Device* device, D3D12_DESCRIPTOR_HEAP_TYPE type, u32 num_descriptors,
bool shader_visible)
{
D3D12_DESCRIPTOR_HEAP_DESC desc = {type, static_cast<UINT>(num_descriptors),
shader_visible ? D3D12_DESCRIPTOR_HEAP_FLAG_SHADER_VISIBLE :
D3D12_DESCRIPTOR_HEAP_FLAG_NONE};
HRESULT hr = device->CreateDescriptorHeap(&desc, IID_PPV_ARGS(m_descriptor_heap.ReleaseAndGetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateDescriptorHeap() failed: %08X", hr);
return false;
}
m_heap_base_cpu = m_descriptor_heap->GetCPUDescriptorHandleForHeapStart();
if (shader_visible)
m_heap_base_gpu = m_descriptor_heap->GetGPUDescriptorHandleForHeapStart();
m_num_descriptors = num_descriptors;
m_descriptor_increment_size = device->GetDescriptorHandleIncrementSize(type);
m_shader_visible = shader_visible;
// Set all slots to unallocated (1)
const u32 bitset_count = num_descriptors / BITSET_SIZE + (((num_descriptors % BITSET_SIZE) != 0) ? 1 : 0);
m_free_slots.resize(bitset_count);
for (BitSetType& bs : m_free_slots)
bs.flip();
return true;
}
void D3D12DescriptorHeapManager::Destroy()
{
for (BitSetType& bs : m_free_slots)
{
DebugAssert(bs.all());
}
m_shader_visible = false;
m_num_descriptors = 0;
m_descriptor_increment_size = 0;
m_heap_base_cpu = {};
m_heap_base_gpu = {};
m_descriptor_heap.Reset();
m_free_slots.clear();
}
bool D3D12DescriptorHeapManager::Allocate(D3D12DescriptorHandle* handle)
{
// Start past the temporary slots, no point in searching those.
for (u32 group = 0; group < m_free_slots.size(); group++)
{
BitSetType& bs = m_free_slots[group];
if (bs.none())
continue;
u32 bit = 0;
for (; bit < BITSET_SIZE; bit++)
{
if (bs[bit])
break;
}
u32 index = group * BITSET_SIZE + bit;
bs[bit] = false;
handle->index = index;
handle->cpu_handle.ptr = m_heap_base_cpu.ptr + index * m_descriptor_increment_size;
handle->gpu_handle.ptr = m_shader_visible ? (m_heap_base_gpu.ptr + index * m_descriptor_increment_size) : 0;
return true;
}
Panic("Out of fixed descriptors");
return false;
}
void D3D12DescriptorHeapManager::Free(u32 index)
{
DebugAssert(index < m_num_descriptors);
u32 group = index / BITSET_SIZE;
u32 bit = index % BITSET_SIZE;
m_free_slots[group][bit] = true;
}
void D3D12DescriptorHeapManager::Free(D3D12DescriptorHandle* handle)
{
if (handle->index == D3D12DescriptorHandle::INVALID_INDEX)
return;
Free(handle->index);
handle->Clear();
}
D3D12DescriptorAllocator::D3D12DescriptorAllocator() = default;
D3D12DescriptorAllocator::~D3D12DescriptorAllocator() = default;
bool D3D12DescriptorAllocator::Create(ID3D12Device* device, D3D12_DESCRIPTOR_HEAP_TYPE type, u32 num_descriptors)
{
const D3D12_DESCRIPTOR_HEAP_DESC desc = {type, static_cast<UINT>(num_descriptors),
D3D12_DESCRIPTOR_HEAP_FLAG_SHADER_VISIBLE};
const HRESULT hr = device->CreateDescriptorHeap(&desc, IID_PPV_ARGS(m_descriptor_heap.ReleaseAndGetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateDescriptorHeap() failed: %08X", hr);
return false;
}
m_num_descriptors = num_descriptors;
m_descriptor_increment_size = device->GetDescriptorHandleIncrementSize(type);
m_heap_base_cpu = m_descriptor_heap->GetCPUDescriptorHandleForHeapStart();
m_heap_base_gpu = m_descriptor_heap->GetGPUDescriptorHandleForHeapStart();
return true;
}
void D3D12DescriptorAllocator::Destroy()
{
m_descriptor_heap.Reset();
m_descriptor_increment_size = 0;
m_num_descriptors = 0;
m_current_offset = 0;
m_heap_base_cpu = {};
m_heap_base_gpu = {};
}
bool D3D12DescriptorAllocator::Allocate(u32 num_handles, D3D12DescriptorHandle* out_base_handle)
{
if ((m_current_offset + num_handles) > m_num_descriptors)
return false;
out_base_handle->index = m_current_offset;
out_base_handle->cpu_handle.ptr = m_heap_base_cpu.ptr + m_current_offset * m_descriptor_increment_size;
out_base_handle->gpu_handle.ptr = m_heap_base_gpu.ptr + m_current_offset * m_descriptor_increment_size;
m_current_offset += num_handles;
return true;
}
void D3D12DescriptorAllocator::Reset()
{
m_current_offset = 0;
}

244
src/core/gpu/d3d12_descriptor_heap_manager.h Normal file
View File

@ -0,0 +1,244 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "common/hash_combine.h"
#include "common/types.h"
#include "common/windows_headers.h"
#include <bitset>
#include <cstring>
#include <d3d12.h>
#include <unordered_map>
#include <vector>
#include <wrl/client.h>
// This class provides an abstraction for D3D12 descriptor heaps.
struct D3D12DescriptorHandle final
{
enum : u32
{
INVALID_INDEX = 0xFFFFFFFF
};
D3D12_CPU_DESCRIPTOR_HANDLE cpu_handle{};
D3D12_GPU_DESCRIPTOR_HANDLE gpu_handle{};
u32 index = INVALID_INDEX;
ALWAYS_INLINE operator bool() const { return index != INVALID_INDEX; }
ALWAYS_INLINE operator D3D12_CPU_DESCRIPTOR_HANDLE() const { return cpu_handle; }
ALWAYS_INLINE operator D3D12_GPU_DESCRIPTOR_HANDLE() const { return gpu_handle; }
ALWAYS_INLINE bool operator==(const D3D12DescriptorHandle& rhs) const { return (index == rhs.index); }
ALWAYS_INLINE bool operator!=(const D3D12DescriptorHandle& rhs) const { return (index != rhs.index); }
ALWAYS_INLINE bool operator<(const D3D12DescriptorHandle& rhs) const { return (index < rhs.index); }
ALWAYS_INLINE bool operator<=(const D3D12DescriptorHandle& rhs) const { return (index <= rhs.index); }
ALWAYS_INLINE bool operator>(const D3D12DescriptorHandle& rhs) const { return (index > rhs.index); }
ALWAYS_INLINE bool operator>=(const D3D12DescriptorHandle& rhs) const { return (index >= rhs.index); }
ALWAYS_INLINE void Clear()
{
cpu_handle = {};
gpu_handle = {};
index = INVALID_INDEX;
}
};
class D3D12DescriptorHeapManager final
{
public:
D3D12DescriptorHeapManager();
~D3D12DescriptorHeapManager();
ID3D12DescriptorHeap* GetDescriptorHeap() const { return m_descriptor_heap.Get(); }
u32 GetDescriptorIncrementSize() const { return m_descriptor_increment_size; }
bool Create(ID3D12Device* device, D3D12_DESCRIPTOR_HEAP_TYPE type, u32 num_descriptors, bool shader_visible);
void Destroy();
bool Allocate(D3D12DescriptorHandle* handle);
void Free(D3D12DescriptorHandle* handle);
void Free(u32 index);
private:
Microsoft::WRL::ComPtr<ID3D12DescriptorHeap> m_descriptor_heap;
u32 m_num_descriptors = 0;
u32 m_descriptor_increment_size = 0;
bool m_shader_visible = false;
D3D12_CPU_DESCRIPTOR_HANDLE m_heap_base_cpu = {};
D3D12_GPU_DESCRIPTOR_HANDLE m_heap_base_gpu = {};
static constexpr u32 BITSET_SIZE = 1024;
using BitSetType = std::bitset<BITSET_SIZE>;
std::vector<BitSetType> m_free_slots = {};
};
class D3D12DescriptorAllocator
{
public:
D3D12DescriptorAllocator();
~D3D12DescriptorAllocator();
ALWAYS_INLINE ID3D12DescriptorHeap* GetDescriptorHeap() const { return m_descriptor_heap.Get(); }
ALWAYS_INLINE u32 GetDescriptorIncrementSize() const { return m_descriptor_increment_size; }
bool Create(ID3D12Device* device, D3D12_DESCRIPTOR_HEAP_TYPE type, u32 num_descriptors);
void Destroy();
bool Allocate(u32 num_handles, D3D12DescriptorHandle* out_base_handle);
void Reset();
private:
Microsoft::WRL::ComPtr<ID3D12DescriptorHeap> m_descriptor_heap;
u32 m_descriptor_increment_size = 0;
u32 m_num_descriptors = 0;
u32 m_current_offset = 0;
D3D12_CPU_DESCRIPTOR_HANDLE m_heap_base_cpu = {};
D3D12_GPU_DESCRIPTOR_HANDLE m_heap_base_gpu = {};
};
template<u32 NumSamplers>
class D3D12GroupedSamplerAllocator : private D3D12DescriptorAllocator
{
struct Key
{
u32 idx[NumSamplers];
ALWAYS_INLINE bool operator==(const Key& rhs) const { return (std::memcmp(idx, rhs.idx, sizeof(idx)) == 0); }
ALWAYS_INLINE bool operator!=(const Key& rhs) const { return (std::memcmp(idx, rhs.idx, sizeof(idx)) != 0); }
};
struct KeyHash
{
ALWAYS_INLINE std::size_t operator()(const Key& key) const
{
size_t seed = 0;
for (u32 i : key.idx)
hash_combine(seed, i);
return seed;
}
};
public:
D3D12GroupedSamplerAllocator();
~D3D12GroupedSamplerAllocator();
using D3D12DescriptorAllocator::GetDescriptorHeap;
using D3D12DescriptorAllocator::GetDescriptorIncrementSize;
bool Create(ID3D12Device* device, u32 num_descriptors);
void Destroy();
bool LookupSingle(ID3D12Device* device, D3D12DescriptorHandle* gpu_handle, const D3D12DescriptorHandle& cpu_handle);
bool LookupGroup(ID3D12Device* device, D3D12DescriptorHandle* gpu_handle, const D3D12DescriptorHandle* cpu_handles);
// Clears cache but doesn't reset allocator.
void InvalidateCache();
void Reset();
bool ShouldReset() const;
private:
std::unordered_map<Key, D3D12DescriptorHandle, KeyHash> m_groups;
};
template<u32 NumSamplers>
D3D12GroupedSamplerAllocator<NumSamplers>::D3D12GroupedSamplerAllocator() = default;
template<u32 NumSamplers>
D3D12GroupedSamplerAllocator<NumSamplers>::~D3D12GroupedSamplerAllocator() = default;
template<u32 NumSamplers>
bool D3D12GroupedSamplerAllocator<NumSamplers>::Create(ID3D12Device* device, u32 num_descriptors)
{
return D3D12DescriptorAllocator::Create(device, D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER, num_descriptors);
}
template<u32 NumSamplers>
void D3D12GroupedSamplerAllocator<NumSamplers>::Destroy()
{
D3D12DescriptorAllocator::Destroy();
}
template<u32 NumSamplers>
void D3D12GroupedSamplerAllocator<NumSamplers>::Reset()
{
m_groups.clear();
D3D12DescriptorAllocator::Reset();
}
template<u32 NumSamplers>
void D3D12GroupedSamplerAllocator<NumSamplers>::InvalidateCache()
{
m_groups.clear();
}
template<u32 NumSamplers>
bool D3D12GroupedSamplerAllocator<NumSamplers>::LookupSingle(ID3D12Device* device, D3D12DescriptorHandle* gpu_handle,
const D3D12DescriptorHandle& cpu_handle)
{
Key key;
key.idx[0] = cpu_handle.index;
for (u32 i = 1; i < NumSamplers; i++)
key.idx[i] = 0;
auto it = m_groups.find(key);
if (it != m_groups.end())
{
*gpu_handle = it->second;
return true;
}
if (!Allocate(1, gpu_handle))
return false;
device->CopyDescriptorsSimple(1, *gpu_handle, cpu_handle, D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER);
m_groups.emplace(key, *gpu_handle);
return true;
}
template<u32 NumSamplers>
bool D3D12GroupedSamplerAllocator<NumSamplers>::LookupGroup(ID3D12Device* device, D3D12DescriptorHandle* gpu_handle,
const D3D12DescriptorHandle* cpu_handles)
{
Key key;
for (u32 i = 0; i < NumSamplers; i++)
key.idx[i] = cpu_handles[i].index;
auto it = m_groups.find(key);
if (it != m_groups.end())
{
*gpu_handle = it->second;
return true;
}
if (!Allocate(NumSamplers, gpu_handle))
return false;
D3D12_CPU_DESCRIPTOR_HANDLE dst_handle = *gpu_handle;
UINT dst_size = NumSamplers;
D3D12_CPU_DESCRIPTOR_HANDLE src_handles[NumSamplers];
UINT src_sizes[NumSamplers];
for (u32 i = 0; i < NumSamplers; i++)
{
src_handles[i] = cpu_handles[i];
src_sizes[i] = 1;
}
device->CopyDescriptors(1, &dst_handle, &dst_size, NumSamplers, src_handles, src_sizes,
D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER);
m_groups.emplace(key, *gpu_handle);
return true;
}
template<u32 NumSamplers>
bool D3D12GroupedSamplerAllocator<NumSamplers>::ShouldReset() const
{
// We only reset the sampler heap if more than half of the descriptors are used.
// This saves descriptor copying when there isn't a large number of sampler configs per frame.
return m_groups.size() >= (D3D12_MAX_SHADER_VISIBLE_SAMPLER_HEAP_SIZE / 2);
}

2028
src/core/gpu/d3d12_device.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

343
src/core/gpu/d3d12_device.h Normal file
View File

@ -0,0 +1,343 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "d3d12_descriptor_heap_manager.h"
#include "d3d12_stream_buffer.h"
#include "gpu_device.h"
#include "gpu_texture.h"
#include "common/windows_headers.h"
#include <array>
#include <atomic>
#include <condition_variable>
#include <d3d12.h>
#include <deque>
#include <dxgi1_5.h>
#include <functional>
#include <memory>
#include <mutex>
#include <string>
#include <thread>
#include <unordered_map>
#include <vector>
#include <wrl/client.h>
class D3D12Framebuffer;
class D3D12Pipeline;
class D3D12SwapChain;
class D3D12Texture;
class D3D12TextureBuffer;
namespace D3D12MA {
class Allocator;
}
class D3D12Device final : public GPUDevice
{
public:
friend D3D12Texture;
template<typename T>
using ComPtr = Microsoft::WRL::ComPtr<T>;
enum : u32
{
NUM_COMMAND_LISTS = 3,
/// Start/End timestamp queries.
NUM_TIMESTAMP_QUERIES_PER_CMDLIST = 2,
};
static void LookupNativeFormat(GPUTexture::Format format, DXGI_FORMAT* d3d_format, DXGI_FORMAT* srv_format,
DXGI_FORMAT* rtv_format, DXGI_FORMAT* dsv_format);
public:
D3D12Device();
~D3D12Device() override;
RenderAPI GetRenderAPI() const override;
bool HasSurface() const override;
bool UpdateWindow() override;
void ResizeWindow(s32 new_window_width, s32 new_window_height, float new_window_scale) override;
static AdapterAndModeList StaticGetAdapterAndModeList();
AdapterAndModeList GetAdapterAndModeList() override;
void DestroySurface() override;
std::string GetShaderCacheBaseName(const std::string_view& type) const override;
std::unique_ptr<GPUTexture> CreateTexture(u32 width, u32 height, u32 layers, u32 levels, u32 samples,
GPUTexture::Type type, GPUTexture::Format format,
const void* data = nullptr, u32 data_stride = 0,
bool dynamic = false) override;
std::unique_ptr<GPUSampler> CreateSampler(const GPUSampler::Config& config) override;
std::unique_ptr<GPUTextureBuffer> CreateTextureBuffer(GPUTextureBuffer::Format format, u32 size_in_elements) override;
bool DownloadTexture(GPUTexture* texture, u32 x, u32 y, u32 width, u32 height, void* out_data,
u32 out_data_stride) override;
bool SupportsTextureFormat(GPUTexture::Format format) const override;
void CopyTextureRegion(GPUTexture* dst, u32 dst_x, u32 dst_y, u32 dst_layer, u32 dst_level, GPUTexture* src,
u32 src_x, u32 src_y, u32 src_layer, u32 src_level, u32 width, u32 height) override;
void ResolveTextureRegion(GPUTexture* dst, u32 dst_x, u32 dst_y, u32 dst_layer, u32 dst_level, GPUTexture* src,
u32 src_x, u32 src_y, u32 src_layer, u32 src_level, u32 width, u32 height) override;
void ClearRenderTarget(GPUTexture* t, u32 c) override;
void ClearDepth(GPUTexture* t, float d) override;
void InvalidateRenderTarget(GPUTexture* t) override;
std::unique_ptr<GPUFramebuffer> CreateFramebuffer(GPUTexture* rt_or_ds, GPUTexture* ds = nullptr) override;
std::unique_ptr<GPUShader> CreateShaderFromBinary(GPUShaderStage stage, gsl::span<const u8> data) override;
std::unique_ptr<GPUShader> CreateShaderFromSource(GPUShaderStage stage, const std::string_view& source,
std::vector<u8>* out_binary = nullptr) override;
std::unique_ptr<GPUPipeline> CreatePipeline(const GPUPipeline::GraphicsConfig& config) override;
void PushDebugGroup(const char* fmt, ...) override;
void PopDebugGroup() override;
void InsertDebugMessage(const char* fmt, ...) override;
void MapVertexBuffer(u32 vertex_size, u32 vertex_count, void** map_ptr, u32* map_space,
u32* map_base_vertex) override;
void UnmapVertexBuffer(u32 vertex_size, u32 vertex_count) override;
void MapIndexBuffer(u32 index_count, DrawIndex** map_ptr, u32* map_space, u32* map_base_index) override;
void UnmapIndexBuffer(u32 used_index_count) override;
void PushUniformBuffer(const void* data, u32 data_size) override;
void* MapUniformBuffer(u32 size) override;
void UnmapUniformBuffer(u32 size) override;
void SetFramebuffer(GPUFramebuffer* fb) override;
void SetPipeline(GPUPipeline* pipeline) override;
void SetTextureSampler(u32 slot, GPUTexture* texture, GPUSampler* sampler) override;
void SetTextureBuffer(u32 slot, GPUTextureBuffer* buffer) override;
void SetViewport(s32 x, s32 y, s32 width, s32 height) override;
void SetScissor(s32 x, s32 y, s32 width, s32 height) override;
void Draw(u32 vertex_count, u32 base_vertex) override;
void DrawIndexed(u32 index_count, u32 base_index, u32 base_vertex) override;
bool SetGPUTimingEnabled(bool enabled) override;
float GetAndResetAccumulatedGPUTime() override;
void SetVSync(bool enabled) override;
bool BeginPresent(bool skip_present) override;
void EndPresent() override;
// Global state accessors
ALWAYS_INLINE static D3D12Device& GetInstance() { return *static_cast<D3D12Device*>(g_gpu_device.get()); }
ALWAYS_INLINE IDXGIAdapter1* GetAdapter() const { return m_adapter.Get(); }
ALWAYS_INLINE ID3D12Device1* GetDevice() const { return m_device.Get(); }
ALWAYS_INLINE ID3D12CommandQueue* GetCommandQueue() const { return m_command_queue.Get(); }
ALWAYS_INLINE D3D12MA::Allocator* GetAllocator() const { return m_allocator.Get(); }
void WaitForGPUIdle();
// Descriptor manager access.
D3D12DescriptorHeapManager& GetDescriptorHeapManager() { return m_descriptor_heap_manager; }
D3D12DescriptorHeapManager& GetRTVHeapManager() { return m_rtv_heap_manager; }
D3D12DescriptorHeapManager& GetDSVHeapManager() { return m_dsv_heap_manager; }
D3D12DescriptorHeapManager& GetSamplerHeapManager() { return m_sampler_heap_manager; }
const D3D12DescriptorHandle& GetNullSRVDescriptor() const { return m_null_srv_descriptor; }
// These command buffers are allocated per-frame. They are valid until the command buffer
// is submitted, after that you should call these functions again.
ALWAYS_INLINE ID3D12GraphicsCommandList4* GetCommandList() const
{
return m_command_lists[m_current_command_list].command_lists[1].Get();
}
ALWAYS_INLINE D3D12StreamBuffer& GetTextureUploadBuffer() { return m_texture_upload_buffer; }
ID3D12GraphicsCommandList4* GetInitCommandList();
// Root signature access.
ComPtr<ID3DBlob> SerializeRootSignature(const D3D12_ROOT_SIGNATURE_DESC* desc);
ComPtr<ID3D12RootSignature> CreateRootSignature(const D3D12_ROOT_SIGNATURE_DESC* desc);
/// Fence value for current command list.
u64 GetCurrentFenceValue() const { return m_current_fence_value; }
/// Last "completed" fence.
u64 GetCompletedFenceValue() const { return m_completed_fence_value; }
// Schedule a d3d12 resource for destruction later on. This will occur when the command buffer
// is next re-used, and the GPU has finished working with the specified resource.
void DeferObjectDestruction(ComPtr<ID3D12Object> resource);
void DeferResourceDestruction(ComPtr<D3D12MA::Allocation> allocation, ComPtr<ID3D12Resource> resource);
void DeferDescriptorDestruction(D3D12DescriptorHeapManager& heap, D3D12DescriptorHandle* descriptor);
// Wait for a fence to be completed.
// Also invokes callbacks for completion.
void WaitForFence(u64 fence_counter);
/// Ends any render pass, executes the command buffer, and invalidates cached state.
void SubmitCommandList(bool wait_for_completion);
void SubmitCommandList(bool wait_for_completion, const char* reason, ...);
void SubmitCommandListAndRestartRenderPass(const char* reason);
void UnbindFramebuffer(D3D12Framebuffer* fb);
void UnbindFramebuffer(D3D12Texture* tex);
void UnbindPipeline(D3D12Pipeline* pl);
void UnbindTexture(D3D12Texture* tex);
void UnbindTextureBuffer(D3D12TextureBuffer* buf);
protected:
bool CreateDevice(const std::string_view& adapter) override;
void DestroyDevice() override;
private:
enum DIRTY_FLAG : u32
{
DIRTY_FLAG_INITIAL = (1 << 0),
DIRTY_FLAG_PIPELINE_LAYOUT = (1 << 1),
DIRTY_FLAG_CONSTANT_BUFFER = (1 << 2),
DIRTY_FLAG_TEXTURES = (1 << 3),
DIRTY_FLAG_SAMPLERS = (1 << 3),
ALL_DIRTY_STATE = DIRTY_FLAG_INITIAL | DIRTY_FLAG_PIPELINE_LAYOUT | DIRTY_FLAG_CONSTANT_BUFFER |
DIRTY_FLAG_TEXTURES | DIRTY_FLAG_SAMPLERS,
};
struct CommandList
{
// [0] - Init (upload) command buffer, [1] - draw command buffer
std::array<ComPtr<ID3D12CommandAllocator>, 2> command_allocators;
std::array<ComPtr<ID3D12GraphicsCommandList4>, 2> command_lists;
D3D12DescriptorAllocator descriptor_allocator;
D3D12GroupedSamplerAllocator<MAX_TEXTURE_SAMPLERS> sampler_allocator;
u64 fence_counter = 0;
bool init_list_used = false;
bool needs_fence_wait = false;
bool has_timestamp_query = false;
};
using SamplerMap = std::unordered_map<u64, D3D12DescriptorHandle>;
static void GetAdapterAndModeList(AdapterAndModeList* ret, IDXGIFactory5* factory);
void SetFeatures();
bool CreateSwapChain();
bool CreateSwapChainRTV();
void DestroySwapChainRTVs();
void DestroySwapChain();
bool CreateCommandLists();
void DestroyCommandLists();
bool CreateRootSignatures();
void DestroyRootSignatures();
bool CreateBuffers();
void DestroyBuffers();
bool CreateDescriptorHeaps();
void DestroyDescriptorHeaps();
bool CreateTimestampQuery();
void DestroyTimestampQuery();
D3D12DescriptorHandle GetSampler(const GPUSampler::Config& config);
void DestroySamplers();
void DestroyDeferredObjects(u64 fence_value);
void RenderBlankFrame();
void MoveToNextCommandList();
bool CreateSRVDescriptor(ID3D12Resource* resource, u32 levels, DXGI_FORMAT format, D3D12DescriptorHandle* dh);
bool CreateRTVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, D3D12DescriptorHandle* dh);
bool CreateDSVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, D3D12DescriptorHandle* dh);
bool CreateUAVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, D3D12DescriptorHandle* dh);
bool CheckDownloadBufferSize(u32 required_size);
void DestroyDownloadBuffer();
/// Set dirty flags on everything to force re-bind at next draw time.
void InvalidateCachedState();
void SetVertexBuffer(ID3D12GraphicsCommandList4* cmdlist);
void SetViewport(ID3D12GraphicsCommandList4* cmdlist);
void SetScissor(ID3D12GraphicsCommandList4* cmdlist);
/// Applies any changed state.
ID3D12RootSignature* GetCurrentRootSignature() const;
void SetInitialPipelineState();
void PreDrawCheck();
void UpdateRootSignature();
template<GPUPipeline::Layout layout>
bool UpdateParametersForLayout(u32 dirty);
bool UpdateRootParameters(u32 dirty);
// Ends a render pass if we're currently in one.
// When Bind() is next called, the pass will be restarted.
void BeginRenderPass();
void BeginSwapChainRenderPass();
void EndRenderPass();
bool InRenderPass();
ComPtr<IDXGIAdapter1> m_adapter;
ComPtr<ID3D12Device1> m_device;
ComPtr<ID3D12CommandQueue> m_command_queue;
ComPtr<D3D12MA::Allocator> m_allocator;
ComPtr<ID3D12Fence> m_fence;
HANDLE m_fence_event = {};
u64 m_current_fence_value = 0;
u64 m_completed_fence_value = 0;
std::array<CommandList, NUM_COMMAND_LISTS> m_command_lists;
u32 m_current_command_list = NUM_COMMAND_LISTS - 1;
D3D_FEATURE_LEVEL m_feature_level = D3D_FEATURE_LEVEL_11_0;
ComPtr<IDXGIFactory5> m_dxgi_factory;
ComPtr<IDXGISwapChain1> m_swap_chain;
std::vector<std::pair<ComPtr<ID3D12Resource>, D3D12DescriptorHandle>> m_swap_chain_buffers;
u32 m_current_swap_chain_buffer = 0;
bool m_allow_tearing_supported = false;
bool m_using_allow_tearing = false;
bool m_is_exclusive_fullscreen = false;
D3D12DescriptorHeapManager m_descriptor_heap_manager;
D3D12DescriptorHeapManager m_rtv_heap_manager;
D3D12DescriptorHeapManager m_dsv_heap_manager;
D3D12DescriptorHeapManager m_sampler_heap_manager;
D3D12DescriptorHandle m_null_srv_descriptor;
D3D12DescriptorHandle m_point_sampler;
ComPtr<ID3D12QueryHeap> m_timestamp_query_heap;
ComPtr<ID3D12Resource> m_timestamp_query_buffer;
ComPtr<D3D12MA::Allocation> m_timestamp_query_allocation;
double m_timestamp_frequency = 0.0;
float m_accumulated_gpu_time = 0.0f;
std::deque<std::pair<u64, std::pair<D3D12MA::Allocation*, ID3D12Object*>>> m_cleanup_resources;
std::deque<std::pair<u64, std::pair<D3D12DescriptorHeapManager*, D3D12DescriptorHandle>>> m_cleanup_descriptors;
std::array<ComPtr<ID3D12RootSignature>, static_cast<u8>(GPUPipeline::Layout::MaxCount)> m_root_signatures = {};
D3D12StreamBuffer m_vertex_buffer;
D3D12StreamBuffer m_index_buffer;
D3D12StreamBuffer m_uniform_buffer;
D3D12StreamBuffer m_texture_upload_buffer;
u32 m_uniform_buffer_position = 0;
bool m_in_render_pass = false;
SamplerMap m_sampler_map;
ComPtr<D3D12MA::Allocation> m_download_buffer_allocation;
ComPtr<ID3D12Resource> m_download_buffer;
u32 m_download_buffer_size = 0;
// Which bindings/state has to be updated before the next draw.
u32 m_dirty_flags = ALL_DIRTY_STATE;
D3D12Framebuffer* m_current_framebuffer = nullptr;
D3D12Pipeline* m_current_pipeline = nullptr;
D3D12_PRIMITIVE_TOPOLOGY m_current_topology = D3D_PRIMITIVE_TOPOLOGY_UNDEFINED;
u32 m_current_vertex_stride = 0;
u32 m_current_blend_constant = 0;
GPUPipeline::Layout m_current_pipeline_layout = GPUPipeline::Layout::SingleTextureAndPushConstants;
std::array<D3D12Texture*, MAX_TEXTURE_SAMPLERS> m_current_textures = {};
std::array<D3D12DescriptorHandle, MAX_TEXTURE_SAMPLERS> m_current_samplers = {};
D3D12TextureBuffer* m_current_texture_buffer = nullptr;
Common::Rectangle<s32> m_current_viewport{0, 0, 1, 1};
Common::Rectangle<s32> m_current_scissor{0, 0, 1, 1};
};

1031
src/core/gpu/d3d12_gpu_device.cpp
View File

File diff suppressed because it is too large Load Diff

107
src/core/gpu/d3d12_gpu_device.h
View File

@ -1,107 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "common/timer.h"
#include "common/windows_headers.h"
#include "d3d12/descriptor_heap_manager.h"
#include "d3d12/staging_texture.h"
#include "d3d12/stream_buffer.h"
#include "d3d12/texture.h"
#include "gpu_device.h"
#include "postprocessing_chain.h"
#include <d3d12.h>
#include <dxgi.h>
#include <memory>
#include <string>
#include <string_view>
#include <vector>
#include <wrl/client.h>
class D3D12GPUDevice final : public GPUDevice
{
public:
template<typename T>
using ComPtr = Microsoft::WRL::ComPtr<T>;
D3D12GPUDevice();
~D3D12GPUDevice();
RenderAPI GetRenderAPI() const override;
bool HasSurface() const override;
//bool CreateDevice(const WindowInfo& wi, bool vsync) override;
//bool SetupDevice() override;
//bool MakeCurrent() override;
//bool DoneCurrent() override;
//bool ChangeWindow(const WindowInfo& new_wi) override;
//void ResizeWindow(s32 new_window_width, s32 new_window_height) override;
//bool SupportsFullscreen() const override;
//bool IsFullscreen() override;
//bool SetFullscreen(bool fullscreen, u32 width, u32 height, float refresh_rate) override;
AdapterAndModeList GetAdapterAndModeList() override;
void DestroySurface() override;
bool SetPostProcessingChain(const std::string_view& config) override;
std::unique_ptr<GPUTexture> CreateTexture(u32 width, u32 height, u32 layers, u32 levels, u32 samples,
GPUTexture::Type type, GPUTexture::Format format, const void* data,
u32 data_stride, bool dynamic = false) override;
bool DownloadTexture(GPUTexture* texture, u32 x, u32 y, u32 width, u32 height, void* out_data,
u32 out_data_stride) override;
bool SupportsTextureFormat(GPUTexture::Format format) const override;
bool GetHostRefreshRate(float* refresh_rate) override;
void SetVSync(bool enabled) override;
//bool Render(bool skip_present) override;
bool SetGPUTimingEnabled(bool enabled) override;
float GetAndResetAccumulatedGPUTime() override;
static AdapterAndModeList StaticGetAdapterAndModeList();
protected:
static AdapterAndModeList GetAdapterAndModeList(IDXGIFactory* dxgi_factory);
//virtual bool CreateResources() override;
//virtual void DestroyResources() override;
bool CreateSwapChain(const DXGI_MODE_DESC* fullscreen_mode);
bool CreateSwapChainRTV();
void DestroySwapChainRTVs();
#if 0
void RenderDisplay(ID3D12GraphicsCommandList* cmdlist, D3D12::Texture* swap_chain_buf);
void RenderSoftwareCursor(ID3D12GraphicsCommandList* cmdlist);
void RenderImGui(ID3D12GraphicsCommandList* cmdlist);
void RenderDisplay(ID3D12GraphicsCommandList* cmdlist, s32 left, s32 top, s32 width, s32 height,
D3D12::Texture* texture, s32 texture_view_x, s32 texture_view_y, s32 texture_view_width,
s32 texture_view_height, bool linear_filter);
void RenderSoftwareCursor(ID3D12GraphicsCommandList* cmdlist, s32 left, s32 top, s32 width, s32 height,
GPUTexture* texture_handle);
#endif
ComPtr<IDXGIFactory> m_dxgi_factory;
ComPtr<IDXGISwapChain> m_swap_chain;
std::vector<D3D12::Texture> m_swap_chain_buffers;
u32 m_current_swap_chain_buffer = 0;
ComPtr<ID3D12RootSignature> m_display_root_signature;
ComPtr<ID3D12PipelineState> m_display_pipeline;
ComPtr<ID3D12PipelineState> m_software_cursor_pipeline;
D3D12::DescriptorHandle m_point_sampler;
D3D12::DescriptorHandle m_linear_sampler;
D3D12::DescriptorHandle m_border_sampler;
D3D12::Texture m_display_pixels_texture;
D3D12::StagingTexture m_readback_staging_texture;
bool m_allow_tearing_supported = false;
bool m_using_allow_tearing = false;
};

199
src/core/gpu/d3d12_pipeline.cpp Normal file
View File

@ -0,0 +1,199 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "d3d12_pipeline.h"
#include "d3d12_builders.h"
#include "d3d12_device.h"
#include "d3d_common.h"
#include "common/assert.h"
#include "common/log.h"
#include <d3dcompiler.h>
Log_SetChannel(D3D12Device);
static u32 s_next_bad_shader_id = 1;
D3D12Shader::D3D12Shader(GPUShaderStage stage, Bytecode bytecode) : GPUShader(stage), m_bytecode(std::move(bytecode))
{
}
D3D12Shader::~D3D12Shader() = default;
void D3D12Shader::SetDebugName(const std::string_view& name)
{
}
std::unique_ptr<GPUShader> D3D12Device::CreateShaderFromBinary(GPUShaderStage stage, gsl::span<const u8> data)
{
// Can't do much at this point.
std::vector bytecode(data.begin(), data.end());
return std::unique_ptr<GPUShader>(new D3D12Shader(stage, std::move(bytecode)));
}
std::unique_ptr<GPUShader> D3D12Device::CreateShaderFromSource(GPUShaderStage stage, const std::string_view& source,
std::vector<u8>* out_binary /*= nullptr*/)
{
std::optional<std::vector<u8>> bytecode = D3DCommon::CompileShader(m_feature_level, m_debug_device, stage, source);
if (!bytecode.has_value())
return {};
std::unique_ptr<GPUShader> ret = CreateShaderFromBinary(stage, bytecode.value());
if (ret && out_binary)
*out_binary = std::move(bytecode.value());
return ret;
}
//////////////////////////////////////////////////////////////////////////
D3D12Pipeline::D3D12Pipeline(Microsoft::WRL::ComPtr<ID3D12PipelineState> pipeline, Layout layout,
D3D12_PRIMITIVE_TOPOLOGY topology, u32 vertex_stride, u32 blend_constants)
: GPUPipeline(), m_pipeline(std::move(pipeline)), m_layout(layout), m_topology(topology),
m_vertex_stride(vertex_stride), m_blend_constants(blend_constants),
m_blend_constants_f(GPUDevice::RGBA8ToFloat(blend_constants))
{
}
D3D12Pipeline::~D3D12Pipeline()
{
D3D12Device::GetInstance().DeferObjectDestruction(std::move(m_pipeline));
}
void D3D12Pipeline::SetDebugName(const std::string_view& name)
{
D3D12::SetObjectName(m_pipeline.Get(), name);
}
std::unique_ptr<GPUPipeline> D3D12Device::CreatePipeline(const GPUPipeline::GraphicsConfig& config)
{
static constexpr std::array<D3D12_PRIMITIVE_TOPOLOGY, static_cast<u32>(GPUPipeline::Primitive::MaxCount)> primitives =
{{
D3D_PRIMITIVE_TOPOLOGY_POINTLIST, // Points
D3D_PRIMITIVE_TOPOLOGY_LINELIST, // Lines
D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST, // Triangles
D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP, // TriangleStrips
}};
static constexpr std::array<D3D12_PRIMITIVE_TOPOLOGY_TYPE, static_cast<u32>(GPUPipeline::Primitive::MaxCount)>
primitive_types = {{
D3D12_PRIMITIVE_TOPOLOGY_TYPE_POINT, // Points
D3D12_PRIMITIVE_TOPOLOGY_TYPE_LINE, // Lines
D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE, // Triangles
D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE, // TriangleStrips
}};
static constexpr u32 MAX_COMPONENTS = 4;
static constexpr const DXGI_FORMAT
format_mapping[static_cast<u8>(GPUPipeline::VertexAttribute::Type::MaxCount)][MAX_COMPONENTS] = {
{DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_R32G32_FLOAT, DXGI_FORMAT_R32G32B32_FLOAT,
DXGI_FORMAT_R32G32B32A32_FLOAT}, // Float
{DXGI_FORMAT_R8_UINT, DXGI_FORMAT_R8G8_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R8G8B8A8_UINT}, // UInt8
{DXGI_FORMAT_R8_SINT, DXGI_FORMAT_R8G8_SINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R8G8B8A8_SINT}, // SInt8
{DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_R8G8_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R8G8B8A8_UNORM}, // UNorm8
{DXGI_FORMAT_R16_UINT, DXGI_FORMAT_R16G16_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R16G16B16A16_UINT}, // UInt16
{DXGI_FORMAT_R16_SINT, DXGI_FORMAT_R16G16_SINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R16G16B16A16_SINT}, // SInt16
{DXGI_FORMAT_R16_UNORM, DXGI_FORMAT_R16G16_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R16G16B16A16_UNORM}, // UNorm16
{DXGI_FORMAT_R32_UINT, DXGI_FORMAT_R32G32_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R32G32B32A32_UINT}, // UInt32
{DXGI_FORMAT_R32_SINT, DXGI_FORMAT_R32G32_SINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R32G32B32A32_SINT}, // SInt32
};
static constexpr std::array<D3D12_CULL_MODE, static_cast<u32>(GPUPipeline::CullMode::MaxCount)> cull_mapping = {{
D3D12_CULL_MODE_NONE, // None
D3D12_CULL_MODE_FRONT, // Front
D3D12_CULL_MODE_BACK, // Back
}};
static constexpr std::array<D3D12_COMPARISON_FUNC, static_cast<u32>(GPUPipeline::DepthFunc::MaxCount)>
compare_mapping = {{
D3D12_COMPARISON_FUNC_NEVER, // Never
D3D12_COMPARISON_FUNC_ALWAYS, // Always
D3D12_COMPARISON_FUNC_LESS, // Less
D3D12_COMPARISON_FUNC_LESS_EQUAL, // LessEqual
D3D12_COMPARISON_FUNC_GREATER, // Greater
D3D12_COMPARISON_FUNC_GREATER_EQUAL, // GreaterEqual
D3D12_COMPARISON_FUNC_EQUAL, // Equal
}};
static constexpr std::array<D3D12_BLEND, static_cast<u32>(GPUPipeline::BlendFunc::MaxCount)> blend_mapping = {{
D3D12_BLEND_ZERO, // Zero
D3D12_BLEND_ONE, // One
D3D12_BLEND_SRC_COLOR, // SrcColor
D3D12_BLEND_INV_SRC_COLOR, // InvSrcColor
D3D12_BLEND_DEST_COLOR, // DstColor
D3D12_BLEND_INV_DEST_COLOR, // InvDstColor
D3D12_BLEND_SRC_ALPHA, // SrcAlpha
D3D12_BLEND_INV_SRC_ALPHA, // InvSrcAlpha
D3D12_BLEND_SRC1_ALPHA, // SrcAlpha1
D3D12_BLEND_INV_SRC1_ALPHA, // InvSrcAlpha1
D3D12_BLEND_DEST_ALPHA, // DstAlpha
D3D12_BLEND_INV_DEST_ALPHA, // InvDstAlpha
D3D12_BLEND_BLEND_FACTOR, // ConstantColor
D3D12_BLEND_INV_BLEND_FACTOR, // InvConstantColor
}};
static constexpr std::array<D3D12_BLEND_OP, static_cast<u32>(GPUPipeline::BlendOp::MaxCount)> op_mapping = {{
D3D12_BLEND_OP_ADD, // Add
D3D12_BLEND_OP_SUBTRACT, // Subtract
D3D12_BLEND_OP_REV_SUBTRACT, // ReverseSubtract
D3D12_BLEND_OP_MIN, // Min
D3D12_BLEND_OP_MAX, // Max
}};
D3D12::GraphicsPipelineBuilder gpb;
gpb.SetRootSignature(m_root_signatures[static_cast<u8>(config.layout)].Get());
gpb.SetVertexShader(static_cast<const D3D12Shader*>(config.vertex_shader)->GetBytecodeData(),
static_cast<const D3D12Shader*>(config.vertex_shader)->GetBytecodeSize());
gpb.SetPixelShader(static_cast<const D3D12Shader*>(config.fragment_shader)->GetBytecodeData(),
static_cast<const D3D12Shader*>(config.fragment_shader)->GetBytecodeSize());
gpb.SetPrimitiveTopologyType(primitive_types[static_cast<u8>(config.primitive)]);
if (!config.input_layout.vertex_attributes.empty())
{
for (u32 i = 0; i < static_cast<u32>(config.input_layout.vertex_attributes.size()); i++)
{
const GPUPipeline::VertexAttribute& va = config.input_layout.vertex_attributes[i];
DebugAssert(va.components > 0 && va.components <= MAX_COMPONENTS);
gpb.AddVertexAttribute(
"ATTR", i, format_mapping[static_cast<u8>(va.type.GetValue())][static_cast<u8>(va.components.GetValue() - 1)],
0, va.offset);
}
}
gpb.SetRasterizationState(D3D12_FILL_MODE_SOLID,
cull_mapping[static_cast<u8>(config.rasterization.cull_mode.GetValue())], false);
if (config.samples > 1)
gpb.SetMultisamples(config.samples);
gpb.SetDepthState(config.depth.depth_test != GPUPipeline::DepthFunc::Always || config.depth.depth_write,
config.depth.depth_write, compare_mapping[static_cast<u8>(config.depth.depth_test.GetValue())]);
gpb.SetNoStencilState();
gpb.SetBlendState(0, config.blend.enable, blend_mapping[static_cast<u8>(config.blend.src_blend.GetValue())],
blend_mapping[static_cast<u8>(config.blend.dst_blend.GetValue())],
op_mapping[static_cast<u8>(config.blend.blend_op.GetValue())],
blend_mapping[static_cast<u8>(config.blend.src_alpha_blend.GetValue())],
blend_mapping[static_cast<u8>(config.blend.dst_alpha_blend.GetValue())],
op_mapping[static_cast<u8>(config.blend.alpha_blend_op.GetValue())], config.blend.write_mask);
if (config.color_format != GPUTexture::Format::Unknown)
{
DXGI_FORMAT color_format;
LookupNativeFormat(config.color_format, nullptr, nullptr, &color_format, nullptr);
gpb.SetRenderTarget(0, color_format);
}
if (config.depth_format != GPUTexture::Format::Unknown)
{
DXGI_FORMAT depth_format;
LookupNativeFormat(config.depth_format, nullptr, nullptr, nullptr, &depth_format);
gpb.SetDepthStencilFormat(depth_format);
}
/* TODO: PIPELINE CACHE */
ComPtr<ID3D12PipelineState> pipeline = gpb.Create(m_device.Get(), false);
if (!pipeline)
return {};
return std::unique_ptr<GPUPipeline>(new D3D12Pipeline(
pipeline, config.layout, primitives[static_cast<u8>(config.primitive)],
config.input_layout.vertex_attributes.empty() ? 0 : config.input_layout.vertex_stride, config.blend.constant));
}

63
src/core/gpu/d3d12_pipeline.h Normal file
View File

@ -0,0 +1,63 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "gpu_device.h"
#include "common/windows_headers.h"
#include <d3d12.h>
#include <vector>
#include <wrl/client.h>
class D3D12Device;
class D3D12Shader final : public GPUShader
{
friend D3D12Device;
public:
using Bytecode = std::vector<u8>;
~D3D12Shader() override;
ALWAYS_INLINE const Bytecode& GetBytecode() const { return m_bytecode; }
ALWAYS_INLINE D3D12_SHADER_BYTECODE GetD3DBytecode() const { return {m_bytecode.data(), m_bytecode.size()}; }
ALWAYS_INLINE const u8* GetBytecodeData() const { return m_bytecode.data(); }
ALWAYS_INLINE u32 GetBytecodeSize() const { return static_cast<u32>(m_bytecode.size()); }
void SetDebugName(const std::string_view& name) override;
private:
D3D12Shader(GPUShaderStage stage, Bytecode bytecode);
Bytecode m_bytecode;
};
class D3D12Pipeline final : public GPUPipeline
{
friend D3D12Device;
public:
~D3D12Pipeline() override;
ALWAYS_INLINE ID3D12PipelineState* GetPipeline() const { return m_pipeline.Get(); }
ALWAYS_INLINE Layout GetLayout() const { return m_layout; }
ALWAYS_INLINE D3D12_PRIMITIVE_TOPOLOGY GetTopology() const { return m_topology; }
ALWAYS_INLINE u32 GetVertexStride() const { return m_vertex_stride; }
ALWAYS_INLINE u32 GetBlendConstants() const { return m_blend_constants; }
ALWAYS_INLINE const std::array<float, 4>& GetBlendConstantsF() const { return m_blend_constants_f; }
ALWAYS_INLINE bool HasVertexStride() const { return (m_vertex_stride > 0); }
void SetDebugName(const std::string_view& name) override;
private:
D3D12Pipeline(Microsoft::WRL::ComPtr<ID3D12PipelineState> pipeline, Layout layout, D3D12_PRIMITIVE_TOPOLOGY topology,
u32 vertex_stride, u32 blend_constants);
Microsoft::WRL::ComPtr<ID3D12PipelineState> m_pipeline;
Layout m_layout;
D3D12_PRIMITIVE_TOPOLOGY m_topology;
u32 m_vertex_stride;
u32 m_blend_constants;
std::array<float, 4> m_blend_constants_f;
};

71
src/core/gpu/d3d12/stream_buffer.cpp → src/core/gpu/d3d12_stream_buffer.cpp
View File

@ -1,57 +1,67 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
// Parts originally from Dolphin Emulator, also written by myself.
#include "stream_buffer.h"
#include "d3d12_stream_buffer.h"
#include "d3d12_device.h"
#include "common/align.h"
#include "common/assert.h"
#include "common/log.h"
#include "context.h"
#include "D3D12MemAlloc.h"
#include <algorithm>
#include <functional>
Log_SetChannel(D3D12::StreamBuffer);
namespace D3D12 {
StreamBuffer::StreamBuffer() = default;
Log_SetChannel(D3D12StreamBuffer);
StreamBuffer::~StreamBuffer()
D3D12StreamBuffer::D3D12StreamBuffer() = default;
D3D12StreamBuffer::~D3D12StreamBuffer()
{
Destroy();
}
bool StreamBuffer::Create(u32 size)
bool D3D12StreamBuffer::Create(u32 size)
{
static const D3D12_HEAP_PROPERTIES heap_properties = {D3D12_HEAP_TYPE_UPLOAD};
const D3D12_RESOURCE_DESC resource_desc = {
D3D12_RESOURCE_DIMENSION_BUFFER, 0, size, 1, 1, 1, DXGI_FORMAT_UNKNOWN, {1, 0}, D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
D3D12_RESOURCE_FLAG_NONE};
Microsoft::WRL::ComPtr<ID3D12Resource> buffer;
D3D12MA::ALLOCATION_DESC allocationDesc = {};
allocationDesc.Flags = D3D12MA::ALLOCATION_FLAG_COMMITTED;
allocationDesc.HeapType = D3D12_HEAP_TYPE_UPLOAD;
HRESULT hr = g_d3d12_context->GetDevice()->CreateCommittedResource(&heap_properties, D3D12_HEAP_FLAG_NONE,
&resource_desc, D3D12_RESOURCE_STATE_GENERIC_READ,
nullptr, IID_PPV_ARGS(buffer.GetAddressOf()));
AssertMsg(SUCCEEDED(hr), "Allocate buffer");
Microsoft::WRL::ComPtr<ID3D12Resource> buffer;
Microsoft::WRL::ComPtr<D3D12MA::Allocation> allocation;
HRESULT hr = D3D12Device::GetInstance().GetAllocator()->CreateResource(
&allocationDesc, &resource_desc, D3D12_RESOURCE_STATE_GENERIC_READ, nullptr, allocation.ReleaseAndGetAddressOf(),
IID_PPV_ARGS(buffer.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateResource() failed: %08X", hr);
return false;
}
static const D3D12_RANGE read_range = {};
u8* host_pointer;
hr = buffer->Map(0, &read_range, reinterpret_cast<void**>(&host_pointer));
AssertMsg(SUCCEEDED(hr), "Map buffer");
if (FAILED(hr))
{
Log_ErrorPrintf("Map() failed: %08X", hr);
return false;
}
Destroy(true);
m_buffer = std::move(buffer);
m_allocation = std::move(allocation);
m_host_pointer = host_pointer;
m_size = size;
m_gpu_pointer = m_buffer->GetGPUVirtualAddress();
return true;
}
bool StreamBuffer::ReserveMemory(u32 num_bytes, u32 alignment)
bool D3D12StreamBuffer::ReserveMemory(u32 num_bytes, u32 alignment)
{
const u32 required_bytes = num_bytes + alignment;
@ -120,15 +130,15 @@ bool StreamBuffer::ReserveMemory(u32 num_bytes, u32 alignment)
return false;
}
void StreamBuffer::CommitMemory(u32 final_num_bytes)
void D3D12StreamBuffer::CommitMemory(u32 final_num_bytes)
{
Assert((m_current_offset + final_num_bytes) <= m_size);
Assert(final_num_bytes <= m_current_space);
DebugAssert((m_current_offset + final_num_bytes) <= m_size);
DebugAssert(final_num_bytes <= m_current_space);
m_current_offset += final_num_bytes;
m_current_space -= final_num_bytes;
}
void StreamBuffer::Destroy(bool defer)
void D3D12StreamBuffer::Destroy(bool defer)
{
if (m_host_pointer)
{
@ -138,8 +148,9 @@ void StreamBuffer::Destroy(bool defer)
}
if (m_buffer && defer)
g_d3d12_context->DeferResourceDestruction(m_buffer.Get());
D3D12Device::GetInstance().DeferResourceDestruction(std::move(m_allocation), std::move(m_buffer));
m_buffer.Reset();
m_allocation.Reset();
m_current_offset = 0;
m_current_space = 0;
@ -147,14 +158,14 @@ void StreamBuffer::Destroy(bool defer)
m_tracked_fences.clear();
}
void StreamBuffer::UpdateCurrentFencePosition()
void D3D12StreamBuffer::UpdateCurrentFencePosition()
{
// Don't create a tracking entry if the GPU is caught up with the buffer.
if (m_current_offset == m_current_gpu_position)
return;
// Has the offset changed since the last fence?
const u64 fence = g_d3d12_context->GetCurrentFenceValue();
const u64 fence = D3D12Device::GetInstance().GetCurrentFenceValue();
if (!m_tracked_fences.empty() && m_tracked_fences.back().first == fence)
{
// Still haven't executed a command buffer, so just update the offset.
@ -166,12 +177,12 @@ void StreamBuffer::UpdateCurrentFencePosition()
m_tracked_fences.emplace_back(fence, m_current_offset);
}
void StreamBuffer::UpdateGPUPosition()
void D3D12StreamBuffer::UpdateGPUPosition()
{
auto start = m_tracked_fences.begin();
auto end = start;
const u64 completed_counter = g_d3d12_context->GetCompletedFenceValue();
const u64 completed_counter = D3D12Device::GetInstance().GetCompletedFenceValue();
while (end != m_tracked_fences.end() && completed_counter >= end->first)
{
m_current_gpu_position = end->second;
@ -182,7 +193,7 @@ void StreamBuffer::UpdateGPUPosition()
m_tracked_fences.erase(start, end);
}
bool StreamBuffer::WaitForClearSpace(u32 num_bytes)
bool D3D12StreamBuffer::WaitForClearSpace(u32 num_bytes)
{
u32 new_offset = 0;
u32 new_space = 0;
@ -249,16 +260,14 @@ bool StreamBuffer::WaitForClearSpace(u32 num_bytes)
// Did any fences satisfy this condition?
// Has the command buffer been executed yet? If not, the caller should execute it.
if (iter == m_tracked_fences.end() || iter->first == g_d3d12_context->GetCurrentFenceValue())
if (iter == m_tracked_fences.end() || iter->first == D3D12Device::GetInstance().GetCurrentFenceValue())
return false;
// Wait until this fence is signaled. This will fire the callback, updating the GPU position.
g_d3d12_context->WaitForFence(iter->first);
D3D12Device::GetInstance().WaitForFence(iter->first);
m_tracked_fences.erase(m_tracked_fences.begin(), m_current_offset == iter->second ? m_tracked_fences.end() : ++iter);
m_current_offset = new_offset;
m_current_space = new_space;
m_current_gpu_position = new_gpu_position;
return true;
}
} // namespace D3D12

16
src/core/gpu/d3d12/stream_buffer.h → src/core/gpu/d3d12_stream_buffer.h
View File

@ -1,22 +1,25 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
// Parts originally from Dolphin Emulator, also written by myself.
#pragma once
#include "common/types.h"
#include "common/windows_headers.h"
#include <d3d12.h>
#include <deque>
#include <utility>
#include <wrl/client.h>
namespace D3D12 {
class StreamBuffer
namespace D3D12MA {
class Allocation;
}
class D3D12StreamBuffer
{
public:
StreamBuffer();
~StreamBuffer();
D3D12StreamBuffer();
~D3D12StreamBuffer();
bool Create(u32 size);
@ -48,11 +51,10 @@ private:
u32 m_current_gpu_position = 0;
Microsoft::WRL::ComPtr<ID3D12Resource> m_buffer;
Microsoft::WRL::ComPtr<D3D12MA::Allocation> m_allocation;
D3D12_GPU_VIRTUAL_ADDRESS m_gpu_pointer = {};
u8* m_host_pointer = nullptr;
// List of fences and the corresponding positions in the buffer
std::deque<std::pair<u64, u32>> m_tracked_fences;
};
} // namespace D3D12

941
src/core/gpu/d3d12_texture.cpp Normal file
View File

@ -0,0 +1,941 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "d3d12_texture.h"
#include "d3d12_builders.h"
#include "d3d12_device.h"
#include "common/align.h"
#include "common/assert.h"
#include "common/bitutils.h"
#include "common/log.h"
#include "common/string_util.h"
#include "D3D12MemAlloc.h"
Log_SetChannel(D3D12Device);
void D3D12Device::LookupNativeFormat(GPUTexture::Format format, DXGI_FORMAT* d3d_format, DXGI_FORMAT* srv_format,
DXGI_FORMAT* rtv_format, DXGI_FORMAT* dsv_format)
{
static constexpr std::array<std::array<DXGI_FORMAT, 4>, static_cast<int>(GPUTexture::Format::MaxCount)>
s_format_mapping = {{
{DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN}, // Unknown
{DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM,
DXGI_FORMAT_UNKNOWN}, // RGBA8
{DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM,
DXGI_FORMAT_UNKNOWN}, // BGRA8
{DXGI_FORMAT_B5G6R5_UNORM, DXGI_FORMAT_B5G6R5_UNORM, DXGI_FORMAT_B5G6R5_UNORM, DXGI_FORMAT_UNKNOWN}, // RGB565
{DXGI_FORMAT_B5G5R5A1_UNORM, DXGI_FORMAT_B5G5R5A1_UNORM, DXGI_FORMAT_B5G5R5A1_UNORM,
DXGI_FORMAT_UNKNOWN}, // RGBA5551
{DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_UNKNOWN}, // R8
{DXGI_FORMAT_R16_TYPELESS, DXGI_FORMAT_R16_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D16_UNORM}, // D16
}};
const auto& mapping = s_format_mapping[static_cast<int>(format)];
if (d3d_format)
*d3d_format = mapping[0];
if (srv_format)
*srv_format = mapping[1];
if (rtv_format)
*rtv_format = mapping[2];
if (dsv_format)
*dsv_format = mapping[3];
}
D3D12Texture::D3D12Texture(u32 width, u32 height, u32 layers, u32 levels, u32 samples, Type type, Format format,
DXGI_FORMAT dxgi_format, ComPtr<ID3D12Resource> resource,
ComPtr<D3D12MA::Allocation> allocation, const D3D12DescriptorHandle& srv_descriptor,
const D3D12DescriptorHandle& write_descriptor, const D3D12DescriptorHandle& uav_descriptor,
WriteDescriptorType wdtype, D3D12_RESOURCE_STATES resource_state)
: GPUTexture(static_cast<u16>(width), static_cast<u16>(height), static_cast<u8>(layers), static_cast<u8>(levels),
static_cast<u8>(samples), type, format),
m_resource(std::move(resource)), m_allocation(std::move(allocation)), m_srv_descriptor(srv_descriptor),
m_write_descriptor(write_descriptor), m_uav_descriptor(uav_descriptor), m_dxgi_format(dxgi_format),
m_resource_state(resource_state), m_write_descriptor_type(wdtype)
{
}
D3D12Texture::~D3D12Texture()
{
Destroy(true);
}
std::unique_ptr<GPUTexture> D3D12Device::CreateTexture(u32 width, u32 height, u32 layers, u32 levels, u32 samples,
GPUTexture::Type type, GPUTexture::Format format,
const void* data /* = nullptr */, u32 data_stride /* = 0 */,
bool dynamic /* = false */)
{
if (!GPUTexture::ValidateConfig(width, height, layers, levels, samples, type, format))
return {};
DXGI_FORMAT dxgi_format, srv_format, rtv_format, dsv_format;
LookupNativeFormat(format, &dxgi_format, &srv_format, &rtv_format, &dsv_format);
const DXGI_FORMAT uav_format = (type == GPUTexture::Type::RWTexture) ? dxgi_format : DXGI_FORMAT_UNKNOWN;
D3D12_RESOURCE_DESC desc = {};
desc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D;
desc.Width = width;
desc.Height = height;
desc.DepthOrArraySize = 1;
desc.MipLevels = static_cast<u8>(levels);
desc.Format = dxgi_format;
desc.SampleDesc.Count = 1;
desc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN;
D3D12MA::ALLOCATION_DESC allocationDesc = {};
allocationDesc.Flags = D3D12MA::ALLOCATION_FLAG_WITHIN_BUDGET;
allocationDesc.HeapType = D3D12_HEAP_TYPE_DEFAULT;
D3D12_CLEAR_VALUE optimized_clear_value = {};
D3D12_RESOURCE_STATES state;
switch (type)
{
case GPUTexture::Type::Texture:
{
desc.Flags = D3D12_RESOURCE_FLAG_NONE;
state = D3D12_RESOURCE_STATE_COPY_DEST;
}
break;
case GPUTexture::Type::RenderTarget:
{
// RT's tend to be larger, so we'll keep them committed for speed.
DebugAssert(levels == 1);
allocationDesc.Flags |= D3D12MA::ALLOCATION_FLAG_COMMITTED;
desc.Flags = D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET;
optimized_clear_value.Format = rtv_format;
state = D3D12_RESOURCE_STATE_RENDER_TARGET;
}
break;
case GPUTexture::Type::DepthStencil:
{
DebugAssert(levels == 1);
allocationDesc.Flags |= D3D12MA::ALLOCATION_FLAG_COMMITTED;
desc.Flags = D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL;
optimized_clear_value.Format = dsv_format;
state = D3D12_RESOURCE_STATE_DEPTH_WRITE;
}
break;
case GPUTexture::Type::RWTexture:
{
DebugAssert(levels == 1);
allocationDesc.Flags |= D3D12MA::ALLOCATION_FLAG_COMMITTED;
state = D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE;
}
break;
default:
return {};
}
if (uav_format != DXGI_FORMAT_UNKNOWN)
desc.Flags |= D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS;
ComPtr<ID3D12Resource> resource;
ComPtr<D3D12MA::Allocation> allocation;
HRESULT hr = m_allocator->CreateResource(
&allocationDesc, &desc, state,
(type == GPUTexture::Type::RenderTarget || type == GPUTexture::Type::DepthStencil) ? &optimized_clear_value :
nullptr,
allocation.GetAddressOf(), IID_PPV_ARGS(resource.GetAddressOf()));
if (FAILED(hr))
{
// OOM isn't fatal.
if (hr != E_OUTOFMEMORY)
Log_ErrorPrintf("Create texture failed: 0x%08X", hr);
return {};
}
D3D12DescriptorHandle srv_descriptor, write_descriptor, uav_descriptor;
D3D12Texture::WriteDescriptorType write_descriptor_type = D3D12Texture::WriteDescriptorType::None;
if (srv_format != DXGI_FORMAT_UNKNOWN)
{
if (!CreateSRVDescriptor(resource.Get(), levels, srv_format, &srv_descriptor))
return {};
}
switch (type)
{
case GPUTexture::Type::RenderTarget:
{
write_descriptor_type = D3D12Texture::WriteDescriptorType::RTV;
if (!CreateRTVDescriptor(resource.Get(), rtv_format, &write_descriptor))
{
m_descriptor_heap_manager.Free(&srv_descriptor);
return {};
}
}
break;
case GPUTexture::Type::DepthStencil:
{
write_descriptor_type = D3D12Texture::WriteDescriptorType::DSV;
if (!CreateDSVDescriptor(resource.Get(), dsv_format, &write_descriptor))
{
m_descriptor_heap_manager.Free(&srv_descriptor);
return {};
}
}
break;
default:
break;
}
if (uav_format != DXGI_FORMAT_UNKNOWN && !CreateUAVDescriptor(resource.Get(), dsv_format, &uav_descriptor))
{
m_descriptor_heap_manager.Free(&write_descriptor);
m_descriptor_heap_manager.Free(&srv_descriptor);
return {};
}
std::unique_ptr<D3D12Texture> tex(new D3D12Texture(width, height, layers, levels, samples, type, format, dxgi_format,
std::move(resource), std::move(allocation), srv_descriptor,
write_descriptor, uav_descriptor, write_descriptor_type, state));
if (data)
{
tex->Update(0, 0, width, height, data, data_stride);
tex->TransitionToState(D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE);
}
return tex;
}
bool D3D12Device::CreateSRVDescriptor(ID3D12Resource* resource, u32 levels, DXGI_FORMAT format,
D3D12DescriptorHandle* dh)
{
if (!m_descriptor_heap_manager.Allocate(dh))
{
Log_ErrorPrint("Failed to allocate SRV descriptor");
return false;
}
D3D12_SHADER_RESOURCE_VIEW_DESC desc = {format, D3D12_SRV_DIMENSION_TEXTURE2D,
D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING};
desc.Texture2D.MipLevels = levels;
m_device->CreateShaderResourceView(resource, &desc, dh->cpu_handle);
return true;
}
bool D3D12Device::CreateRTVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, D3D12DescriptorHandle* dh)
{
if (!m_rtv_heap_manager.Allocate(dh))
{
Log_ErrorPrint("Failed to allocate SRV descriptor");
return false;
}
const D3D12_RENDER_TARGET_VIEW_DESC desc = {format, D3D12_RTV_DIMENSION_TEXTURE2D};
m_device->CreateRenderTargetView(resource, &desc, dh->cpu_handle);
return true;
}
bool D3D12Device::CreateDSVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, D3D12DescriptorHandle* dh)
{
if (!m_dsv_heap_manager.Allocate(dh))
{
Log_ErrorPrint("Failed to allocate SRV descriptor");
return false;
}
const D3D12_DEPTH_STENCIL_VIEW_DESC desc = {format, D3D12_DSV_DIMENSION_TEXTURE2D, D3D12_DSV_FLAG_NONE};
m_device->CreateDepthStencilView(resource, &desc, dh->cpu_handle);
return true;
}
bool D3D12Device::CreateUAVDescriptor(ID3D12Resource* resource, DXGI_FORMAT format, D3D12DescriptorHandle* dh)
{
if (!m_descriptor_heap_manager.Allocate(dh))
{
Log_ErrorPrint("Failed to allocate UAV descriptor");
return false;
}
const D3D12_UNORDERED_ACCESS_VIEW_DESC desc = {format, D3D12_UAV_DIMENSION_TEXTURE2D};
m_device->CreateUnorderedAccessView(resource, nullptr, &desc, dh->cpu_handle);
return true;
}
void D3D12Texture::Destroy(bool defer)
{
D3D12Device& dev = D3D12Device::GetInstance();
dev.UnbindTexture(this);
if (defer)
{
dev.DeferDescriptorDestruction(dev.GetDescriptorHeapManager(), &m_srv_descriptor);
switch (m_write_descriptor_type)
{
case WriteDescriptorType::RTV:
dev.DeferDescriptorDestruction(dev.GetRTVHeapManager(), &m_write_descriptor);
break;
case WriteDescriptorType::DSV:
dev.DeferDescriptorDestruction(dev.GetDSVHeapManager(), &m_write_descriptor);
break;
case WriteDescriptorType::None:
default:
break;
}
if (m_uav_descriptor)
dev.DeferDescriptorDestruction(dev.GetDescriptorHeapManager(), &m_uav_descriptor);
dev.DeferResourceDestruction(std::move(m_allocation), std::move(m_resource));
}
else
{
dev.GetDescriptorHeapManager().Free(&m_srv_descriptor);
switch (m_write_descriptor_type)
{
case WriteDescriptorType::RTV:
dev.GetRTVHeapManager().Free(&m_write_descriptor);
break;
case WriteDescriptorType::DSV:
dev.GetDSVHeapManager().Free(&m_write_descriptor);
break;
case WriteDescriptorType::None:
default:
break;
}
if (m_uav_descriptor)
dev.GetDescriptorHeapManager().Free(&m_uav_descriptor);
m_resource.Reset();
m_allocation.Reset();
}
m_write_descriptor_type = WriteDescriptorType::None;
}
ID3D12GraphicsCommandList4* D3D12Texture::GetCommandBufferForUpdate()
{
D3D12Device& dev = D3D12Device::GetInstance();
if (m_type != Type::Texture || m_use_fence_counter == dev.GetCurrentFenceValue())
{
// Console.WriteLn("Texture update within frame, can't use do beforehand");
dev.EndRenderPass();
return dev.GetCommandList();
}
return dev.GetInitCommandList();
}
void D3D12Texture::CopyTextureDataForUpload(void* dst, const void* src, u32 width, u32 height, u32 pitch,
u32 upload_pitch) const
{
StringUtil::StrideMemCpy(dst, upload_pitch, src, pitch, GetPixelSize() * width, height);
}
ID3D12Resource* D3D12Texture::AllocateUploadStagingBuffer(const void* data, u32 pitch, u32 upload_pitch, u32 width,
u32 height) const
{
const u32 size = upload_pitch * height;
ComPtr<ID3D12Resource> resource;
ComPtr<D3D12MA::Allocation> allocation;
const D3D12MA::ALLOCATION_DESC allocation_desc = {D3D12MA::ALLOCATION_FLAG_NONE, D3D12_HEAP_TYPE_UPLOAD};
const D3D12_RESOURCE_DESC resource_desc = {
D3D12_RESOURCE_DIMENSION_BUFFER, 0, size, 1, 1, 1, DXGI_FORMAT_UNKNOWN, {1, 0}, D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
D3D12_RESOURCE_FLAG_NONE};
HRESULT hr = D3D12Device::GetInstance().GetAllocator()->CreateResource(
&allocation_desc, &resource_desc, D3D12_RESOURCE_STATE_GENERIC_READ, nullptr, allocation.GetAddressOf(),
IID_PPV_ARGS(resource.GetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateResource() failed with %08X", hr);
return nullptr;
}
void* map_ptr;
hr = resource->Map(0, nullptr, &map_ptr);
if (FAILED(hr))
{
Log_ErrorPrintf("Map() failed with %08X", hr);
return nullptr;
}
CopyTextureDataForUpload(map_ptr, data, width, height, pitch, upload_pitch);
const D3D12_RANGE write_range = {0, size};
resource->Unmap(0, &write_range);
// Immediately queue it for freeing after the command buffer finishes, since it's only needed for the copy.
// This adds the reference needed to keep the buffer alive.
ID3D12Resource* ret = resource.Get();
D3D12Device::GetInstance().DeferResourceDestruction(std::move(allocation), std::move(resource));
return ret;
}
bool D3D12Texture::Update(u32 x, u32 y, u32 width, u32 height, const void* data, u32 pitch, u32 layer, u32 level)
{
DebugAssert(layer < m_layers && level < m_levels);
DebugAssert((x + width) <= GetMipWidth(level) && (y + height) <= GetMipHeight(level));
D3D12Device& dev = D3D12Device::GetInstance();
D3D12StreamBuffer& sbuffer = dev.GetTextureUploadBuffer();
const u32 upload_pitch = Common::AlignUpPow2<u32>(pitch, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const u32 required_size = height * upload_pitch;
D3D12_TEXTURE_COPY_LOCATION srcloc;
srcloc.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT;
srcloc.PlacedFootprint.Footprint.Width = width;
srcloc.PlacedFootprint.Footprint.Height = height;
srcloc.PlacedFootprint.Footprint.Depth = 1;
srcloc.PlacedFootprint.Footprint.Format = m_dxgi_format;
srcloc.PlacedFootprint.Footprint.RowPitch = upload_pitch;
// If the texture is larger than half our streaming buffer size, use a separate buffer.
// Otherwise allocation will either fail, or require lots of cmdbuffer submissions.
if (required_size > (sbuffer.GetSize() / 2))
{
srcloc.pResource = AllocateUploadStagingBuffer(data, pitch, upload_pitch, width, height);
if (!srcloc.pResource)
return false;
srcloc.PlacedFootprint.Offset = 0;
}
else
{
if (!sbuffer.ReserveMemory(required_size, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT))
{
D3D12Device::GetInstance().SubmitCommandList(false, "While waiting for %u bytes in texture upload buffer",
required_size);
if (!sbuffer.ReserveMemory(required_size, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT))
{
Log_ErrorPrintf("Failed to reserve texture upload memory (%u bytes).", required_size);
return false;
}
}
srcloc.pResource = sbuffer.GetBuffer();
srcloc.PlacedFootprint.Offset = sbuffer.GetCurrentOffset();
CopyTextureDataForUpload(sbuffer.GetCurrentHostPointer(), data, width, height, pitch, upload_pitch);
sbuffer.CommitMemory(required_size);
}
ID3D12GraphicsCommandList4* cmdlist = GetCommandBufferForUpdate();
// if we're an rt and have been cleared, and the full rect isn't being uploaded, do the clear
if (m_type == Type::RenderTarget)
{
if (x != 0 || y != 0 || width != m_width || height != m_height)
CommitClear(cmdlist);
else
m_state = State::Dirty;
}
// first time the texture is used? don't leave it undefined
if (m_resource_state == D3D12_RESOURCE_STATE_COMMON)
TransitionToState(cmdlist, D3D12_RESOURCE_STATE_COPY_DEST);
else if (m_resource_state != D3D12_RESOURCE_STATE_COPY_DEST)
TransitionSubresourceToState(cmdlist, layer, level, m_resource_state, D3D12_RESOURCE_STATE_COPY_DEST);
D3D12_TEXTURE_COPY_LOCATION dstloc;
dstloc.pResource = m_resource.Get();
dstloc.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX;
dstloc.SubresourceIndex = layer;
const D3D12_BOX srcbox{0u, 0u, 0u, width, height, 1u};
cmdlist->CopyTextureRegion(&dstloc, x, y, 0, &srcloc, &srcbox);
if (m_resource_state != D3D12_RESOURCE_STATE_COPY_DEST)
TransitionSubresourceToState(cmdlist, layer, level, D3D12_RESOURCE_STATE_COPY_DEST, m_resource_state);
return true;
}
bool D3D12Texture::Map(void** map, u32* map_stride, u32 x, u32 y, u32 width, u32 height, u32 layer, u32 level)
{
// TODO: linear textures for dynamic?
if ((x + width) > GetMipWidth(level) || (y + height) > GetMipHeight(level) || layer > m_layers || level > m_levels)
{
return false;
}
D3D12Device& dev = D3D12Device::GetInstance();
if (m_state == State::Cleared && (x != 0 || y != 0 || width != m_width || height != m_height))
CommitClear(GetCommandBufferForUpdate());
// see note in Update() for the reason why.
const u32 aligned_pitch = Common::AlignUpPow2(width * GetPixelSize(), D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const u32 req_size = height * aligned_pitch;
D3D12StreamBuffer& buffer = dev.GetTextureUploadBuffer();
if (req_size >= (buffer.GetSize() / 2))
return false;
if (!buffer.ReserveMemory(req_size, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT))
{
dev.SubmitCommandList(false, "While waiting for %u bytes in texture upload buffer", req_size);
if (!buffer.ReserveMemory(req_size, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT))
Panic("Failed to reserve texture upload memory");
}
// map for writing
*map = buffer.GetCurrentHostPointer();
*map_stride = aligned_pitch;
m_map_x = static_cast<u16>(x);
m_map_y = static_cast<u16>(y);
m_map_width = static_cast<u16>(width);
m_map_height = static_cast<u16>(height);
m_map_layer = static_cast<u8>(layer);
m_map_level = static_cast<u8>(level);
m_state = State::Dirty;
return true;
}
void D3D12Texture::Unmap()
{
D3D12Device& dev = D3D12Device::GetInstance();
D3D12StreamBuffer& sb = dev.GetTextureUploadBuffer();
const u32 aligned_pitch = Common::AlignUpPow2(m_map_width * GetPixelSize(), D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const u32 req_size = m_map_height * aligned_pitch;
const u32 offset = sb.GetCurrentOffset();
sb.CommitMemory(req_size);
ID3D12GraphicsCommandList4* cmdlist = GetCommandBufferForUpdate();
// first time the texture is used? don't leave it undefined
if (m_resource_state == D3D12_RESOURCE_STATE_COMMON)
TransitionToState(cmdlist, D3D12_RESOURCE_STATE_COPY_DEST);
else if (m_resource_state != D3D12_RESOURCE_STATE_COPY_DEST)
TransitionSubresourceToState(cmdlist, m_map_layer, m_map_level, m_resource_state, D3D12_RESOURCE_STATE_COPY_DEST);
D3D12_TEXTURE_COPY_LOCATION srcloc;
srcloc.pResource = sb.GetBuffer();
srcloc.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT;
srcloc.PlacedFootprint.Offset = offset;
srcloc.PlacedFootprint.Footprint.Width = m_map_width;
srcloc.PlacedFootprint.Footprint.Height = m_map_height;
srcloc.PlacedFootprint.Footprint.Depth = 1;
srcloc.PlacedFootprint.Footprint.Format = m_dxgi_format;
srcloc.PlacedFootprint.Footprint.RowPitch = aligned_pitch;
D3D12_TEXTURE_COPY_LOCATION dstloc;
dstloc.pResource = m_resource.Get();
dstloc.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX;
dstloc.SubresourceIndex = m_map_level;
const D3D12_BOX srcbox{0u, 0u, 0u, m_map_width, m_map_height, 1};
cmdlist->CopyTextureRegion(&dstloc, m_map_x, m_map_y, 0, &srcloc, &srcbox);
if (m_resource_state != D3D12_RESOURCE_STATE_COPY_DEST)
TransitionSubresourceToState(cmdlist, m_map_layer, m_map_level, D3D12_RESOURCE_STATE_COPY_DEST, m_resource_state);
m_map_x = 0;
m_map_y = 0;
m_map_width = 0;
m_map_height = 0;
m_map_layer = 0;
m_map_level = 0;
}
void D3D12Texture::CommitClear()
{
if (m_state != GPUTexture::State::Cleared)
return;
D3D12Device& dev = D3D12Device::GetInstance();
dev.EndRenderPass();
CommitClear(dev.GetCommandList());
}
void D3D12Texture::CommitClear(ID3D12GraphicsCommandList* cmdlist)
{
if (IsDepthStencil())
{
TransitionToState(cmdlist, D3D12_RESOURCE_STATE_DEPTH_WRITE);
cmdlist->ClearDepthStencilView(GetWriteDescriptor(), D3D12_CLEAR_FLAG_DEPTH, m_clear_value.depth, 0, 0, nullptr);
}
else
{
TransitionToState(cmdlist, D3D12_RESOURCE_STATE_RENDER_TARGET);
cmdlist->ClearRenderTargetView(GetWriteDescriptor(), D3D12Device::RGBA8ToFloat(m_clear_value.color).data(), 0,
nullptr);
}
SetState(State::Dirty);
}
void D3D12Texture::SetDebugName(const std::string_view& name)
{
D3D12::SetObjectName(m_resource.Get(), name);
}
u32 D3D12Texture::CalculateSubresource(u32 layer, u32 level, u32 num_levels)
{
// D3D11CalcSubresource
return level + layer * num_levels;
}
u32 D3D12Texture::CalculateSubresource(u32 layer, u32 level) const
{
return CalculateSubresource(layer, level, m_levels);
}
void D3D12Texture::TransitionToState(D3D12_RESOURCE_STATES state)
{
TransitionToState(D3D12Device::GetInstance().GetCommandList(), state);
}
void D3D12Texture::TransitionToState(ID3D12GraphicsCommandList* cmdlist, D3D12_RESOURCE_STATES state)
{
if (m_resource_state == state)
return;
const D3D12_RESOURCE_STATES prev_state = m_resource_state;
m_resource_state = state;
const D3D12_RESOURCE_BARRIER barrier = {
D3D12_RESOURCE_BARRIER_TYPE_TRANSITION,
D3D12_RESOURCE_BARRIER_FLAG_NONE,
{{m_resource.Get(), D3D12_RESOURCE_BARRIER_ALL_SUBRESOURCES, prev_state, state}}};
cmdlist->ResourceBarrier(1, &barrier);
}
void D3D12Texture::TransitionSubresourceToState(ID3D12GraphicsCommandList* cmdlist, u32 layer, u32 level,
D3D12_RESOURCE_STATES before_state,
D3D12_RESOURCE_STATES after_state) const
{
TransitionSubresourceToState(cmdlist, m_resource.Get(), CalculateSubresource(layer, level), before_state,
after_state);
}
void D3D12Texture::TransitionSubresourceToState(ID3D12GraphicsCommandList* cmdlist, u32 subresource,
D3D12_RESOURCE_STATES before_state,
D3D12_RESOURCE_STATES after_state) const
{
TransitionSubresourceToState(cmdlist, m_resource.Get(), subresource, before_state, after_state);
}
void D3D12Texture::TransitionSubresourceToState(ID3D12GraphicsCommandList* cmdlist, ID3D12Resource* resource,
u32 subresource, D3D12_RESOURCE_STATES before_state,
D3D12_RESOURCE_STATES after_state)
{
const D3D12_RESOURCE_BARRIER barrier = {D3D12_RESOURCE_BARRIER_TYPE_TRANSITION,
D3D12_RESOURCE_BARRIER_FLAG_NONE,
{{resource, subresource, before_state, after_state}}};
cmdlist->ResourceBarrier(1, &barrier);
}
void D3D12Texture::MakeReadyForSampling()
{
if (m_resource_state == D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE)
return;
D3D12Device& dev = D3D12Device::GetInstance();
if (dev.InRenderPass())
dev.EndRenderPass();
TransitionToState(D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE);
}
bool D3D12Device::DownloadTexture(GPUTexture* texture, u32 x, u32 y, u32 width, u32 height, void* out_data,
u32 out_data_stride)
{
D3D12Texture* T = static_cast<D3D12Texture*>(texture);
T->CommitClear();
const u32 pitch = Common::AlignUp(width * T->GetPixelSize(), D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const u32 size = pitch * height;
const u32 subresource = 0;
if (!CheckDownloadBufferSize(size))
{
Log_ErrorPrintf("Can't read back %ux%u", width, height);
return false;
}
if (InRenderPass())
EndRenderPass();
ID3D12GraphicsCommandList4* cmdlist = GetCommandList();
D3D12_TEXTURE_COPY_LOCATION srcloc;
srcloc.pResource = T->GetResource();
srcloc.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX;
srcloc.SubresourceIndex = subresource;
D3D12_TEXTURE_COPY_LOCATION dstloc;
dstloc.pResource = m_download_buffer.Get();
dstloc.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT;
dstloc.PlacedFootprint.Offset = 0;
dstloc.PlacedFootprint.Footprint.Format = T->GetDXGIFormat();
dstloc.PlacedFootprint.Footprint.Width = width;
dstloc.PlacedFootprint.Footprint.Height = height;
dstloc.PlacedFootprint.Footprint.Depth = 1;
dstloc.PlacedFootprint.Footprint.RowPitch = pitch;
const D3D12_RESOURCE_STATES old_layout = T->GetResourceState();
if (old_layout != D3D12_RESOURCE_STATE_COPY_SOURCE)
T->TransitionSubresourceToState(cmdlist, subresource, old_layout, D3D12_RESOURCE_STATE_COPY_SOURCE);
// TODO: Rules for depth buffers here?
const D3D12_BOX srcbox{static_cast<UINT>(x), static_cast<UINT>(y), 0u,
static_cast<UINT>(x + width), static_cast<UINT>(y + height), 1u};
cmdlist->CopyTextureRegion(&dstloc, 0, 0, 0, &srcloc, &srcbox);
if (old_layout != D3D12_RESOURCE_STATE_COPY_SOURCE)
T->TransitionSubresourceToState(cmdlist, subresource, D3D12_RESOURCE_STATE_COPY_SOURCE, old_layout);
SubmitCommandList(true);
u8* map_pointer;
const D3D12_RANGE read_range{0u, size};
const HRESULT hr = m_download_buffer->Map(0, &read_range, reinterpret_cast<void**>(const_cast<u8**>(&map_pointer)));
if (FAILED(hr))
{
Log_ErrorPrintf("Map() failed with HRESULT %08X", hr);
return false;
}
StringUtil::StrideMemCpy(out_data, out_data_stride, map_pointer, pitch, width * T->GetPixelSize(), height);
m_download_buffer->Unmap(0, nullptr);
return true;
}
bool D3D12Device::CheckDownloadBufferSize(u32 required_size)
{
if (m_download_buffer_size >= required_size)
return true;
DestroyDownloadBuffer();
D3D12MA::ALLOCATION_DESC allocation_desc = {};
allocation_desc.HeapType = D3D12_HEAP_TYPE_READBACK;
const D3D12_RESOURCE_DESC resource_desc = {D3D12_RESOURCE_DIMENSION_BUFFER,
0,
required_size,
1,
1,
1,
DXGI_FORMAT_UNKNOWN,
{1, 0},
D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
D3D12_RESOURCE_FLAG_NONE};
HRESULT hr = m_allocator->CreateResource(&allocation_desc, &resource_desc, D3D12_RESOURCE_STATE_COPY_DEST, nullptr,
m_download_buffer_allocation.ReleaseAndGetAddressOf(),
IID_PPV_ARGS(m_download_buffer.ReleaseAndGetAddressOf()));
if (FAILED(hr))
{
Log_ErrorPrintf("CreateResource() failed with HRESULT %08X", hr);
return false;
}
return true;
}
void D3D12Device::DestroyDownloadBuffer()
{
if (!m_download_buffer)
return;
m_download_buffer.Reset();
m_download_buffer_allocation.Reset();
m_download_buffer_size = 0;
}
D3D12Sampler::D3D12Sampler(D3D12DescriptorHandle descriptor) : m_descriptor(descriptor)
{
}
D3D12Sampler::~D3D12Sampler()
{
// Cleaned up by main class.
}
void D3D12Sampler::SetDebugName(const std::string_view& name)
{
}
D3D12DescriptorHandle D3D12Device::GetSampler(const GPUSampler::Config& config)
{
const auto it = m_sampler_map.find(config.key);
if (it != m_sampler_map.end())
return it->second;
static constexpr std::array<D3D12_TEXTURE_ADDRESS_MODE, static_cast<u8>(GPUSampler::AddressMode::MaxCount)> ta = {{
D3D12_TEXTURE_ADDRESS_MODE_WRAP, // Repeat
D3D12_TEXTURE_ADDRESS_MODE_CLAMP, // ClampToEdge
D3D12_TEXTURE_ADDRESS_MODE_BORDER, // ClampToBorder
}};
static constexpr u8 filter_count = static_cast<u8>(GPUSampler::Filter::MaxCount);
static constexpr D3D12_FILTER filters[filter_count][filter_count][filter_count] = {
{
{D3D12_FILTER_MIN_MAG_MIP_POINT, D3D12_FILTER_MIN_POINT_MAG_LINEAR_MIP_POINT},
{D3D12_FILTER_MIN_LINEAR_MAG_MIP_POINT, D3D12_FILTER_MIN_MAG_LINEAR_MIP_POINT},
},
{
{D3D12_FILTER_MIN_MAG_POINT_MIP_LINEAR, D3D12_FILTER_MIN_POINT_MAG_MIP_LINEAR},
{D3D12_FILTER_MIN_LINEAR_MAG_POINT_MIP_LINEAR, D3D12_FILTER_MIN_MAG_MIP_LINEAR},
}};
D3D12_SAMPLER_DESC desc = {};
desc.AddressU = ta[static_cast<u8>(config.address_u.GetValue())];
desc.AddressV = ta[static_cast<u8>(config.address_v.GetValue())];
desc.AddressW = ta[static_cast<u8>(config.address_w.GetValue())];
std::memcpy(desc.BorderColor, RGBA8ToFloat(config.border_color).data(), sizeof(desc.BorderColor));
desc.MinLOD = static_cast<float>(config.min_lod);
desc.MaxLOD = static_cast<float>(config.max_lod);
if (config.anisotropy > 0)
{
desc.Filter = D3D12_FILTER_ANISOTROPIC;
desc.MaxAnisotropy = config.anisotropy;
}
else
{
desc.Filter = filters[static_cast<u8>(config.mip_filter.GetValue())][static_cast<u8>(config.min_filter.GetValue())]
[static_cast<u8>(config.mag_filter.GetValue())];
desc.MaxAnisotropy = 1;
}
// TODO: Allocate
D3D12DescriptorHandle handle;
if (m_sampler_heap_manager.Allocate(&handle))
m_device->CreateSampler(&desc, handle);
m_sampler_map.emplace(config.key, handle);
return handle;
}
void D3D12Device::DestroySamplers()
{
for (auto& it : m_sampler_map)
{
if (it.second)
m_sampler_heap_manager.Free(&it.second);
}
m_sampler_map.clear();
}
std::unique_ptr<GPUSampler> D3D12Device::CreateSampler(const GPUSampler::Config& config)
{
const D3D12DescriptorHandle handle = GetSampler(config);
if (!handle)
return {};
return std::unique_ptr<GPUSampler>(new D3D12Sampler(std::move(handle)));
}
D3D12Framebuffer::D3D12Framebuffer(GPUTexture* rt, GPUTexture* ds, u32 width, u32 height, D3D12DescriptorHandle rtv,
D3D12DescriptorHandle dsv)
: GPUFramebuffer(rt, ds, width, height), m_rtv(std::move(rtv)), m_dsv(std::move(dsv))
{
}
D3D12Framebuffer::~D3D12Framebuffer()
{
D3D12Device& dev = D3D12Device::GetInstance();
if (m_rtv)
D3D12Device::GetInstance().DeferDescriptorDestruction(dev.GetRTVHeapManager(), &m_rtv);
if (m_dsv)
D3D12Device::GetInstance().DeferDescriptorDestruction(dev.GetDSVHeapManager(), &m_dsv);
}
void D3D12Framebuffer::SetDebugName(const std::string_view& name)
{
}
std::unique_ptr<GPUFramebuffer> D3D12Device::CreateFramebuffer(GPUTexture* rt_or_ds, GPUTexture* ds /*= nullptr*/)
{
DebugAssert((rt_or_ds || ds) && (!rt_or_ds || rt_or_ds->IsRenderTarget() || (rt_or_ds->IsDepthStencil() && !ds)));
D3D12Texture* RT = static_cast<D3D12Texture*>((rt_or_ds && rt_or_ds->IsDepthStencil()) ? nullptr : rt_or_ds);
D3D12Texture* DS = static_cast<D3D12Texture*>((rt_or_ds && rt_or_ds->IsDepthStencil()) ? rt_or_ds : ds);
const u32 width = RT ? RT->GetWidth() : DS->GetWidth();
const u32 height = RT ? RT->GetHeight() : DS->GetHeight();
D3D12DescriptorHandle rtv, dsv;
if (RT)
rtv = RT->GetWriteDescriptor();
if (DS)
dsv = DS->GetWriteDescriptor();
return std::unique_ptr<GPUFramebuffer>(new D3D12Framebuffer(RT, DS, width, height, std::move(rtv), std::move(dsv)));
}
D3D12TextureBuffer::D3D12TextureBuffer(Format format, u32 size_in_elements) : GPUTextureBuffer(format, size_in_elements)
{
}
D3D12TextureBuffer::~D3D12TextureBuffer()
{
Destroy(true);
}
bool D3D12TextureBuffer::Create(D3D12Device& dev)
{
static constexpr std::array<DXGI_FORMAT, static_cast<u8>(GPUTextureBuffer::Format::MaxCount)> format_mapping = {{
DXGI_FORMAT_R16_UINT, // R16UI
}};
if (!m_buffer.Create(GetSizeInBytes()))
return false;
if (!dev.GetDescriptorHeapManager().Allocate(&m_descriptor))
return {};
D3D12_SHADER_RESOURCE_VIEW_DESC desc = {format_mapping[static_cast<u8>(m_format)], D3D12_SRV_DIMENSION_BUFFER,
D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING};
desc.Buffer.NumElements = m_size_in_elements;
dev.GetDevice()->CreateShaderResourceView(m_buffer.GetBuffer(), &desc, m_descriptor);
return true;
}
void D3D12TextureBuffer::Destroy(bool defer)
{
D3D12Device& dev = D3D12Device::GetInstance();
if (m_descriptor)
{
if (defer)
dev.DeferDescriptorDestruction(dev.GetDescriptorHeapManager(), &m_descriptor);
else
dev.GetDescriptorHeapManager().Free(&m_descriptor);
}
}
void* D3D12TextureBuffer::Map(u32 required_elements)
{
const u32 esize = GetElementSize(m_format);
const u32 req_size = esize * required_elements;
if (!m_buffer.ReserveMemory(req_size, esize))
{
D3D12Device::GetInstance().SubmitCommandListAndRestartRenderPass("out of space in texture buffer");
if (!m_buffer.ReserveMemory(req_size, esize))
Panic("Failed to allocate texture buffer space.");
}
m_current_position = m_buffer.GetCurrentOffset() / esize;
return m_buffer.GetCurrentHostPointer();
}
void D3D12TextureBuffer::Unmap(u32 used_elements)
{
m_buffer.CommitMemory(GetElementSize(m_format) * used_elements);
}
std::unique_ptr<GPUTextureBuffer> D3D12Device::CreateTextureBuffer(GPUTextureBuffer::Format format,
u32 size_in_elements)
{
std::unique_ptr<D3D12TextureBuffer> tb = std::make_unique<D3D12TextureBuffer>(format, size_in_elements);
if (!tb->Create(*this))
tb.reset();
return tb;
}

166
src/core/gpu/d3d12_texture.h Normal file
View File

@ -0,0 +1,166 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "d3d12_descriptor_heap_manager.h"
#include "d3d12_stream_buffer.h"
#include "gpu_device.h"
#include "gpu_texture.h"
#include <d3d12.h>
#include <limits>
#include <memory>
namespace D3D12MA {
class Allocation;
}
class D3D12Device;
class D3D12Texture final : public GPUTexture
{
friend D3D12Device;
public:
template<typename T>
using ComPtr = Microsoft::WRL::ComPtr<T>;
~D3D12Texture() override;
void Destroy(bool defer);
ALWAYS_INLINE const D3D12DescriptorHandle& GetSRVDescriptor() const { return m_srv_descriptor; }
ALWAYS_INLINE const D3D12DescriptorHandle& GetWriteDescriptor() const { return m_write_descriptor; }
ALWAYS_INLINE const D3D12DescriptorHandle& GetUAVDescriptor() const { return m_uav_descriptor; }
ALWAYS_INLINE D3D12_RESOURCE_STATES GetResourceState() const { return m_resource_state; }
ALWAYS_INLINE DXGI_FORMAT GetDXGIFormat() const { return m_dxgi_format; }
ALWAYS_INLINE ID3D12Resource* GetResource() const { return m_resource.Get(); }
bool IsValid() const override { return static_cast<bool>(m_resource); }
bool Update(u32 x, u32 y, u32 width, u32 height, const void* data, u32 pitch, u32 layer = 0, u32 level = 0) override;
bool Map(void** map, u32* map_stride, u32 x, u32 y, u32 width, u32 height, u32 layer = 0, u32 level = 0) override;
void Unmap() override;
void MakeReadyForSampling() override;
void SetDebugName(const std::string_view& name) override;
void TransitionToState(D3D12_RESOURCE_STATES state);
void CommitClear();
void CommitClear(ID3D12GraphicsCommandList* cmdlist);
static u32 CalculateSubresource(u32 layer, u32 level, u32 num_levels);
u32 CalculateSubresource(u32 layer, u32 level) const;
void TransitionToState(ID3D12GraphicsCommandList* cmdlist, D3D12_RESOURCE_STATES state);
void TransitionSubresourceToState(ID3D12GraphicsCommandList* cmdlist, u32 layer, u32 level,
D3D12_RESOURCE_STATES before_state, D3D12_RESOURCE_STATES after_state) const;
void TransitionSubresourceToState(ID3D12GraphicsCommandList* cmdlist, u32 subresource,
D3D12_RESOURCE_STATES before_state, D3D12_RESOURCE_STATES after_state) const;
static void TransitionSubresourceToState(ID3D12GraphicsCommandList* cmdlist, ID3D12Resource* resource,
u32 subresource, D3D12_RESOURCE_STATES before_state,
D3D12_RESOURCE_STATES after_state);
// Call when the texture is bound to the pipeline, or read from in a copy.
ALWAYS_INLINE void SetUseFenceValue(u64 counter) { m_use_fence_counter = counter; }
private:
enum class WriteDescriptorType : u8
{
None,
RTV,
DSV
};
D3D12Texture(u32 width, u32 height, u32 layers, u32 levels, u32 samples, Type type, Format format,
DXGI_FORMAT dxgi_format, ComPtr<ID3D12Resource> resource, ComPtr<D3D12MA::Allocation> allocation,
const D3D12DescriptorHandle& srv_descriptor, const D3D12DescriptorHandle& write_descriptor,
const D3D12DescriptorHandle& uav_descriptor, WriteDescriptorType wdtype,
D3D12_RESOURCE_STATES resource_state);
ID3D12GraphicsCommandList4* GetCommandBufferForUpdate();
ID3D12Resource* AllocateUploadStagingBuffer(const void* data, u32 pitch, u32 upload_pitch, u32 width,
u32 height) const;
void CopyTextureDataForUpload(void* dst, const void* src, u32 width, u32 height, u32 pitch, u32 upload_pitch) const;
ComPtr<ID3D12Resource> m_resource;
ComPtr<D3D12MA::Allocation> m_allocation;
D3D12DescriptorHandle m_srv_descriptor = {};
D3D12DescriptorHandle m_write_descriptor = {};
D3D12DescriptorHandle m_uav_descriptor = {};
DXGI_FORMAT m_dxgi_format = DXGI_FORMAT_UNKNOWN;
D3D12_RESOURCE_STATES m_resource_state = D3D12_RESOURCE_STATE_COMMON;
WriteDescriptorType m_write_descriptor_type = WriteDescriptorType::None;
// Contains the fence counter when the texture was last used.
// When this matches the current fence counter, the texture was used this command buffer.
u64 m_use_fence_counter = 0;
u16 m_map_x = 0;
u16 m_map_y = 0;
u16 m_map_width = 0;
u16 m_map_height = 0;
u8 m_map_layer = 0;
u8 m_map_level = 0;
};
class D3D12Sampler final : public GPUSampler
{
friend D3D12Device;
public:
~D3D12Sampler() override;
ALWAYS_INLINE const D3D12DescriptorHandle& GetDescriptor() const { return m_descriptor; }
void SetDebugName(const std::string_view& name) override;
private:
D3D12Sampler(D3D12DescriptorHandle descriptor);
D3D12DescriptorHandle m_descriptor;
};
class D3D12Framebuffer final : public GPUFramebuffer
{
friend D3D12Device;
public:
~D3D12Framebuffer() override;
ALWAYS_INLINE const D3D12DescriptorHandle& GetRTV() const { return m_rtv; }
ALWAYS_INLINE const D3D12DescriptorHandle& GetDSV() const { return m_dsv; }
void SetDebugName(const std::string_view& name) override;
private:
D3D12Framebuffer(GPUTexture* rt, GPUTexture* ds, u32 width, u32 height, D3D12DescriptorHandle rtv,
D3D12DescriptorHandle dsv);
D3D12DescriptorHandle m_rtv;
D3D12DescriptorHandle m_dsv;
};
class D3D12TextureBuffer final : public GPUTextureBuffer
{
friend D3D12Device;
public:
D3D12TextureBuffer(Format format, u32 size_in_elements);
~D3D12TextureBuffer() override;
ALWAYS_INLINE const D3D12DescriptorHandle& GetDescriptor() const { return m_descriptor; }
bool Create(D3D12Device& dev);
void Destroy(bool defer);
// Inherited via GPUTextureBuffer
void* Map(u32 required_elements) override;
void Unmap(u32 used_elements) override;
private:
D3D12StreamBuffer m_buffer;
D3D12DescriptorHandle m_descriptor;
};

388
src/core/gpu/d3d_common.cpp Normal file
View File

@ -0,0 +1,388 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#include "d3d_common.h"
#include "common/file_system.h"
#include "common/log.h"
#include "common/rectangle.h"
#include "common/string_util.h"
#include "fmt/format.h"
#include <d3dcompiler.h>
#include <dxgi1_5.h>
Log_SetChannel(D3DCommon);
static unsigned s_next_bad_shader_id = 1;
Microsoft::WRL::ComPtr<IDXGIFactory5> D3DCommon::CreateFactory(bool debug)
{
UINT flags = 0;
if (debug)
flags |= DXGI_CREATE_FACTORY_DEBUG;
Microsoft::WRL::ComPtr<IDXGIFactory5> factory;
const HRESULT hr = CreateDXGIFactory2(flags, IID_PPV_ARGS(factory.GetAddressOf()));
if (FAILED(hr))
Log_ErrorPrintf("Failed to create DXGI factory: %08X", hr);
return factory;
}
static std::string FixupDuplicateAdapterNames(const std::vector<std::string>& adapter_names, std::string adapter_name)
{
if (std::any_of(adapter_names.begin(), adapter_names.end(),
[&adapter_name](const std::string& other) { return (adapter_name == other); }))
{
std::string original_adapter_name = std::move(adapter_name);
u32 current_extra = 2;
do
{
adapter_name = fmt::format("{} ({})", original_adapter_name.c_str(), current_extra);
current_extra++;
} while (std::any_of(adapter_names.begin(), adapter_names.end(),
[&adapter_name](const std::string& other) { return (adapter_name == other); }));
}
return adapter_name;
}
std::vector<std::string> D3DCommon::GetAdapterNames(IDXGIFactory5* factory)
{
std::vector<std::string> adapter_names;
Microsoft::WRL::ComPtr<IDXGIAdapter1> adapter;
for (u32 index = 0;; index++)
{
const HRESULT hr = factory->EnumAdapters1(index, adapter.ReleaseAndGetAddressOf());
if (hr == DXGI_ERROR_NOT_FOUND)
break;
if (FAILED(hr))
{
Log_ErrorPrintf("IDXGIFactory2::EnumAdapters() returned %08X", hr);
continue;
}
adapter_names.push_back(FixupDuplicateAdapterNames(adapter_names, GetAdapterName(adapter.Get())));
}
return adapter_names;
}
std::vector<std::string> D3DCommon::GetFullscreenModes(IDXGIFactory5* factory, const std::string_view& adapter_name)
{
std::vector<std::string> modes;
HRESULT hr;
Microsoft::WRL::ComPtr<IDXGIAdapter1> adapter = GetChosenOrFirstAdapter(factory, adapter_name);
if (!adapter)
return modes;
Microsoft::WRL::ComPtr<IDXGIOutput> output;
if (FAILED(hr = adapter->EnumOutputs(0, &output)))
{
Log_ErrorPrintf("EnumOutputs() failed: %08X", hr);
return modes;
}
UINT num_modes = 0;
if (FAILED(hr = output->GetDisplayModeList(DXGI_FORMAT_R8G8B8A8_UNORM, 0, &num_modes, nullptr)))
{
Log_ErrorPrintf("GetDisplayModeList() failed: %08X", hr);
return modes;
}
std::vector<DXGI_MODE_DESC> dmodes(num_modes);
if (FAILED(hr = output->GetDisplayModeList(DXGI_FORMAT_R8G8B8A8_UNORM, 0, &num_modes, dmodes.data())))
{
Log_ErrorPrintf("GetDisplayModeList() (2) failed: %08X", hr);
return modes;
}
for (const DXGI_MODE_DESC& mode : dmodes)
{
modes.push_back(GPUDevice::GetFullscreenModeString(mode.Width, mode.Height,
static_cast<float>(mode.RefreshRate.Numerator) /
static_cast<float>(mode.RefreshRate.Denominator)));
}
return modes;
}
bool D3DCommon::GetRequestedExclusiveFullscreenModeDesc(IDXGIFactory5* factory, const RECT& window_rect, u32 width,
u32 height, float refresh_rate, DXGI_FORMAT format,
DXGI_MODE_DESC* fullscreen_mode, IDXGIOutput** output)
{
// We need to find which monitor the window is located on.
const Common::Rectangle<s32> client_rc_vec(window_rect.left, window_rect.top, window_rect.right, window_rect.bottom);
// The window might be on a different adapter to which we are rendering.. so we have to enumerate them all.
HRESULT hr;
Microsoft::WRL::ComPtr<IDXGIOutput> first_output, intersecting_output;
for (u32 adapter_index = 0; !intersecting_output; adapter_index++)
{
Microsoft::WRL::ComPtr<IDXGIAdapter1> adapter;
hr = factory->EnumAdapters1(adapter_index, adapter.GetAddressOf());
if (hr == DXGI_ERROR_NOT_FOUND)
break;
else if (FAILED(hr))
continue;
for (u32 output_index = 0;; output_index++)
{
Microsoft::WRL::ComPtr<IDXGIOutput> this_output;
DXGI_OUTPUT_DESC output_desc;
hr = adapter->EnumOutputs(output_index, this_output.GetAddressOf());
if (hr == DXGI_ERROR_NOT_FOUND)
break;
else if (FAILED(hr) || FAILED(this_output->GetDesc(&output_desc)))
continue;
const Common::Rectangle<s32> output_rc(output_desc.DesktopCoordinates.left, output_desc.DesktopCoordinates.top,
output_desc.DesktopCoordinates.right,
output_desc.DesktopCoordinates.bottom);
if (!client_rc_vec.Intersects(output_rc))
{
intersecting_output = std::move(this_output);
break;
}
// Fallback to the first monitor.
if (!first_output)
first_output = std::move(this_output);
}
}
if (!intersecting_output)
{
if (!first_output)
{
Log_ErrorPrintf("No DXGI output found. Can't use exclusive fullscreen.");
return false;
}
Log_WarningPrint("No DXGI output found for window, using first.");
intersecting_output = std::move(first_output);
}
DXGI_MODE_DESC request_mode = {};
request_mode.Width = width;
request_mode.Height = height;
request_mode.Format = format;
request_mode.RefreshRate.Numerator = static_cast<UINT>(std::floor(refresh_rate * 1000.0f));
request_mode.RefreshRate.Denominator = 1000u;
if (FAILED(hr = intersecting_output->FindClosestMatchingMode(&request_mode, fullscreen_mode, nullptr)) ||
request_mode.Format != format)
{
Log_ErrorPrintf("Failed to find closest matching mode, hr=%08X", hr);
return false;
}
*output = intersecting_output.Get();
intersecting_output->AddRef();
return true;
}
Microsoft::WRL::ComPtr<IDXGIAdapter1> D3DCommon::GetAdapterByName(IDXGIFactory5* factory, const std::string_view& name)
{
if (name.empty())
return {};
// This might seem a bit odd to cache the names.. but there's a method to the madness.
// We might have two GPUs with the same name... :)
std::vector<std::string> adapter_names;
Microsoft::WRL::ComPtr<IDXGIAdapter1> adapter;
for (u32 index = 0;; index++)
{
const HRESULT hr = factory->EnumAdapters1(index, adapter.ReleaseAndGetAddressOf());
if (hr == DXGI_ERROR_NOT_FOUND)
break;
if (FAILED(hr))
{
Log_ErrorPrintf("IDXGIFactory2::EnumAdapters() returned %08X");
continue;
}
std::string adapter_name = FixupDuplicateAdapterNames(adapter_names, GetAdapterName(adapter.Get()));
if (adapter_name == name)
{
Log_VerbosePrintf("Found adapter '%s'", adapter_name.c_str());
return adapter;
}
adapter_names.push_back(std::move(adapter_name));
}
Log_ErrorPrintf(fmt::format("Adapter '{}' not found.", name).c_str());
return {};
}
Microsoft::WRL::ComPtr<IDXGIAdapter1> D3DCommon::GetFirstAdapter(IDXGIFactory5* factory)
{
Microsoft::WRL::ComPtr<IDXGIAdapter1> adapter;
HRESULT hr = factory->EnumAdapters1(0, adapter.GetAddressOf());
if (FAILED(hr))
Log_ErrorPrintf("IDXGIFactory2::EnumAdapters() for first adapter returned %08X", hr);
return adapter;
}
Microsoft::WRL::ComPtr<IDXGIAdapter1> D3DCommon::GetChosenOrFirstAdapter(IDXGIFactory5* factory,
const std::string_view& name)
{
Microsoft::WRL::ComPtr<IDXGIAdapter1> adapter = GetAdapterByName(factory, name);
if (!adapter)
adapter = GetFirstAdapter(factory);
return adapter;
}
std::string D3DCommon::GetAdapterName(IDXGIAdapter1* adapter)
{
std::string ret;
DXGI_ADAPTER_DESC1 desc;
HRESULT hr = adapter->GetDesc1(&desc);
if (SUCCEEDED(hr))
{
ret = StringUtil::WideStringToUTF8String(desc.Description);
}
else
{
Log_ErrorPrintf("IDXGIAdapter1::GetDesc() returned %08X", hr);
}
if (ret.empty())
ret = "(Unknown)";
return ret;
}
std::string D3DCommon::GetDriverVersionFromLUID(const LUID& luid)
{
std::string ret;
HKEY hKey;
if (RegOpenKeyExW(HKEY_LOCAL_MACHINE, L"SOFTWARE\\Microsoft\\DirectX", 0, KEY_READ, &hKey) == ERROR_SUCCESS)
{
DWORD max_key_len = 0, adapter_count = 0;
if (RegQueryInfoKeyW(hKey, nullptr, nullptr, nullptr, &adapter_count, &max_key_len, nullptr, nullptr, nullptr,
nullptr, nullptr, nullptr) == ERROR_SUCCESS)
{
std::vector<WCHAR> current_name(max_key_len + 1);
for (DWORD i = 0; i < adapter_count; ++i)
{
DWORD subKeyLength = static_cast<DWORD>(current_name.size());
if (RegEnumKeyExW(hKey, i, current_name.data(), &subKeyLength, nullptr, nullptr, nullptr, nullptr) ==
ERROR_SUCCESS)
{
LUID current_luid = {};
DWORD current_luid_size = sizeof(uint64_t);
if (RegGetValueW(hKey, current_name.data(), L"AdapterLuid", RRF_RT_QWORD, nullptr, &current_luid,
&current_luid_size) == ERROR_SUCCESS &&
current_luid.HighPart == luid.HighPart && current_luid.LowPart == luid.LowPart)
{
LARGE_INTEGER driver_version = {};
DWORD driver_version_size = sizeof(driver_version);
if (RegGetValueW(hKey, current_name.data(), L"DriverVersion", RRF_RT_QWORD, nullptr, &driver_version,
&driver_version_size) == ERROR_SUCCESS)
{
WORD nProduct = HIWORD(driver_version.HighPart);
WORD nVersion = LOWORD(driver_version.HighPart);
WORD nSubVersion = HIWORD(driver_version.LowPart);
WORD nBuild = LOWORD(driver_version.LowPart);
ret = fmt::format("{}.{}.{}.{}", nProduct, nVersion, nSubVersion, nBuild);
}
}
}
}
}
RegCloseKey(hKey);
}
return ret;
}
std::optional<std::vector<u8>> D3DCommon::CompileShader(D3D_FEATURE_LEVEL feature_level, bool debug_device,
GPUShaderStage stage, const std::string_view& source)
{
const char* target;
switch (feature_level)
{
case D3D_FEATURE_LEVEL_10_0:
{
static constexpr std::array<const char*, 4> targets = {{"vs_4_0", "ps_4_0", "cs_4_0"}};
target = targets[static_cast<int>(stage)];
}
break;
case D3D_FEATURE_LEVEL_10_1:
{
static constexpr std::array<const char*, 4> targets = {{"vs_4_1", "ps_4_1", "cs_4_1"}};
target = targets[static_cast<int>(stage)];
}
break;
case D3D_FEATURE_LEVEL_11_0:
{
static constexpr std::array<const char*, 4> targets = {{"vs_5_0", "ps_5_0", "cs_5_0"}};
target = targets[static_cast<int>(stage)];
}
break;
case D3D_FEATURE_LEVEL_11_1:
default:
{
static constexpr std::array<const char*, 4> targets = {{"vs_5_1", "ps_5_1", "cs_5_1"}};
target = targets[static_cast<int>(stage)];
}
break;
}
static constexpr UINT flags_non_debug = D3DCOMPILE_OPTIMIZATION_LEVEL3;
static constexpr UINT flags_debug = D3DCOMPILE_SKIP_OPTIMIZATION | D3DCOMPILE_DEBUG;
Microsoft::WRL::ComPtr<ID3DBlob> blob;
Microsoft::WRL::ComPtr<ID3DBlob> error_blob;
const HRESULT hr =
D3DCompile(source.data(), source.size(), "0", nullptr, nullptr, "main", target,
debug_device ? flags_debug : flags_non_debug, 0, blob.GetAddressOf(), error_blob.GetAddressOf());
std::string error_string;
if (error_blob)
{
error_string.append(static_cast<const char*>(error_blob->GetBufferPointer()), error_blob->GetBufferSize());
error_blob.Reset();
}
if (FAILED(hr))
{
Log_ErrorPrintf("Failed to compile '%s':\n%s", target, error_string.c_str());
auto fp = FileSystem::OpenManagedCFile(
GPUDevice::GetShaderDumpPath(fmt::format("bad_shader_{}.txt", s_next_bad_shader_id++)).c_str(), "wb");
if (fp)
{
std::fwrite(source.data(), source.size(), 1, fp.get());
std::fprintf(fp.get(), "\n\nCompile as %s failed: %08X\n", target, hr);
std::fwrite(error_string.c_str(), error_string.size(), 1, fp.get());
}
return {};
}
if (!error_string.empty())
Log_WarningPrintf("'%s' compiled with warnings:\n%s", target, error_string.c_str());
return std::vector<u8>(static_cast<const u8*>(blob->GetBufferPointer()),
static_cast<const u8*>(blob->GetBufferPointer()) + blob->GetBufferSize());
}

55
src/core/gpu/d3d_common.h Normal file
View File

@ -0,0 +1,55 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "gpu_device.h"
#include "common/types.h"
#include "common/windows_headers.h"
#include <d3dcommon.h>
#include <dxgiformat.h>
#include <optional>
#include <string>
#include <vector>
#include <wrl/client.h>
struct IDXGIFactory5;
struct IDXGIAdapter1;
struct IDXGIOutput;
struct DXGI_MODE_DESC;
namespace D3DCommon {
// create a dxgi factory
Microsoft::WRL::ComPtr<IDXGIFactory5> CreateFactory(bool debug);
// returns a list of all adapter names
std::vector<std::string> GetAdapterNames(IDXGIFactory5* factory);
// returns a list of fullscreen modes for the specified adapter
std::vector<std::string> GetFullscreenModes(IDXGIFactory5* factory, const std::string_view& adapter_name);
// returns the fullscreen mode to use for the specified dimensions
bool GetRequestedExclusiveFullscreenModeDesc(IDXGIFactory5* factory, const RECT& window_rect, u32 width, u32 height,
float refresh_rate, DXGI_FORMAT format, DXGI_MODE_DESC* fullscreen_mode,
IDXGIOutput** output);
// get an adapter based on name
Microsoft::WRL::ComPtr<IDXGIAdapter1> GetAdapterByName(IDXGIFactory5* factory, const std::string_view& name);
// returns the first adapter in the system
Microsoft::WRL::ComPtr<IDXGIAdapter1> GetFirstAdapter(IDXGIFactory5* factory);
// returns the adapter specified in the configuration, or the default
Microsoft::WRL::ComPtr<IDXGIAdapter1> GetChosenOrFirstAdapter(IDXGIFactory5* factory, const std::string_view& name);
// returns a utf-8 string of the specified adapter's name
std::string GetAdapterName(IDXGIAdapter1* adapter);
// returns the driver version from the registry as a string
std::string GetDriverVersionFromLUID(const LUID& luid);
std::optional<std::vector<u8>> CompileShader(D3D_FEATURE_LEVEL feature_level, bool debug_device, GPUShaderStage stage,
const std::string_view& source);
} // namespace D3DCommon

930
src/core/gpu/d3d_shaders.h
View File

@ -1,930 +0,0 @@
#if 0
cbuffer UBOBlock : register(b0)
{
float4 u_src_rect;
};
void main(in uint vertex_id : SV_VertexID,
out float2 v_tex0 : TEXCOORD0,
out float4 o_pos : SV_Position)
{
float2 pos = float2(float((vertex_id << 1) & 2u), float(vertex_id & 2u));
v_tex0 = u_src_rect.xy + pos * u_src_rect.zw;
o_pos = float4(pos * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);
}
//
// Generated by Microsoft (R) HLSL Shader Compiler 10.1
//
//
// Buffer Definitions:
//
// cbuffer UBOBlock
// {
//
// float4 u_src_rect; // Offset: 0 Size: 16
//
// }
//
//
// Resource Bindings:
//
// Name Type Format Dim HLSL Bind Count
// ------------------------------ ---------- ------- ----------- -------------- ------
// UBOBlock cbuffer NA NA cb0 1
//
//
//
// Input signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// SV_VertexID 0 x 0 VERTID uint x
//
//
// Output signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// TEXCOORD 0 xy 0 NONE float xy
// SV_Position 0 xyzw 1 POS float xyzw
//
vs_4_0
dcl_constantbuffer CB0[1], immediateIndexed
dcl_input_sgv v0.x, vertex_id
dcl_output o0.xy
dcl_output_siv o1.xyzw, position
dcl_temps 1
ishl r0.x, v0.x, l(1)
and r0.x, r0.x, l(2)
and r0.z, v0.x, l(2)
utof r0.xy, r0.xzxx
mad o0.xy, r0.xyxx, cb0[0].zwzz, cb0[0].xyxx
mad o1.xy, r0.xyxx, l(2.000000, -2.000000, 0.000000, 0.000000), l(-1.000000, 1.000000, 0.000000, 0.000000)
mov o1.zw, l(0,0,0,1.000000)
ret
// Approximately 8 instruction slots used
#endif
const BYTE static s_display_vs_bytecode[] =
{
68, 88, 66, 67, 37, 97,
157, 234, 112, 10, 38, 98,
114, 228, 143, 118, 71, 158,
122, 195, 1, 0, 0, 0,
72, 3, 0, 0, 5, 0,
0, 0, 52, 0, 0, 0,
248, 0, 0, 0, 44, 1,
0, 0, 132, 1, 0, 0,
204, 2, 0, 0, 82, 68,
69, 70, 188, 0, 0, 0,
1, 0, 0, 0, 72, 0,
0, 0, 1, 0, 0, 0,
28, 0, 0, 0, 0, 4,
254, 255, 0, 129, 0, 0,
148, 0, 0, 0, 60, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 1, 0, 0, 0,
85, 66, 79, 66, 108, 111,
99, 107, 0, 171, 171, 171,
60, 0, 0, 0, 1, 0,
0, 0, 96, 0, 0, 0,
16, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
120, 0, 0, 0, 0, 0,
0, 0, 16, 0, 0, 0,
2, 0, 0, 0, 132, 0,
0, 0, 0, 0, 0, 0,
117, 95, 115, 114, 99, 95,
114, 101, 99, 116, 0, 171,
1, 0, 3, 0, 1, 0,
4, 0, 0, 0, 0, 0,
0, 0, 0, 0, 77, 105,
99, 114, 111, 115, 111, 102,
116, 32, 40, 82, 41, 32,
72, 76, 83, 76, 32, 83,
104, 97, 100, 101, 114, 32,
67, 111, 109, 112, 105, 108,
101, 114, 32, 49, 48, 46,
49, 0, 73, 83, 71, 78,
44, 0, 0, 0, 1, 0,
0, 0, 8, 0, 0, 0,
32, 0, 0, 0, 0, 0,
0, 0, 6, 0, 0, 0,
1, 0, 0, 0, 0, 0,
0, 0, 1, 1, 0, 0,
83, 86, 95, 86, 101, 114,
116, 101, 120, 73, 68, 0,
79, 83, 71, 78, 80, 0,
0, 0, 2, 0, 0, 0,
8, 0, 0, 0, 56, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 3, 0,
0, 0, 0, 0, 0, 0,
3, 12, 0, 0, 65, 0,
0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 3, 0,
0, 0, 1, 0, 0, 0,
15, 0, 0, 0, 84, 69,
88, 67, 79, 79, 82, 68,
0, 83, 86, 95, 80, 111,
115, 105, 116, 105, 111, 110,
0, 171, 171, 171, 83, 72,
68, 82, 64, 1, 0, 0,
64, 0, 1, 0, 80, 0,
0, 0, 89, 0, 0, 4,
70, 142, 32, 0, 0, 0,
0, 0, 1, 0, 0, 0,
96, 0, 0, 4, 18, 16,
16, 0, 0, 0, 0, 0,
6, 0, 0, 0, 101, 0,
0, 3, 50, 32, 16, 0,
0, 0, 0, 0, 103, 0,
0, 4, 242, 32, 16, 0,
1, 0, 0, 0, 1, 0,
0, 0, 104, 0, 0, 2,
1, 0, 0, 0, 41, 0,
0, 7, 18, 0, 16, 0,
0, 0, 0, 0, 10, 16,
16, 0, 0, 0, 0, 0,
1, 64, 0, 0, 1, 0,
0, 0, 1, 0, 0, 7,
18, 0, 16, 0, 0, 0,
0, 0, 10, 0, 16, 0,
0, 0, 0, 0, 1, 64,
0, 0, 2, 0, 0, 0,
1, 0, 0, 7, 66, 0,
16, 0, 0, 0, 0, 0,
10, 16, 16, 0, 0, 0,
0, 0, 1, 64, 0, 0,
2, 0, 0, 0, 86, 0,
0, 5, 50, 0, 16, 0,
0, 0, 0, 0, 134, 0,
16, 0, 0, 0, 0, 0,
50, 0, 0, 11, 50, 32,
16, 0, 0, 0, 0, 0,
70, 0, 16, 0, 0, 0,
0, 0, 230, 138, 32, 0,
0, 0, 0, 0, 0, 0,
0, 0, 70, 128, 32, 0,
0, 0, 0, 0, 0, 0,
0, 0, 50, 0, 0, 15,
50, 32, 16, 0, 1, 0,
0, 0, 70, 0, 16, 0,
0, 0, 0, 0, 2, 64,
0, 0, 0, 0, 0, 64,
0, 0, 0, 192, 0, 0,
0, 0, 0, 0, 0, 0,
2, 64, 0, 0, 0, 0,
128, 191, 0, 0, 128, 63,
0, 0, 0, 0, 0, 0,
0, 0, 54, 0, 0, 8,
194, 32, 16, 0, 1, 0,
0, 0, 2, 64, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 128, 63, 62, 0,
0, 1, 83, 84, 65, 84,
116, 0, 0, 0, 8, 0,
0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 3, 0,
0, 0, 2, 0, 0, 0,
1, 0, 0, 0, 2, 0,
0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0
};
#if 0
Texture2D samp0 : register(t0);
SamplerState samp0_ss : register(s0);
void main(in float2 v_tex0 : TEXCOORD0,
out float4 o_col0 : SV_Target)
{
#ifdef ALPHA
o_col0 = samp0.Sample(samp0_ss, v_tex0);
#else
o_col0 = float4(samp0.Sample(samp0_ss, v_tex0).rgb, 1.0);
#endif
}
//
// Generated by Microsoft (R) HLSL Shader Compiler 10.1
//
//
// Resource Bindings:
//
// Name Type Format Dim HLSL Bind Count
// ------------------------------ ---------- ------- ----------- -------------- ------
// samp0_ss sampler NA NA s0 1
// samp0 texture float4 2d t0 1
//
//
//
// Input signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// TEXCOORD 0 xy 0 NONE float xy
//
//
// Output signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// SV_Target 0 xyzw 0 TARGET float xyzw
//
ps_4_0
dcl_sampler s0, mode_default
dcl_resource_texture2d (float,float,float,float) t0
dcl_input_ps linear v0.xy
dcl_output o0.xyzw
dcl_temps 1
sample r0.xyzw, v0.xyxx, t0.xyzw, s0
mov o0.xyz, r0.xyzx
mov o0.w, l(1.000000)
ret
// Approximately 4 instruction slots used
#endif
const BYTE static s_display_ps_bytecode[] =
{
68, 88, 66, 67, 192, 215,
150, 96, 210, 93, 209, 128,
113, 254, 100, 56, 49, 113,
128, 72, 1, 0, 0, 0,
80, 2, 0, 0, 5, 0,
0, 0, 52, 0, 0, 0,
208, 0, 0, 0, 4, 1,
0, 0, 56, 1, 0, 0,
212, 1, 0, 0, 82, 68,
69, 70, 148, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 2, 0, 0, 0,
28, 0, 0, 0, 0, 4,
255, 255, 0, 129, 0, 0,
107, 0, 0, 0, 92, 0,
0, 0, 3, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 1, 0, 0, 0,
101, 0, 0, 0, 2, 0,
0, 0, 5, 0, 0, 0,
4, 0, 0, 0, 255, 255,
255, 255, 0, 0, 0, 0,
1, 0, 0, 0, 13, 0,
0, 0, 115, 97, 109, 112,
48, 95, 115, 115, 0, 115,
97, 109, 112, 48, 0, 77,
105, 99, 114, 111, 115, 111,
102, 116, 32, 40, 82, 41,
32, 72, 76, 83, 76, 32,
83, 104, 97, 100, 101, 114,
32, 67, 111, 109, 112, 105,
108, 101, 114, 32, 49, 48,
46, 49, 0, 171, 73, 83,
71, 78, 44, 0, 0, 0,
1, 0, 0, 0, 8, 0,
0, 0, 32, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 3, 0, 0, 0,
0, 0, 0, 0, 3, 3,
0, 0, 84, 69, 88, 67,
79, 79, 82, 68, 0, 171,
171, 171, 79, 83, 71, 78,
44, 0, 0, 0, 1, 0,
0, 0, 8, 0, 0, 0,
32, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
3, 0, 0, 0, 0, 0,
0, 0, 15, 0, 0, 0,
83, 86, 95, 84, 97, 114,
103, 101, 116, 0, 171, 171,
83, 72, 68, 82, 148, 0,
0, 0, 64, 0, 0, 0,
37, 0, 0, 0, 90, 0,
0, 3, 0, 96, 16, 0,
0, 0, 0, 0, 88, 24,
0, 4, 0, 112, 16, 0,
0, 0, 0, 0, 85, 85,
0, 0, 98, 16, 0, 3,
50, 16, 16, 0, 0, 0,
0, 0, 101, 0, 0, 3,
242, 32, 16, 0, 0, 0,
0, 0, 104, 0, 0, 2,
1, 0, 0, 0, 69, 0,
0, 9, 242, 0, 16, 0,
0, 0, 0, 0, 70, 16,
16, 0, 0, 0, 0, 0,
70, 126, 16, 0, 0, 0,
0, 0, 0, 96, 16, 0,
0, 0, 0, 0, 54, 0,
0, 5, 114, 32, 16, 0,
0, 0, 0, 0, 70, 2,
16, 0, 0, 0, 0, 0,
54, 0, 0, 5, 130, 32,
16, 0, 0, 0, 0, 0,
1, 64, 0, 0, 0, 0,
128, 63, 62, 0, 0, 1,
83, 84, 65, 84, 116, 0,
0, 0, 4, 0, 0, 0,
1, 0, 0, 0, 0, 0,
0, 0, 2, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
2, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0
};
#if 0
//
// Generated by Microsoft (R) HLSL Shader Compiler 10.1
//
//
// Resource Bindings:
//
// Name Type Format Dim HLSL Bind Count
// ------------------------------ ---------- ------- ----------- -------------- ------
// samp0_ss sampler NA NA s0 1
// samp0 texture float4 2d t0 1
//
//
//
// Input signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// TEXCOORD 0 xy 0 NONE float xy
//
//
// Output signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// SV_Target 0 xyzw 0 TARGET float xyzw
//
ps_4_0
dcl_sampler s0, mode_default
dcl_resource_texture2d(float, float, float, float) t0
dcl_input_ps linear v0.xy
dcl_output o0.xyzw
sample o0.xyzw, v0.xyxx, t0.xyzw, s0
ret
// Approximately 2 instruction slots used
#endif
const BYTE static s_display_ps_alpha_bytecode[] =
{
68, 88, 66, 67, 140, 134,
46, 29, 68, 36, 193, 23,
94, 171, 102, 123, 183, 66,
19, 177, 1, 0, 0, 0,
32, 2, 0, 0, 5, 0,
0, 0, 52, 0, 0, 0,
208, 0, 0, 0, 4, 1,
0, 0, 56, 1, 0, 0,
164, 1, 0, 0, 82, 68,
69, 70, 148, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 2, 0, 0, 0,
28, 0, 0, 0, 0, 4,
255, 255, 0, 129, 0, 0,
107, 0, 0, 0, 92, 0,
0, 0, 3, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 1, 0, 0, 0,
101, 0, 0, 0, 2, 0,
0, 0, 5, 0, 0, 0,
4, 0, 0, 0, 255, 255,
255, 255, 0, 0, 0, 0,
1, 0, 0, 0, 13, 0,
0, 0, 115, 97, 109, 112,
48, 95, 115, 115, 0, 115,
97, 109, 112, 48, 0, 77,
105, 99, 114, 111, 115, 111,
102, 116, 32, 40, 82, 41,
32, 72, 76, 83, 76, 32,
83, 104, 97, 100, 101, 114,
32, 67, 111, 109, 112, 105,
108, 101, 114, 32, 49, 48,
46, 49, 0, 171, 73, 83,
71, 78, 44, 0, 0, 0,
1, 0, 0, 0, 8, 0,
0, 0, 32, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 3, 0, 0, 0,
0, 0, 0, 0, 3, 3,
0, 0, 84, 69, 88, 67,
79, 79, 82, 68, 0, 171,
171, 171, 79, 83, 71, 78,
44, 0, 0, 0, 1, 0,
0, 0, 8, 0, 0, 0,
32, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
3, 0, 0, 0, 0, 0,
0, 0, 15, 0, 0, 0,
83, 86, 95, 84, 97, 114,
103, 101, 116, 0, 171, 171,
83, 72, 68, 82, 100, 0,
0, 0, 64, 0, 0, 0,
25, 0, 0, 0, 90, 0,
0, 3, 0, 96, 16, 0,
0, 0, 0, 0, 88, 24,
0, 4, 0, 112, 16, 0,
0, 0, 0, 0, 85, 85,
0, 0, 98, 16, 0, 3,
50, 16, 16, 0, 0, 0,
0, 0, 101, 0, 0, 3,
242, 32, 16, 0, 0, 0,
0, 0, 69, 0, 0, 9,
242, 32, 16, 0, 0, 0,
0, 0, 70, 16, 16, 0,
0, 0, 0, 0, 70, 126,
16, 0, 0, 0, 0, 0,
0, 96, 16, 0, 0, 0,
0, 0, 62, 0, 0, 1,
83, 84, 65, 84, 116, 0,
0, 0, 2, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 2, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0
};
#if 0
cbuffer vertexBuffer : register(b0)
{
float4x4 ProjectionMatrix;
};
struct VS_INPUT
{
float2 pos : POSITION;
float4 col : COLOR0;
float2 uv : TEXCOORD0;
};
struct PS_INPUT
{
float4 pos : SV_POSITION;
float4 col : COLOR0;
float2 uv : TEXCOORD0;
};
PS_INPUT vs_main(VS_INPUT input)
{
PS_INPUT output;
output.pos = mul(ProjectionMatrix, float4(input.pos.xy, 0.f, 1.f));
output.col = input.col;
output.uv = input.uv;
return output;
}
sampler sampler0 : register(s0);
Texture2D texture0 : register(t0);
float4 ps_main(PS_INPUT input) : SV_Target
{
float4 out_col = input.col * texture0.Sample(sampler0, input.uv);
return out_col;
}
//
// Generated by Microsoft (R) HLSL Shader Compiler 10.1
//
//
// Buffer Definitions:
//
// cbuffer vertexBuffer
// {
//
// float4x4 ProjectionMatrix; // Offset: 0 Size: 64
//
// }
//
//
// Resource Bindings:
//
// Name Type Format Dim HLSL Bind Count
// ------------------------------ ---------- ------- ----------- -------------- ------
// vertexBuffer cbuffer NA NA cb0 1
//
//
//
// Input signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// POSITION 0 xy 0 NONE float xy
// COLOR 0 xyzw 1 NONE float xyzw
// TEXCOORD 0 xy 2 NONE float xy
//
//
// Output signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// SV_POSITION 0 xyzw 0 POS float xyzw
// COLOR 0 xyzw 1 NONE float xyzw
// TEXCOORD 0 xy 2 NONE float xy
//
vs_5_0
dcl_globalFlags refactoringAllowed
dcl_constantbuffer CB0[4], immediateIndexed
dcl_input v0.xy
dcl_input v1.xyzw
dcl_input v2.xy
dcl_output_siv o0.xyzw, position
dcl_output o1.xyzw
dcl_output o2.xy
dcl_temps 1
mul r0.xyzw, v0.yyyy, cb0[1].xyzw
mad r0.xyzw, cb0[0].xyzw, v0.xxxx, r0.xyzw
add o0.xyzw, r0.xyzw, cb0[3].xyzw
mov o1.xyzw, v1.xyzw
mov o2.xy, v2.xyxx
ret
// Approximately 6 instruction slots used
#endif
const BYTE static s_imgui_vs_bytecode[] =
{
68, 88, 66, 67, 115, 32,
121, 4, 196, 220, 218, 4,
63, 218, 114, 0, 25, 78,
192, 97, 1, 0, 0, 0,
220, 3, 0, 0, 5, 0,
0, 0, 52, 0, 0, 0,
80, 1, 0, 0, 192, 1,
0, 0, 52, 2, 0, 0,
64, 3, 0, 0, 82, 68,
69, 70, 20, 1, 0, 0,
1, 0, 0, 0, 108, 0,
0, 0, 1, 0, 0, 0,
60, 0, 0, 0, 0, 5,
254, 255, 0, 129, 0, 0,
236, 0, 0, 0, 82, 68,
49, 49, 60, 0, 0, 0,
24, 0, 0, 0, 32, 0,
0, 0, 40, 0, 0, 0,
36, 0, 0, 0, 12, 0,
0, 0, 0, 0, 0, 0,
92, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 1, 0,
0, 0, 118, 101, 114, 116,
101, 120, 66, 117, 102, 102,
101, 114, 0, 171, 171, 171,
92, 0, 0, 0, 1, 0,
0, 0, 132, 0, 0, 0,
64, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
172, 0, 0, 0, 0, 0,
0, 0, 64, 0, 0, 0,
2, 0, 0, 0, 200, 0,
0, 0, 0, 0, 0, 0,
255, 255, 255, 255, 0, 0,
0, 0, 255, 255, 255, 255,
0, 0, 0, 0, 80, 114,
111, 106, 101, 99, 116, 105,
111, 110, 77, 97, 116, 114,
105, 120, 0, 102, 108, 111,
97, 116, 52, 120, 52, 0,
171, 171, 3, 0, 3, 0,
4, 0, 4, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 189, 0,
0, 0, 77, 105, 99, 114,
111, 115, 111, 102, 116, 32,
40, 82, 41, 32, 72, 76,
83, 76, 32, 83, 104, 97,
100, 101, 114, 32, 67, 111,
109, 112, 105, 108, 101, 114,
32, 49, 48, 46, 49, 0,
73, 83, 71, 78, 104, 0,
0, 0, 3, 0, 0, 0,
8, 0, 0, 0, 80, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 3, 0,
0, 0, 0, 0, 0, 0,
3, 3, 0, 0, 89, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 3, 0,
0, 0, 1, 0, 0, 0,
15, 15, 0, 0, 95, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 3, 0,
0, 0, 2, 0, 0, 0,
3, 3, 0, 0, 80, 79,
83, 73, 84, 73, 79, 78,
0, 67, 79, 76, 79, 82,
0, 84, 69, 88, 67, 79,
79, 82, 68, 0, 79, 83,
71, 78, 108, 0, 0, 0,
3, 0, 0, 0, 8, 0,
0, 0, 80, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 3, 0, 0, 0,
0, 0, 0, 0, 15, 0,
0, 0, 92, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 3, 0, 0, 0,
1, 0, 0, 0, 15, 0,
0, 0, 98, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 3, 0, 0, 0,
2, 0, 0, 0, 3, 12,
0, 0, 83, 86, 95, 80,
79, 83, 73, 84, 73, 79,
78, 0, 67, 79, 76, 79,
82, 0, 84, 69, 88, 67,
79, 79, 82, 68, 0, 171,
83, 72, 69, 88, 4, 1,
0, 0, 80, 0, 1, 0,
65, 0, 0, 0, 106, 8,
0, 1, 89, 0, 0, 4,
70, 142, 32, 0, 0, 0,
0, 0, 4, 0, 0, 0,
95, 0, 0, 3, 50, 16,
16, 0, 0, 0, 0, 0,
95, 0, 0, 3, 242, 16,
16, 0, 1, 0, 0, 0,
95, 0, 0, 3, 50, 16,
16, 0, 2, 0, 0, 0,
103, 0, 0, 4, 242, 32,
16, 0, 0, 0, 0, 0,
1, 0, 0, 0, 101, 0,
0, 3, 242, 32, 16, 0,
1, 0, 0, 0, 101, 0,
0, 3, 50, 32, 16, 0,
2, 0, 0, 0, 104, 0,
0, 2, 1, 0, 0, 0,
56, 0, 0, 8, 242, 0,
16, 0, 0, 0, 0, 0,
86, 21, 16, 0, 0, 0,
0, 0, 70, 142, 32, 0,
0, 0, 0, 0, 1, 0,
0, 0, 50, 0, 0, 10,
242, 0, 16, 0, 0, 0,
0, 0, 70, 142, 32, 0,
0, 0, 0, 0, 0, 0,
0, 0, 6, 16, 16, 0,
0, 0, 0, 0, 70, 14,
16, 0, 0, 0, 0, 0,
0, 0, 0, 8, 242, 32,
16, 0, 0, 0, 0, 0,
70, 14, 16, 0, 0, 0,
0, 0, 70, 142, 32, 0,
0, 0, 0, 0, 3, 0,
0, 0, 54, 0, 0, 5,
242, 32, 16, 0, 1, 0,
0, 0, 70, 30, 16, 0,
1, 0, 0, 0, 54, 0,
0, 5, 50, 32, 16, 0,
2, 0, 0, 0, 70, 16,
16, 0, 2, 0, 0, 0,
62, 0, 0, 1, 83, 84,
65, 84, 148, 0, 0, 0,
6, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0,
6, 0, 0, 0, 3, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 2, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0
};
#if 0
//
// Generated by Microsoft (R) HLSL Shader Compiler 10.1
//
//
// Resource Bindings:
//
// Name Type Format Dim HLSL Bind Count
// ------------------------------ ---------- ------- ----------- -------------- ------
// sampler0 sampler NA NA s0 1
// texture0 texture float4 2d t0 1
//
//
//
// Input signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// SV_POSITION 0 xyzw 0 POS float
// COLOR 0 xyzw 1 NONE float xyzw
// TEXCOORD 0 xy 2 NONE float xy
//
//
// Output signature:
//
// Name Index Mask Register SysValue Format Used
// -------------------- ----- ------ -------- -------- ------- ------
// SV_Target 0 xyzw 0 TARGET float xyzw
//
ps_5_0
dcl_globalFlags refactoringAllowed
dcl_sampler s0, mode_default
dcl_resource_texture2d (float,float,float,float) t0
dcl_input_ps linear v1.xyzw
dcl_input_ps linear v2.xy
dcl_output o0.xyzw
dcl_temps 1
sample_indexable(texture2d)(float,float,float,float) r0.xyzw, v2.xyxx, t0.xyzw, s0
mul o0.xyzw, r0.xyzw, v1.xyzw
ret
// Approximately 3 instruction slots used
#endif
const BYTE static s_imgui_ps_bytecode[] =
{
68, 88, 66, 67, 219, 131,
126, 77, 0, 201, 229, 66,
111, 92, 42, 162, 229, 97,
51, 114, 1, 0, 0, 0,
224, 2, 0, 0, 5, 0,
0, 0, 52, 0, 0, 0,
244, 0, 0, 0, 104, 1,
0, 0, 156, 1, 0, 0,
68, 2, 0, 0, 82, 68,
69, 70, 184, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 2, 0, 0, 0,
60, 0, 0, 0, 0, 5,
255, 255, 0, 129, 0, 0,
142, 0, 0, 0, 82, 68,
49, 49, 60, 0, 0, 0,
24, 0, 0, 0, 32, 0,
0, 0, 40, 0, 0, 0,
36, 0, 0, 0, 12, 0,
0, 0, 0, 0, 0, 0,
124, 0, 0, 0, 3, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 1, 0,
0, 0, 133, 0, 0, 0,
2, 0, 0, 0, 5, 0,
0, 0, 4, 0, 0, 0,
255, 255, 255, 255, 0, 0,
0, 0, 1, 0, 0, 0,
13, 0, 0, 0, 115, 97,
109, 112, 108, 101, 114, 48,
0, 116, 101, 120, 116, 117,
114, 101, 48, 0, 77, 105,
99, 114, 111, 115, 111, 102,
116, 32, 40, 82, 41, 32,
72, 76, 83, 76, 32, 83,
104, 97, 100, 101, 114, 32,
67, 111, 109, 112, 105, 108,
101, 114, 32, 49, 48, 46,
49, 0, 171, 171, 73, 83,
71, 78, 108, 0, 0, 0,
3, 0, 0, 0, 8, 0,
0, 0, 80, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 3, 0, 0, 0,
0, 0, 0, 0, 15, 0,
0, 0, 92, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 3, 0, 0, 0,
1, 0, 0, 0, 15, 15,
0, 0, 98, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 3, 0, 0, 0,
2, 0, 0, 0, 3, 3,
0, 0, 83, 86, 95, 80,
79, 83, 73, 84, 73, 79,
78, 0, 67, 79, 76, 79,
82, 0, 84, 69, 88, 67,
79, 79, 82, 68, 0, 171,
79, 83, 71, 78, 44, 0,
0, 0, 1, 0, 0, 0,
8, 0, 0, 0, 32, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 3, 0,
0, 0, 0, 0, 0, 0,
15, 0, 0, 0, 83, 86,
95, 84, 97, 114, 103, 101,
116, 0, 171, 171, 83, 72,
69, 88, 160, 0, 0, 0,
80, 0, 0, 0, 40, 0,
0, 0, 106, 8, 0, 1,
90, 0, 0, 3, 0, 96,
16, 0, 0, 0, 0, 0,
88, 24, 0, 4, 0, 112,
16, 0, 0, 0, 0, 0,
85, 85, 0, 0, 98, 16,
0, 3, 242, 16, 16, 0,
1, 0, 0, 0, 98, 16,
0, 3, 50, 16, 16, 0,
2, 0, 0, 0, 101, 0,
0, 3, 242, 32, 16, 0,
0, 0, 0, 0, 104, 0,
0, 2, 1, 0, 0, 0,
69, 0, 0, 139, 194, 0,
0, 128, 67, 85, 21, 0,
242, 0, 16, 0, 0, 0,
0, 0, 70, 16, 16, 0,
2, 0, 0, 0, 70, 126,
16, 0, 0, 0, 0, 0,
0, 96, 16, 0, 0, 0,
0, 0, 56, 0, 0, 7,
242, 32, 16, 0, 0, 0,
0, 0, 70, 14, 16, 0,
0, 0, 0, 0, 70, 30,
16, 0, 1, 0, 0, 0,
62, 0, 0, 1, 83, 84,
65, 84, 148, 0, 0, 0,
3, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0,
3, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0,
0, 0, 0, 0
};

30
src/core/gpu/gl/context_egl.cpp
View File

@ -9,7 +9,9 @@
Log_SetChannel(GL::ContextEGL);
namespace GL {
ContextEGL::ContextEGL(const WindowInfo& wi) : Context(wi) {}
ContextEGL::ContextEGL(const WindowInfo& wi) : Context(wi)
{
}
ContextEGL::~ContextEGL()
{
@ -245,10 +247,27 @@ bool ContextEGL::CheckConfigSurfaceFormat(EGLConfig config, WindowInfo::SurfaceF
switch (format)
{
case GPUTexture::Format::Undefined:
return true;
{
if (red_size == 5 && green_size == 6 && red_size == 5)
{
m_wi.surface_format = GPUTexture::Format::RGB565;
}
else if (red_size == 5 && green_size == 5 && red_size == 5 && alpha_size == 1)
{
m_wi.surface_format = GPUTexture::Format::RGBA5551;
}
else if (red_size == 8 && green_size == 8 && blue_size == 8)
{
m_wi.surface_format = GPUTexture::Format::RGBA8;
}
else
{
Log_ErrorPrintf("Unknown surface format: R=%u, G=%u, B=%u, A=%u", red_size, green_size, blue_size, alpha_size);
m_wi.surface_format = GPUTexture::Format::RGBA8;
}
case GPUTexture::Format::RGB8:
return (red_size == 8 && green_size == 8 && blue_size == 8);
return true;
}
case GPUTexture::Format::RGBA8:
return (red_size == 8 && green_size == 8 && blue_size == 8 && alpha_size == 8);
@ -256,6 +275,9 @@ bool ContextEGL::CheckConfigSurfaceFormat(EGLConfig config, WindowInfo::SurfaceF
case GPUTexture::Format::RGB565:
return (red_size == 5 && green_size == 6 && blue_size == 5);
case GPUTexture::Format::RGBA5551:
return (red_size == 5 && green_size == 5 && blue_size == 5 && alpha_size == 1);
default:
return false;
}

4
src/core/gpu/gl/context_wgl.cpp
View File

@ -10,9 +10,6 @@
Log_SetChannel(GL::ContextWGL);
// TODO: get rid of this
#pragma comment(lib, "opengl32.lib")
static void* GetProcAddressCallback(const char* name)
{
void* addr = wglGetProcAddress(name);
@ -231,6 +228,7 @@ HDC ContextWGL::GetDCAndSetPixelFormat(HWND hwnd)
return {};
}
m_wi.surface_format = GPUTexture::Format::RGBA8;
return hDC;
}

6
src/core/gpu/gpu_device.cpp
View File

@ -33,7 +33,7 @@ Log_SetChannel(GPUDevice);
#ifdef _WIN32
#include "common/windows_headers.h"
#include "d3d11_device.h"
#include "d3d12_gpu_device.h"
#include "d3d12_device.h"
#endif
#ifdef __APPLE__
@ -263,6 +263,7 @@ bool GPUDevice::Create(const std::string_view& adapter, const std::string_view&
void GPUDevice::Destroy()
{
m_post_processing_chain.reset();
if (HasSurface())
DestroySurface();
DestroyResources();
@ -1450,8 +1451,7 @@ std::unique_ptr<GPUDevice> GPUDevice::CreateDeviceForAPI(RenderAPI api)
#ifdef _WIN32
case RenderAPI::D3D12:
// return std::make_unique<D3D12GPUDevice>();
return {};
return std::make_unique<D3D12Device>();
case RenderAPI::D3D11:
return std::make_unique<D3D11Device>();

545
src/core/gpu/imgui_impl_dx12.cpp
View File

@ -1,545 +0,0 @@
// dear imgui: Renderer Backend for DirectX12
// This needs to be used along with a Platform Backend (e.g. Win32)
// Implemented features:
// [X] Renderer: User texture binding. Use 'D3D12_GPU_DESCRIPTOR_HANDLE' as ImTextureID. Read the FAQ about ImTextureID!
// [X] Renderer: Support for large meshes (64k+ vertices) with 16-bit indices.
// Important: to compile on 32-bit systems, this backend requires code to be compiled with '#define ImTextureID ImU64'.
// This is because we need ImTextureID to carry a 64-bit value and by default ImTextureID is defined as void*.
// To build this on 32-bit systems:
// - [Solution 1] IDE/msbuild: in "Properties/C++/Preprocessor Definitions" add 'ImTextureID=ImU64' (this is what we do in the 'example_win32_direct12/example_win32_direct12.vcxproj' project file)
// - [Solution 2] IDE/msbuild: in "Properties/C++/Preprocessor Definitions" add 'IMGUI_USER_CONFIG="my_imgui_config.h"' and inside 'my_imgui_config.h' add '#define ImTextureID ImU64' and as many other options as you like.
// - [Solution 3] IDE/msbuild: edit imconfig.h and add '#define ImTextureID ImU64' (prefer solution 2 to create your own config file!)
// - [Solution 4] command-line: add '/D ImTextureID=ImU64' to your cl.exe command-line (this is what we do in the example_win32_direct12/build_win32.bat file)
// You can use unmodified imgui_impl_* files in your project. See examples/ folder for examples of using this.
// Prefer including the entire imgui/ repository into your project (either as a copy or as a submodule), and only build the backends you need.
// If you are new to Dear ImGui, read documentation from the docs/ folder + read the top of imgui.cpp.
// Read online: https://github.com/ocornut/imgui/tree/master/docs
// CHANGELOG
// (minor and older changes stripped away, please see git history for details)
// 2021-06-29: Reorganized backend to pull data from a single structure to facilitate usage with multiple-contexts (all g_XXXX access changed to bd->XXXX).
// 2021-05-19: DirectX12: Replaced direct access to ImDrawCmd::TextureId with a call to ImDrawCmd::GetTexID(). (will become a requirement)
// 2021-02-18: DirectX12: Change blending equation to preserve alpha in output buffer.
// 2021-01-11: DirectX12: Improve Windows 7 compatibility (for D3D12On7) by loading d3d12.dll dynamically.
// 2020-09-16: DirectX12: Avoid rendering calls with zero-sized scissor rectangle since it generates a validation layer warning.
// 2020-09-08: DirectX12: Clarified support for building on 32-bit systems by redefining ImTextureID.
// 2019-10-18: DirectX12: *BREAKING CHANGE* Added extra ID3D12DescriptorHeap parameter to ImGui_ImplDX12_Init() function.
// 2019-05-29: DirectX12: Added support for large mesh (64K+ vertices), enable ImGuiBackendFlags_RendererHasVtxOffset flag.
// 2019-04-30: DirectX12: Added support for special ImDrawCallback_ResetRenderState callback to reset render state.
// 2019-03-29: Misc: Various minor tidying up.
// 2018-12-03: Misc: Added #pragma comment statement to automatically link with d3dcompiler.lib when using D3DCompile().
// 2018-11-30: Misc: Setting up io.BackendRendererName so it can be displayed in the About Window.
// 2018-06-12: DirectX12: Moved the ID3D12GraphicsCommandList* parameter from NewFrame() to RenderDrawData().
// 2018-06-08: Misc: Extracted imgui_impl_dx12.cpp/.h away from the old combined DX12+Win32 example.
// 2018-06-08: DirectX12: Use draw_data->DisplayPos and draw_data->DisplaySize to setup projection matrix and clipping rectangle (to ease support for future multi-viewport).
// 2018-02-22: Merged into master with all Win32 code synchronized to other examples.
#include "common/windows_headers.h"
#include "common/assert.h"
#include "d3d12/context.h"
#include "d3d12/texture.h"
#include "d3d12/stream_buffer.h"
#include "imgui.h"
#include "imgui_impl_dx12.h"
// DirectX
#include <d3d12.h>
#include <dxgi1_4.h>
#include <d3dcompiler.h>
#ifdef _MSC_VER
#pragma comment(lib, "d3dcompiler") // Automatically link with d3dcompiler.lib as we are using D3DCompile() below.
#endif
// If we're doing more than this... wtf?
static constexpr u32 VERTEX_BUFFER_SIZE = 8 * 1024 * 1024;
static constexpr u32 INDEX_BUFFER_SIZE = 4 * 1024 * 1024;
struct ImGui_ImplDX12_Data
{
D3D12::StreamBuffer VertexStreamBuffer;
D3D12::StreamBuffer IndexStreamBuffer;
D3D12::Texture FontTexture;
ID3D12RootSignature* pRootSignature = nullptr;
ID3D12PipelineState* pPipelineState = nullptr;
DXGI_FORMAT RTVFormat = DXGI_FORMAT_UNKNOWN;
};
struct VERTEX_CONSTANT_BUFFER
{
float mvp[4][4];
};
// Backend data stored in io.BackendRendererUserData to allow support for multiple Dear ImGui contexts
// It is STRONGLY preferred that you use docking branch with multi-viewports (== single Dear ImGui context + multiple windows) instead of multiple Dear ImGui contexts.
static ImGui_ImplDX12_Data* ImGui_ImplDX12_GetBackendData()
{
return ImGui::GetCurrentContext() ? (ImGui_ImplDX12_Data*)ImGui::GetIO().BackendRendererUserData : NULL;
}
// Functions
static void ImGui_ImplDX12_SetupRenderState(ImDrawData* draw_data, ID3D12GraphicsCommandList* ctx)
{
ImGui_ImplDX12_Data* bd = ImGui_ImplDX12_GetBackendData();
// Setup orthographic projection matrix into our constant buffer
// Our visible imgui space lies from draw_data->DisplayPos (top left) to draw_data->DisplayPos+data_data->DisplaySize (bottom right).
VERTEX_CONSTANT_BUFFER vertex_constant_buffer;
{
float L = draw_data->DisplayPos.x;
float R = draw_data->DisplayPos.x + draw_data->DisplaySize.x;
float T = draw_data->DisplayPos.y;
float B = draw_data->DisplayPos.y + draw_data->DisplaySize.y;
float mvp[4][4] =
{
{ 2.0f/(R-L), 0.0f, 0.0f, 0.0f },
{ 0.0f, 2.0f/(T-B), 0.0f, 0.0f },
{ 0.0f, 0.0f, 0.5f, 0.0f },
{ (R+L)/(L-R), (T+B)/(B-T), 0.5f, 1.0f },
};
memcpy(&vertex_constant_buffer.mvp, mvp, sizeof(mvp));
}
// Setup viewport
D3D12_VIEWPORT vp;
memset(&vp, 0, sizeof(D3D12_VIEWPORT));
vp.Width = draw_data->DisplaySize.x;
vp.Height = draw_data->DisplaySize.y;
vp.MinDepth = 0.0f;
vp.MaxDepth = 1.0f;
vp.TopLeftX = vp.TopLeftY = 0.0f;
ctx->RSSetViewports(1, &vp);
// Bind shader and vertex buffers
unsigned int stride = sizeof(ImDrawVert);
D3D12_VERTEX_BUFFER_VIEW vbv;
memset(&vbv, 0, sizeof(D3D12_VERTEX_BUFFER_VIEW));
vbv.BufferLocation = bd->VertexStreamBuffer.GetCurrentGPUPointer();
vbv.SizeInBytes = bd->VertexStreamBuffer.GetCurrentSpace();
vbv.StrideInBytes = stride;
ctx->IASetVertexBuffers(0, 1, &vbv);
D3D12_INDEX_BUFFER_VIEW ibv;
memset(&ibv, 0, sizeof(D3D12_INDEX_BUFFER_VIEW));
ibv.BufferLocation = bd->IndexStreamBuffer.GetCurrentGPUPointer();
ibv.SizeInBytes = bd->IndexStreamBuffer.GetCurrentSpace();
ibv.Format = sizeof(ImDrawIdx) == 2 ? DXGI_FORMAT_R16_UINT : DXGI_FORMAT_R32_UINT;
ctx->IASetIndexBuffer(&ibv);
ctx->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST);
ctx->SetPipelineState(bd->pPipelineState);
ctx->SetGraphicsRootSignature(bd->pRootSignature);
ctx->SetGraphicsRoot32BitConstants(0, 16, &vertex_constant_buffer, 0);
// Setup blend factor
const float blend_factor[4] = { 0.f, 0.f, 0.f, 0.f };
ctx->OMSetBlendFactor(blend_factor);
}
template<typename T>
static inline void SafeRelease(T*& res)
{
if (res)
res->Release();
res = NULL;
}
// Render function
void ImGui_ImplDX12_RenderDrawData(ImDrawData* draw_data)
{
// Avoid rendering when minimized
if (draw_data->DisplaySize.x <= 0.0f || draw_data->DisplaySize.y <= 0.0f)
return;
// FIXME: I'm assuming that this only gets called once per frame!
// If not, we can't just re-allocate the IB or VB, we'll have to do a proper allocator.
ImGui_ImplDX12_Data* bd = ImGui_ImplDX12_GetBackendData();
const u32 needed_vb = draw_data->TotalVtxCount * sizeof(ImDrawVert);
const u32 needed_ib = draw_data->TotalIdxCount * sizeof(ImDrawIdx);
if (!bd->VertexStreamBuffer.ReserveMemory(needed_vb, sizeof(ImDrawVert)) ||
!bd->IndexStreamBuffer.ReserveMemory(needed_ib, sizeof(ImDrawIdx)))
{
g_d3d12_context->ExecuteCommandList(false);
if (!bd->VertexStreamBuffer.ReserveMemory(needed_vb, sizeof(ImDrawVert)) ||
!bd->IndexStreamBuffer.ReserveMemory(needed_ib, sizeof(ImDrawIdx)))
{
Panic("Failed to allocate space for imgui vertices/indices");
}
}
// Upload vertex/index data into a single contiguous GPU buffer
ImDrawVert* vtx_dst = (ImDrawVert*)bd->VertexStreamBuffer.GetCurrentHostPointer();
ImDrawIdx* idx_dst = (ImDrawIdx*)bd->IndexStreamBuffer.GetCurrentHostPointer();
for (int n = 0; n < draw_data->CmdListsCount; n++)
{
const ImDrawList* cmd_list = draw_data->CmdLists[n];
memcpy(vtx_dst, cmd_list->VtxBuffer.Data, cmd_list->VtxBuffer.Size * sizeof(ImDrawVert));
memcpy(idx_dst, cmd_list->IdxBuffer.Data, cmd_list->IdxBuffer.Size * sizeof(ImDrawIdx));
vtx_dst += cmd_list->VtxBuffer.Size;
idx_dst += cmd_list->IdxBuffer.Size;
}
// Setup desired DX state (must happen before commit, because it uses the offsets)
ID3D12GraphicsCommandList* ctx = g_d3d12_context->GetCommandList();
ImGui_ImplDX12_SetupRenderState(draw_data, ctx);
bd->VertexStreamBuffer.CommitMemory(needed_vb);
bd->IndexStreamBuffer.CommitMemory(needed_ib);
// Render command lists
// (Because we merged all buffers into a single one, we maintain our own offset into them)
int global_vtx_offset = 0;
int global_idx_offset = 0;
ImVec2 clip_off = draw_data->DisplayPos;
const D3D12::Texture* last_texture = nullptr;
for (int n = 0; n < draw_data->CmdListsCount; n++)
{
const ImDrawList* cmd_list = draw_data->CmdLists[n];
for (int cmd_i = 0; cmd_i < cmd_list->CmdBuffer.Size; cmd_i++)
{
const ImDrawCmd* pcmd = &cmd_list->CmdBuffer[cmd_i];
if (pcmd->UserCallback != NULL)
{
// User callback, registered via ImDrawList::AddCallback()
// (ImDrawCallback_ResetRenderState is a special callback value used by the user to request the renderer to reset render state.)
if (pcmd->UserCallback == ImDrawCallback_ResetRenderState)
ImGui_ImplDX12_SetupRenderState(draw_data, ctx);
else
pcmd->UserCallback(cmd_list, pcmd);
}
else
{
// Project scissor/clipping rectangles into framebuffer space
ImVec2 clip_min(pcmd->ClipRect.x - clip_off.x, pcmd->ClipRect.y - clip_off.y);
ImVec2 clip_max(pcmd->ClipRect.z - clip_off.x, pcmd->ClipRect.w - clip_off.y);
if (clip_max.x <= clip_min.x || clip_max.y <= clip_min.y)
continue;
// Apply Scissor/clipping rectangle, Bind texture, Draw
const D3D12_RECT r = { (LONG)clip_min.x, (LONG)clip_min.y, (LONG)clip_max.x, (LONG)clip_max.y };
const D3D12::Texture* tex = (D3D12::Texture*)pcmd->GetTexID();
if (tex && last_texture != tex)
{
#if 0
// for when we redo the descriptor stuff
D3D12::DescriptorHandle handle;
if (!g_d3d12_context->GetDescriptorAllocator().Allocate(1, &handle))
{
// ugh.
g_d3d12_context->ExecuteCommandList(false);
ctx = g_d3d12_context->GetCommandList();
ImGui_ImplDX12_SetupRenderState(draw_data, ctx);
if (!g_d3d12_context->GetDescriptorAllocator().Allocate(1, &handle))
Panic("Failed to allocate descriptor after cmdlist kick");
}
g_d3d12_context->GetDevice()->CopyDescriptorsSimple(1, handle, tex->GetSRVDescriptor(), D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV);
ctx->SetGraphicsRootDescriptorTable(1, handle);
#else
ctx->SetGraphicsRootDescriptorTable(1, tex->GetSRVDescriptor());
#endif
last_texture = tex;
}
ctx->RSSetScissorRects(1, &r);
ctx->DrawIndexedInstanced(pcmd->ElemCount, 1, pcmd->IdxOffset + global_idx_offset, pcmd->VtxOffset + global_vtx_offset, 0);
}
}
global_idx_offset += cmd_list->IdxBuffer.Size;
global_vtx_offset += cmd_list->VtxBuffer.Size;
}
}
bool ImGui_ImplDX12_CreateFontsTexture()
{
// Build texture atlas
ImGuiIO& io = ImGui::GetIO();
ImGui_ImplDX12_Data* bd = ImGui_ImplDX12_GetBackendData();
unsigned char* pixels;
int width, height;
io.Fonts->GetTexDataAsRGBA32(&pixels, &width, &height);
// Upload texture to graphics system
if (bd->FontTexture.GetWidth() != static_cast<u32>(width) || bd->FontTexture.GetHeight() != static_cast<u32>(height))
{
if (!bd->FontTexture.Create(width, height, 1, 1, 1, DXGI_FORMAT_R8G8B8A8_UNORM,
DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN,
D3D12_RESOURCE_FLAG_NONE))
{
return false;
}
}
#if 0
if (!bd->FontTexture.LoadData(g_d3d12_context->GetInitCommandList(), 0, 0, 0, width, height, pixels, width * sizeof(u32)))
return false;
#else
if (!bd->FontTexture.LoadData(0, 0, width, height, pixels, width * sizeof(u32)))
return false;
#endif
io.Fonts->SetTexID((ImTextureID)&bd->FontTexture);
return true;
}
bool ImGui_ImplDX12_CreateDeviceObjects()
{
ImGui_ImplDX12_Data* bd = ImGui_ImplDX12_GetBackendData();
if (bd->pPipelineState)
ImGui_ImplDX12_DestroyDeviceObjects();
// Create the root signature
{
D3D12_DESCRIPTOR_RANGE descRange = {};
descRange.RangeType = D3D12_DESCRIPTOR_RANGE_TYPE_SRV;
descRange.NumDescriptors = 1;
descRange.BaseShaderRegister = 0;
descRange.RegisterSpace = 0;
descRange.OffsetInDescriptorsFromTableStart = 0;
D3D12_ROOT_PARAMETER param[2] = {};
param[0].ParameterType = D3D12_ROOT_PARAMETER_TYPE_32BIT_CONSTANTS;
param[0].Constants.ShaderRegister = 0;
param[0].Constants.RegisterSpace = 0;
param[0].Constants.Num32BitValues = 16;
param[0].ShaderVisibility = D3D12_SHADER_VISIBILITY_VERTEX;
param[1].ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE;
param[1].DescriptorTable.NumDescriptorRanges = 1;
param[1].DescriptorTable.pDescriptorRanges = &descRange;
param[1].ShaderVisibility = D3D12_SHADER_VISIBILITY_PIXEL;
// Bilinear sampling is required by default. Set 'io.Fonts->Flags |= ImFontAtlasFlags_NoBakedLines' or 'style.AntiAliasedLinesUseTex = false' to allow point/nearest sampling.
D3D12_STATIC_SAMPLER_DESC staticSampler = {};
staticSampler.Filter = D3D12_FILTER_MIN_MAG_MIP_LINEAR;
staticSampler.AddressU = D3D12_TEXTURE_ADDRESS_MODE_CLAMP;
staticSampler.AddressV = D3D12_TEXTURE_ADDRESS_MODE_CLAMP;
staticSampler.AddressW = D3D12_TEXTURE_ADDRESS_MODE_CLAMP;
staticSampler.MipLODBias = 0.f;
staticSampler.MaxAnisotropy = 0;
staticSampler.ComparisonFunc = D3D12_COMPARISON_FUNC_ALWAYS;
staticSampler.BorderColor = D3D12_STATIC_BORDER_COLOR_TRANSPARENT_BLACK;
staticSampler.MinLOD = 0.f;
staticSampler.MaxLOD = 0.f;
staticSampler.ShaderRegister = 0;
staticSampler.RegisterSpace = 0;
staticSampler.ShaderVisibility = D3D12_SHADER_VISIBILITY_PIXEL;
D3D12_ROOT_SIGNATURE_DESC desc = {};
desc.NumParameters = _countof(param);
desc.pParameters = param;
desc.NumStaticSamplers = 1;
desc.pStaticSamplers = &staticSampler;
desc.Flags =
D3D12_ROOT_SIGNATURE_FLAG_ALLOW_INPUT_ASSEMBLER_INPUT_LAYOUT |
D3D12_ROOT_SIGNATURE_FLAG_DENY_HULL_SHADER_ROOT_ACCESS |
D3D12_ROOT_SIGNATURE_FLAG_DENY_DOMAIN_SHADER_ROOT_ACCESS |
D3D12_ROOT_SIGNATURE_FLAG_DENY_GEOMETRY_SHADER_ROOT_ACCESS;
auto blob = g_d3d12_context->SerializeRootSignature(&desc);
if (!blob)
return false;
g_d3d12_context->GetDevice()->CreateRootSignature(0, blob->GetBufferPointer(), blob->GetBufferSize(), IID_PPV_ARGS(&bd->pRootSignature));
}
// By using D3DCompile() from <d3dcompiler.h> / d3dcompiler.lib, we introduce a dependency to a given version of d3dcompiler_XX.dll (see D3DCOMPILER_DLL_A)
// If you would like to use this DX12 sample code but remove this dependency you can:
// 1) compile once, save the compiled shader blobs into a file or source code and pass them to CreateVertexShader()/CreatePixelShader() [preferred solution]
// 2) use code to detect any version of the DLL and grab a pointer to D3DCompile from the DLL.
// See https://github.com/ocornut/imgui/pull/638 for sources and details.
D3D12_GRAPHICS_PIPELINE_STATE_DESC psoDesc;
memset(&psoDesc, 0, sizeof(D3D12_GRAPHICS_PIPELINE_STATE_DESC));
psoDesc.NodeMask = 1;
psoDesc.PrimitiveTopologyType = D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE;
psoDesc.pRootSignature = bd->pRootSignature;
psoDesc.SampleMask = UINT_MAX;
psoDesc.NumRenderTargets = 1;
psoDesc.RTVFormats[0] = bd->RTVFormat;
psoDesc.SampleDesc.Count = 1;
psoDesc.Flags = D3D12_PIPELINE_STATE_FLAG_NONE;
ID3DBlob* vertexShaderBlob;
ID3DBlob* pixelShaderBlob;
// Create the vertex shader
{
static const char* vertexShader =
"cbuffer vertexBuffer : register(b0) \
{\
float4x4 ProjectionMatrix; \
};\
struct VS_INPUT\
{\
float2 pos : POSITION;\
float4 col : COLOR0;\
float2 uv : TEXCOORD0;\
};\
\
struct PS_INPUT\
{\
float4 pos : SV_POSITION;\
float4 col : COLOR0;\
float2 uv : TEXCOORD0;\
};\
\
PS_INPUT main(VS_INPUT input)\
{\
PS_INPUT output;\
output.pos = mul( ProjectionMatrix, float4(input.pos.xy, 0.f, 1.f));\
output.col = input.col;\
output.uv = input.uv;\
return output;\
}";
if (FAILED(D3DCompile(vertexShader, strlen(vertexShader), NULL, NULL, NULL, "main", "vs_5_0", 0, 0, &vertexShaderBlob, NULL)))
return false; // NB: Pass ID3D10Blob* pErrorBlob to D3DCompile() to get error showing in (const char*)pErrorBlob->GetBufferPointer(). Make sure to Release() the blob!
psoDesc.VS = { vertexShaderBlob->GetBufferPointer(), vertexShaderBlob->GetBufferSize() };
// Create the input layout
static D3D12_INPUT_ELEMENT_DESC local_layout[] =
{
{ "POSITION", 0, DXGI_FORMAT_R32G32_FLOAT, 0, (UINT)IM_OFFSETOF(ImDrawVert, pos), D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 0, DXGI_FORMAT_R32G32_FLOAT, 0, (UINT)IM_OFFSETOF(ImDrawVert, uv), D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "COLOR", 0, DXGI_FORMAT_R8G8B8A8_UNORM, 0, (UINT)IM_OFFSETOF(ImDrawVert, col), D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
};
psoDesc.InputLayout = { local_layout, 3 };
}
// Create the pixel shader
{
static const char* pixelShader =
"struct PS_INPUT\
{\
float4 pos : SV_POSITION;\
float4 col : COLOR0;\
float2 uv : TEXCOORD0;\
};\
SamplerState sampler0 : register(s0);\
Texture2D texture0 : register(t0);\
\
float4 main(PS_INPUT input) : SV_Target\
{\
float4 out_col = input.col * texture0.Sample(sampler0, input.uv); \
return out_col; \
}";
if (FAILED(D3DCompile(pixelShader, strlen(pixelShader), NULL, NULL, NULL, "main", "ps_5_0", 0, 0, &pixelShaderBlob, NULL)))
{
vertexShaderBlob->Release();
return false; // NB: Pass ID3D10Blob* pErrorBlob to D3DCompile() to get error showing in (const char*)pErrorBlob->GetBufferPointer(). Make sure to Release() the blob!
}
psoDesc.PS = { pixelShaderBlob->GetBufferPointer(), pixelShaderBlob->GetBufferSize() };
}
// Create the blending setup
{
D3D12_BLEND_DESC& desc = psoDesc.BlendState;
desc.AlphaToCoverageEnable = false;
desc.RenderTarget[0].BlendEnable = true;
desc.RenderTarget[0].SrcBlend = D3D12_BLEND_SRC_ALPHA;
desc.RenderTarget[0].DestBlend = D3D12_BLEND_INV_SRC_ALPHA;
desc.RenderTarget[0].BlendOp = D3D12_BLEND_OP_ADD;
desc.RenderTarget[0].SrcBlendAlpha = D3D12_BLEND_ONE;
desc.RenderTarget[0].DestBlendAlpha = D3D12_BLEND_INV_SRC_ALPHA;
desc.RenderTarget[0].BlendOpAlpha = D3D12_BLEND_OP_ADD;
desc.RenderTarget[0].RenderTargetWriteMask = D3D12_COLOR_WRITE_ENABLE_ALL;
}
// Create the rasterizer state
{
D3D12_RASTERIZER_DESC& desc = psoDesc.RasterizerState;
desc.FillMode = D3D12_FILL_MODE_SOLID;
desc.CullMode = D3D12_CULL_MODE_NONE;
desc.FrontCounterClockwise = FALSE;
desc.DepthBias = D3D12_DEFAULT_DEPTH_BIAS;
desc.DepthBiasClamp = D3D12_DEFAULT_DEPTH_BIAS_CLAMP;
desc.SlopeScaledDepthBias = D3D12_DEFAULT_SLOPE_SCALED_DEPTH_BIAS;
desc.DepthClipEnable = true;
desc.MultisampleEnable = FALSE;
desc.AntialiasedLineEnable = FALSE;
desc.ForcedSampleCount = 0;
desc.ConservativeRaster = D3D12_CONSERVATIVE_RASTERIZATION_MODE_OFF;
}
// Create depth-stencil State
{
D3D12_DEPTH_STENCIL_DESC& desc = psoDesc.DepthStencilState;
desc.DepthEnable = false;
desc.DepthWriteMask = D3D12_DEPTH_WRITE_MASK_ALL;
desc.DepthFunc = D3D12_COMPARISON_FUNC_ALWAYS;
desc.StencilEnable = false;
desc.FrontFace.StencilFailOp = desc.FrontFace.StencilDepthFailOp = desc.FrontFace.StencilPassOp = D3D12_STENCIL_OP_KEEP;
desc.FrontFace.StencilFunc = D3D12_COMPARISON_FUNC_ALWAYS;
desc.BackFace = desc.FrontFace;
}
HRESULT result_pipeline_state = g_d3d12_context->GetDevice()->CreateGraphicsPipelineState(&psoDesc, IID_PPV_ARGS(&bd->pPipelineState));
vertexShaderBlob->Release();
pixelShaderBlob->Release();
if (result_pipeline_state != S_OK)
return false;
return true;
}
void ImGui_ImplDX12_DestroyDeviceObjects()
{
ImGui_ImplDX12_Data* bd = ImGui_ImplDX12_GetBackendData();
if (!bd)
return;
ImGuiIO& io = ImGui::GetIO();
SafeRelease(bd->pRootSignature);
SafeRelease(bd->pPipelineState);
bd->FontTexture.Destroy(false);
bd->VertexStreamBuffer.Destroy(false);
bd->IndexStreamBuffer.Destroy(false);
io.Fonts->SetTexID(NULL); // We copied bd->pFontTextureView to io.Fonts->TexID so let's clear that as well.
}
bool ImGui_ImplDX12_Init(DXGI_FORMAT rtv_format)
{
ImGuiIO& io = ImGui::GetIO();
IM_ASSERT(io.BackendRendererUserData == NULL && "Already initialized a renderer backend!");
// Setup backend capabilities flags
ImGui_ImplDX12_Data* bd = IM_NEW(ImGui_ImplDX12_Data)();
io.BackendRendererUserData = (void*)bd;
io.BackendRendererName = "imgui_impl_dx12";
io.BackendFlags |= ImGuiBackendFlags_RendererHasVtxOffset; // We can honor the ImDrawCmd::VtxOffset field, allowing for large meshes.
bd->RTVFormat = rtv_format;
if (!bd->VertexStreamBuffer.Create(VERTEX_BUFFER_SIZE) || !bd->IndexStreamBuffer.Create(INDEX_BUFFER_SIZE))
return false;
return ImGui_ImplDX12_CreateDeviceObjects();
}
void ImGui_ImplDX12_Shutdown()
{
ImGui_ImplDX12_Data* bd = ImGui_ImplDX12_GetBackendData();
IM_ASSERT(bd != NULL && "No renderer backend to shutdown, or already shutdown?");
ImGuiIO& io = ImGui::GetIO();
ImGui_ImplDX12_DestroyDeviceObjects();
io.BackendRendererName = NULL;
io.BackendRendererUserData = NULL;
IM_DELETE(bd);
}
void ImGui_ImplDX12_NewFrame()
{
ImGui_ImplDX12_Data* bd = ImGui_ImplDX12_GetBackendData();
IM_ASSERT(bd != NULL && "Did you call ImGui_ImplDX12_Init()?");
if (!bd->pPipelineState)
ImGui_ImplDX12_CreateDeviceObjects();
}

14
src/core/gpu/imgui_impl_dx12.h
View File

@ -1,14 +0,0 @@
// dear imgui: Renderer Backend for DirectX12
// This needs to be used along with a Platform Backend (e.g. Win32)
#pragma once
#include "imgui.h" // IMGUI_IMPL_API
bool ImGui_ImplDX12_Init(DXGI_FORMAT rtv_format);
void ImGui_ImplDX12_Shutdown();
void ImGui_ImplDX12_RenderDrawData(ImDrawData* draw_data);
// Use if you want to reset your rendering device without losing Dear ImGui state.
void ImGui_ImplDX12_DestroyDeviceObjects();
bool ImGui_ImplDX12_CreateDeviceObjects();
bool ImGui_ImplDX12_CreateFontsTexture();

4
src/core/gpu/postprocessing_shadergen.cpp
View File

@ -15,8 +15,8 @@ std::string PostProcessingShaderGen::GeneratePostProcessingVertexShader(const Po
std::stringstream ss;
WriteHeader(ss);
DeclareTexture(ss, "samp0", 0);
WriteUniformBuffer(ss, shader, false);
DeclareTexture(ss, "samp0", 0);
DeclareVertexEntryPoint(ss, {}, 0, 1, {}, true);
ss << R"(
@ -38,8 +38,8 @@ std::string PostProcessingShaderGen::GeneratePostProcessingFragmentShader(const
std::stringstream ss;
WriteHeader(ss);
DeclareTexture(ss, "samp0", 0);
WriteUniformBuffer(ss, shader, false);
DeclareTexture(ss, "samp0", 0);
// Rename main, since we need to set up globals
if (!m_glsl)

12
src/core/gpu/vulkan_device.cpp
View File

@ -1222,17 +1222,22 @@ void VulkanDevice::SubmitCommandBuffer(bool wait_for_completion, const char* rea
const std::string reason_str(StringUtil::StdStringFromFormatV(reason, ap));
va_end(ap);
Log_WarningPrintf("Vulkan: Executing command buffer due to '%s'", reason_str.c_str());
Log_WarningPrintf("Executing command buffer due to '%s'", reason_str.c_str());
SubmitCommandBuffer(wait_for_completion);
}
void VulkanDevice::SubmitCommandBufferAndRestartRenderPass(const char* reason)
{
Log_WarningPrintf("Vulkan: Executing command buffer due to '%s'", reason);
if (InRenderPass())
EndRenderPass();
VulkanFramebuffer* fb = m_current_framebuffer;
VulkanPipeline* pl = m_current_pipeline;
SubmitCommandBuffer(false, "%s", reason);
if (fb)
SetFramebuffer(fb);
SetPipeline(pl);
BeginRenderPass();
}
@ -2539,9 +2544,6 @@ void VulkanDevice::BeginRenderPass()
{
DebugAssert(!InRenderPass());
if (m_current_render_pass != VK_NULL_HANDLE)
EndRenderPass();
VkRenderPassBeginInfo bi = {VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO, nullptr};
std::array<VkClearValue, 2> clear_values;

68
src/core/gpu/vulkan_texture.cpp
View File

@ -333,10 +333,6 @@ bool VulkanTexture::Update(u32 x, u32 y, u32 width, u32 height, const void* data
const VkCommandBuffer cmdbuf = GetCommandBufferForUpdate();
// first time the texture is used? don't leave it undefined
if (m_layout == Layout::Undefined)
TransitionToLayout(cmdbuf, Layout::TransferDst);
// if we're an rt and have been cleared, and the full rect isn't being uploaded, do the clear
if (m_type == Type::RenderTarget)
{
@ -346,6 +342,10 @@ bool VulkanTexture::Update(u32 x, u32 y, u32 width, u32 height, const void* data
m_state = State::Dirty;
}
// first time the texture is used? don't leave it undefined
if (m_layout == Layout::Undefined)
TransitionToLayout(cmdbuf, Layout::TransferDst);
UpdateFromBuffer(cmdbuf, x, y, width, height, layer, level, upload_pitch, buffer, buffer_offset);
TransitionToLayout(cmdbuf, Layout::ShaderReadOnly);
return true;
@ -416,23 +416,6 @@ void VulkanTexture::Unmap()
m_map_level = 0;
}
#if 0
void VulkanTexture::Swap(GSTexture* tex)
{
GSTexture::Swap(tex);
std::swap(m_image, static_cast<VulkanTexture*>(tex)->m_image);
std::swap(m_allocation, static_cast<VulkanTexture*>(tex)->m_allocation);
std::swap(m_view, static_cast<VulkanTexture*>(tex)->m_view);
std::swap(m_vk_format, static_cast<VulkanTexture*>(tex)->m_vk_format);
std::swap(m_layout, static_cast<VulkanTexture*>(tex)->m_layout);
std::swap(m_use_fence_counter, static_cast<VulkanTexture*>(tex)->m_use_fence_counter);
std::swap(m_clear_value, static_cast<VulkanTexture*>(tex)->m_clear_value);
std::swap(m_map_area, static_cast<VulkanTexture*>(tex)->m_map_area);
std::swap(m_map_level, static_cast<VulkanTexture*>(tex)->m_map_level);
std::swap(m_framebuffers, static_cast<VulkanTexture*>(tex)->m_framebuffers);
}
#endif
void VulkanTexture::CommitClear()
{
if (m_state != GPUTexture::State::Cleared)
@ -710,44 +693,6 @@ VkDescriptorSet VulkanTexture::GetDescriptorSetWithSampler(VkSampler sampler)
return ds;
}
#if 0
VkFramebuffer VulkanTexture::GetLinkedFramebuffer(VulkanTexture* depth_texture, bool feedback_loop)
{
pxAssertRel(m_type != Type::Texture, "Texture is a render target");
for (const auto& [other_tex, fb, other_feedback_loop] : m_framebuffers)
{
if (other_tex == depth_texture && other_feedback_loop == feedback_loop)
return fb;
}
const VkRenderPass rp = VulkanDevice::GetInstance()->GetRenderPass(
(m_type != GSTexture::Type::DepthStencil) ? m_vk_format : VK_FORMAT_UNDEFINED,
(m_type != GSTexture::Type::DepthStencil) ? (depth_texture ? depth_texture->m_vk_format : VK_FORMAT_UNDEFINED) :
m_vk_format,
VK_ATTACHMENT_LOAD_OP_LOAD, VK_ATTACHMENT_STORE_OP_STORE, VK_ATTACHMENT_LOAD_OP_LOAD, VK_ATTACHMENT_STORE_OP_STORE,
VK_ATTACHMENT_LOAD_OP_DONT_CARE, VK_ATTACHMENT_STORE_OP_DONT_CARE, feedback_loop);
if (!rp)
return VK_NULL_HANDLE;
Vulkan::FramebufferBuilder fbb;
fbb.AddAttachment(m_view);
if (depth_texture)
fbb.AddAttachment(depth_texture->m_view);
fbb.SetSize(m_size.x, m_size.y, 1);
fbb.SetRenderPass(rp);
VkFramebuffer fb = fbb.Create(VulkanDevice::GetInstance()->GetVulkanDevice());
if (!fb)
return VK_NULL_HANDLE;
m_framebuffers.emplace_back(depth_texture, fb, feedback_loop);
if (depth_texture)
depth_texture->m_framebuffers.emplace_back(this, fb, feedback_loop);
return fb;
}
#endif
void VulkanTexture::MakeReadyForSampling()
{
if (m_layout == Layout::ShaderReadOnly)
@ -778,6 +723,7 @@ bool VulkanDevice::DownloadTexture(GPUTexture* texture, u32 x, u32 y, u32 width,
u32 out_data_stride)
{
VulkanTexture* T = static_cast<VulkanTexture*>(texture);
T->CommitClear();
const u32 pitch = Common::AlignUp(width * T->GetPixelSize(), GetBufferCopyRowPitchAlignment());
const u32 size = pitch * height;
@ -835,9 +781,7 @@ bool VulkanDevice::DownloadTexture(GPUTexture* texture, u32 x, u32 y, u32 width,
if (res != VK_SUCCESS)
LOG_VULKAN_ERROR(res, "vmaInvalidateAllocation() failed, readback may be incorrect: ");
StringUtil::StrideMemCpy(out_data, out_data_stride, m_download_buffer_map, pitch, std::min(pitch, out_data_stride),
height);
StringUtil::StrideMemCpy(out_data, out_data_stride, m_download_buffer_map, pitch, width * T->GetPixelSize(), height);
return true;
}

2
src/core/gpu/window_info.cpp
View File

@ -25,8 +25,6 @@ void WindowInfo::SetSurfaceless()
#include "common/windows_headers.h"
#include <dwmapi.h>
#pragma comment(lib, "Dwmapi.lib")
static bool GetRefreshRateFromDWM(HWND hwnd, float* refresh_rate)
{
BOOL composition_enabled;

2
src/core/gpu/window_info.h
View File

@ -33,6 +33,8 @@ struct WindowInfo
void* surface_handle = nullptr;
#endif
ALWAYS_INLINE bool IsSurfaceless() const { return type == Type::Surfaceless; }
// Changes the window to be surfaceless (i.e. no handle/size/etc).
void SetSurfaceless();

7
src/core/gpu_hw.cpp
View File

@ -509,6 +509,12 @@ bool GPU_HW::CreateBuffers()
return false;
}
GL_OBJECT_NAME(m_vram_texture, "VRAM Texture");
GL_OBJECT_NAME(m_vram_depth_texture, "VRAM Depth Texture");
GL_OBJECT_NAME(m_vram_read_texture, "VRAM Read Texture");
GL_OBJECT_NAME(m_display_texture, "Display Texture");
GL_OBJECT_NAME(m_vram_readback_texture, "VRAM Readback Texture");
// vram framebuffer has both colour and depth
if (!(m_vram_framebuffer = g_gpu_device->CreateFramebuffer(m_vram_texture.get(), m_vram_depth_texture.get())) ||
!(m_vram_update_depth_framebuffer = g_gpu_device->CreateFramebuffer(m_vram_depth_texture.get())) ||
@ -972,6 +978,7 @@ bool GPU_HW::CompilePipelines()
if (!fs)
return false;
GL_OBJECT_NAME(fs, "Downsample Composite Pass Fragment Shader");
plconfig.layout = GPUPipeline::Layout::MultiTextureAndUBO;
plconfig.fragment_shader = fs.get();
plconfig.color_format = VRAM_RT_FORMAT;
if (!(m_downsample_composite_pass_pipeline = g_gpu_device->CreatePipeline(plconfig)))

1171
src/core/gpu_hw_d3d12.cpp
View File

File diff suppressed because it is too large Load Diff

107
src/core/gpu_hw_d3d12.h
View File

@ -1,107 +0,0 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#if 0
#include "common/dimensional_array.h"
#include "gpu/d3d12/staging_texture.h"
#include "gpu/d3d12/stream_buffer.h"
#include "gpu/d3d12/texture.h"
#include "gpu_hw.h"
#include "texture_replacements.h"
#include <array>
#include <memory>
#include <tuple>
class GPU_HW_D3D12 final : public GPU_HW
{
public:
template<typename T>
using ComPtr = Microsoft::WRL::ComPtr<T>;
GPU_HW_D3D12();
~GPU_HW_D3D12() override;
bool Initialize() override;
void Reset(bool clear_vram) override;
void RestoreGraphicsAPIState() override;
void UpdateSettings() override;
protected:
void ClearDisplay() override;
void UpdateDisplay() override;
void ReadVRAM(u32 x, u32 y, u32 width, u32 height) override;
void FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color) override;
void UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask, bool check_mask) override;
void CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height) override;
private:
enum : u32
{
MAX_PUSH_CONSTANTS_SIZE = 64,
TEXTURE_REPLACEMENT_BUFFER_SIZE = 64 * 1024 * 1024,
};
void SetCapabilities();
void DestroyResources();
bool CreateRootSignatures();
bool CreateSamplers();
bool CreateBuffers();
void ClearFramebuffer();
void DestroyBuffers();
bool CreateVertexBuffer();
bool CreateUniformBuffer();
bool CreateTextureBuffer();
bool CompilePipelines();
void DestroyPipelines();
bool CreateTextureReplacementStreamBuffer();
bool BlitVRAMReplacementTexture(const TextureReplacementTexture* tex, u32 dst_x, u32 dst_y, u32 width, u32 height);
ComPtr<ID3D12RootSignature> m_batch_root_signature;
ComPtr<ID3D12RootSignature> m_single_sampler_root_signature;
D3D12::Texture m_vram_texture;
D3D12::Texture m_vram_depth_texture;
D3D12::Texture m_vram_read_texture;
D3D12::Texture m_vram_readback_texture;
D3D12::StagingTexture m_vram_readback_staging_texture;
D3D12::Texture m_display_texture;
D3D12::DescriptorHandle m_point_sampler;
D3D12::DescriptorHandle m_linear_sampler;
D3D12::StreamBuffer m_vertex_stream_buffer;
D3D12::StreamBuffer m_uniform_stream_buffer;
D3D12::StreamBuffer m_texture_stream_buffer;
D3D12::DescriptorHandle m_texture_stream_buffer_srv;
u32 m_current_uniform_buffer_offset = 0;
// [depth_test][render_mode][texture_mode][transparency_mode][dithering][interlacing]
DimensionalArray<ComPtr<ID3D12PipelineState>, 2, 2, 5, 9, 4, 2> m_batch_pipelines;
// [wrapped][interlaced]
DimensionalArray<ComPtr<ID3D12PipelineState>, 2, 2> m_vram_fill_pipelines;
// [depth_test]
std::array<ComPtr<ID3D12PipelineState>, 2> m_vram_write_pipelines;
std::array<ComPtr<ID3D12PipelineState>, 2> m_vram_copy_pipelines;
ComPtr<ID3D12PipelineState> m_vram_readback_pipeline;
ComPtr<ID3D12PipelineState> m_vram_update_depth_pipeline;
// [depth_24][interlace_mode]
DimensionalArray<ComPtr<ID3D12PipelineState>, 3, 2> m_display_pipelines;
ComPtr<ID3D12PipelineState> m_copy_pipeline;
D3D12::Texture m_vram_write_replacement_texture;
D3D12::StreamBuffer m_texture_replacment_stream_buffer;
};
#endif

2
src/core/gpu_hw_shadergen.cpp
View File

@ -1417,8 +1417,8 @@ std::string GPU_HW_ShaderGen::GenerateAdaptiveDownsampleBlurFragmentShader()
std::stringstream ss;
WriteHeader(ss);
WriteCommonFunctions(ss);
DeclareTexture(ss, "samp0", 0, false);
WriteAdaptiveDownsampleUniformBuffer(ss);
DeclareTexture(ss, "samp0", 0, false);
// mipmap_blur.glsl ported from parallel-rsx.
DeclareFragmentEntryPoint(ss, 0, 1, {}, false, 1, false, false, false, false);

4
src/duckstation-qt/displaysettingswidget.cpp
View File

@ -13,7 +13,7 @@
// For enumerating adapters.
#ifdef _WIN32
#include "core/gpu/d3d11_device.h"
#include "core/gpu/d3d12_gpu_device.h"
#include "core/gpu/d3d12_device.h"
#endif
#ifdef WITH_VULKAN
#include "core/gpu/vulkan_device.h"
@ -200,7 +200,7 @@ void DisplaySettingsWidget::populateGPUAdaptersAndResolutions()
break;
case GPURenderer::HardwareD3D12:
aml = D3D12GPUDevice::StaticGetAdapterAndModeList();
aml = D3D12Device::StaticGetAdapterAndModeList();
break;
#endif
#ifdef WITH_VULKAN

1
src/frontend-common/cubeb_audio_stream.cpp
View File

@ -16,7 +16,6 @@ Log_SetChannel(CubebAudioStream);
#ifdef _WIN32
#include "common/windows_headers.h"
#include <objbase.h>
#pragma comment(lib, "Ole32.lib")
#endif
static void StateCallback(cubeb_stream* stream, void* user_ptr, cubeb_state state);