Merge pull request #13556 from hrydgard/d3d11-depth-texture

D3D11 depth texture support

Common/GPU/D3D11/thin3d_d3d11.cpp

@@ -1212,6 +1212,8 @@ public:
 			colorRTView->Release();
 		if (colorSRView)
 			colorSRView->Release();
+		if (depthSRView)
+			depthSRView->Release();
 		if (depthStencilTex)
 			depthStencilTex->Release();
 		if (depthStencilRTView)
@@ -1223,6 +1225,7 @@ public:
 	ID3D11Texture2D *colorTex = nullptr;
 	ID3D11RenderTargetView *colorRTView = nullptr;
 	ID3D11ShaderResourceView *colorSRView = nullptr;
+	ID3D11ShaderResourceView *depthSRView = nullptr;
 	DXGI_FORMAT colorFormat = DXGI_FORMAT_UNKNOWN;
 
 	ID3D11Texture2D *depthStencilTex = nullptr;
@@ -1273,11 +1276,11 @@ Framebuffer *D3D11DrawContext::CreateFramebuffer(const FramebufferDesc &desc) {
 		descDepth.Height = desc.height;
 		descDepth.MipLevels = 1;
 		descDepth.ArraySize = 1;
-		descDepth.Format = fb->depthStencilFormat;
+		descDepth.Format = DXGI_FORMAT_R24G8_TYPELESS;  // so we can create an R24X8 view of it.
 		descDepth.SampleDesc.Count = 1;
 		descDepth.SampleDesc.Quality = 0;
 		descDepth.Usage = D3D11_USAGE_DEFAULT;
-		descDepth.BindFlags = D3D11_BIND_DEPTH_STENCIL;
+		descDepth.BindFlags = D3D11_BIND_DEPTH_STENCIL | D3D11_BIND_SHADER_RESOURCE;
 		descDepth.CPUAccessFlags = 0;
 		descDepth.MiscFlags = 0;
 		hr = device_->CreateTexture2D(&descDepth, nullptr, &fb->depthStencilTex);
@@ -1286,7 +1289,7 @@ Framebuffer *D3D11DrawContext::CreateFramebuffer(const FramebufferDesc &desc) {
 			return nullptr;
 		}
 		D3D11_DEPTH_STENCIL_VIEW_DESC descDSV{};
-		descDSV.Format = descDepth.Format;
+		descDSV.Format = DXGI_FORMAT_D24_UNORM_S8_UINT;
 		descDSV.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2D;
 		descDSV.Texture2D.MipSlice = 0;
 		hr = device_->CreateDepthStencilView(fb->depthStencilTex, &descDSV, &fb->depthStencilRTView);
@@ -1294,6 +1297,17 @@ Framebuffer *D3D11DrawContext::CreateFramebuffer(const FramebufferDesc &desc) {
 			delete fb;
 			return nullptr;
 		}
+
+		D3D11_SHADER_RESOURCE_VIEW_DESC depthViewDesc{};
+		depthViewDesc.Format = DXGI_FORMAT_R24_UNORM_X8_TYPELESS;
+		depthViewDesc.Texture2D.MostDetailedMip = 0;
+		depthViewDesc.Texture2D.MipLevels = 1;
+		depthViewDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D;
+		hr = device_->CreateShaderResourceView(fb->depthStencilTex, &depthViewDesc, &fb->depthSRView);
+		if (FAILED(hr)) {
+			WARN_LOG(G3D, "Failed to create SRV for depth buffer.");
+			fb->depthSRView = nullptr;
+		}
 	}
 
 	return fb;
@@ -1587,7 +1601,18 @@ void D3D11DrawContext::BindFramebufferAsRenderTarget(Framebuffer *fbo, const Ren
 // color must be 0, for now.
 void D3D11DrawContext::BindFramebufferAsTexture(Framebuffer *fbo, int binding, FBChannel channelBit, int attachment) {
 	D3D11Framebuffer *fb = (D3D11Framebuffer *)fbo;
-	context_->PSSetShaderResources(binding, 1, &fb->colorSRView);
+	switch (channelBit) {
+	case FBChannel::FB_COLOR_BIT:
+		context_->PSSetShaderResources(binding, 1, &fb->colorSRView);
+		break;
+	case FBChannel::FB_DEPTH_BIT:
+		if (fb->depthSRView) {
+			context_->PSSetShaderResources(binding, 1, &fb->depthSRView);
+		}
+		break;
+	default:
+		break;
+	}
 }
 
 uintptr_t D3D11DrawContext::GetFramebufferAPITexture(Framebuffer *fbo, int channelBit, int attachment) {

GPU/Common/DepalettizeShaderCommon.cpp

@@ -38,6 +38,12 @@ void GenerateDepalShader300(char *buffer, GEBufferFormat pixelFormat, ShaderLang
 		WRITE(p, "SamplerState texSamp : register(s0);\n");
 		WRITE(p, "Texture2D<float4> tex : register(t0);\n");
 		WRITE(p, "Texture2D<float4> pal : register(t3);\n");
+		// Support for depth.
+		if (pixelFormat == GE_FORMAT_DEPTH16) {
+			WRITE(p, "cbuffer params : register(b0) {\n");
+			WRITE(p, "  float z_scale; float z_offset;\n");
+			WRITE(p, "};\n");
+		}
 	} else if (language == GLSL_VULKAN) {
 		WRITE(p, "#version 450\n");
 		WRITE(p, "#extension GL_ARB_separate_shader_objects : enable\n");
@@ -53,7 +59,6 @@ void GenerateDepalShader300(char *buffer, GEBufferFormat pixelFormat, ShaderLang
 			WRITE(p, "  float z_scale; float z_offset;\n");
 			WRITE(p, "};\n");
 		}
-
 	} else {
 		if (gl_extensions.IsGLES) {
 			WRITE(p, "#version 300 es\n");

GPU/D3D11/TextureCacheD3D11.cpp

@@ -25,6 +25,7 @@
 #include "Core/Reporting.h"
 #include "GPU/ge_constants.h"
 #include "GPU/GPUState.h"
+#include "GPU/Common/GPUStateUtils.h"
 #include "GPU/D3D11/FragmentShaderGeneratorD3D11.h"
 #include "GPU/D3D11/TextureCacheD3D11.h"
 #include "GPU/D3D11/FramebufferManagerD3D11.h"
@@ -39,6 +40,12 @@
 #include "ext/xxhash.h"
 #include "Common/Math/math_util.h"
 
+// For depth depal
+struct DepthPushConstants {
+	float z_scale;
+	float z_offset;
+	float pad[2];
+};
 
 #define INVALID_TEX (ID3D11ShaderResourceView *)(-1LL)
 
@@ -113,6 +120,10 @@ TextureCacheD3D11::TextureCacheD3D11(Draw::DrawContext *draw)
 	isBgraBackend_ = true;
 	lastBoundTexture = INVALID_TEX;
 
+	D3D11_BUFFER_DESC desc{ sizeof(DepthPushConstants), D3D11_USAGE_DYNAMIC, D3D11_BIND_CONSTANT_BUFFER, D3D11_CPU_ACCESS_WRITE };
+	HRESULT hr = device_->CreateBuffer(&desc, nullptr, &depalConstants_);
+	_dbg_assert_(SUCCEEDED(hr));
+
 	HRESULT result = 0;
 
 	SetupTextureDecoder();
@@ -121,6 +132,8 @@ TextureCacheD3D11::TextureCacheD3D11(Draw::DrawContext *draw)
 }
 
 TextureCacheD3D11::~TextureCacheD3D11() {
+	depalConstants_->Release();
+
 	// pFramebufferVertexDecl->Release();
 	Clear(true);
 }
@@ -355,8 +368,9 @@ void TextureCacheD3D11::ApplyTextureFramebuffer(VirtualFramebuffer *framebuffer,
 	ID3D11PixelShader *pshader = nullptr;
 	uint32_t clutMode = gstate.clutformat & 0xFFFFFF;
 	bool need_depalettize = IsClutFormat(texFormat);
+	bool depth = channel == NOTIFY_FB_DEPTH;
 	if (need_depalettize && !g_Config.bDisableSlowFramebufEffects) {
-		pshader = depalShaderCache_->GetDepalettizePixelShader(clutMode, framebuffer->drawnFormat);
+		pshader = depalShaderCache_->GetDepalettizePixelShader(clutMode, depth ? GE_FORMAT_DEPTH16 : framebuffer->drawnFormat);
 	}
 
 	if (pshader) {
@@ -382,8 +396,20 @@ void TextureCacheD3D11::ApplyTextureFramebuffer(VirtualFramebuffer *framebuffer,
 		draw_->BindFramebufferAsRenderTarget(depalFBO, { Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE }, "ApplyTextureFramebuffer_DepalShader");
 		context_->PSSetShaderResources(3, 1, &clutTexture);
 		context_->PSSetSamplers(3, 1, &stockD3D11.samplerPoint2DWrap);
-		framebufferManagerD3D11_->BindFramebufferAsColorTexture(0, framebuffer, BINDFBCOLOR_SKIP_COPY | BINDFBCOLOR_FORCE_SELF);
+		draw_->BindFramebufferAsTexture(framebuffer->fbo, 0, depth ? Draw::FB_DEPTH_BIT : Draw::FB_COLOR_BIT, 0);
 		context_->PSSetSamplers(0, 1, &stockD3D11.samplerPoint2DWrap);
+
+		if (depth) {
+			DepthScaleFactors scaleFactors = GetDepthScaleFactors();
+			DepthPushConstants push;
+			push.z_scale = scaleFactors.scale;
+			push.z_offset = scaleFactors.offset;
+			D3D11_MAPPED_SUBRESOURCE map;
+			context_->Map(depalConstants_, 0, D3D11_MAP_WRITE_DISCARD, 0, &map);
+			memcpy(map.pData, &push, sizeof(push));
+			context_->Unmap(depalConstants_, 0);
+			context_->PSSetConstantBuffers(0, 1, &depalConstants_);
+		}
 		shaderApply.Shade();
 
 		context_->PSSetShaderResources(0, 1, &nullTexture);  // Make D3D11 validation happy. Really of no consequence since we rebind anyway.

GPU/D3D11/TextureCacheD3D11.h

@@ -91,6 +91,7 @@ private:
 	SamplerCacheD3D11 samplerCache_;
 
 	ID3D11ShaderResourceView *lastBoundTexture;
+	ID3D11Buffer *depalConstants_;
 
 	int decimationCounter_;
 	int texelsScaledThisFrame_;
@@ -99,6 +100,7 @@ private:
 	FramebufferManagerD3D11 *framebufferManagerD3D11_;
 	DepalShaderCacheD3D11 *depalShaderCache_;
 	ShaderManagerD3D11 *shaderManager_;
+
 };
 
 DXGI_FORMAT GetClutDestFormatD3D11(GEPaletteFormat format);