Merge pull request #13626 from hrydgard/more-framebuffer-cleanup

More framebuffer code cleanup

Common/GPU/D3D9/thin3d_d3d9.cpp

@@ -1081,7 +1081,6 @@ public:
 
 	int width;
 	int height;
-	FBColorDepth colorDepth;
 };
 
 Framebuffer *D3D9Context::CreateFramebuffer(const FramebufferDesc &desc) {
@@ -1090,7 +1089,6 @@ Framebuffer *D3D9Context::CreateFramebuffer(const FramebufferDesc &desc) {
 	D3D9Framebuffer *fbo = new D3D9Framebuffer{};
 	fbo->width = desc.width;
 	fbo->height = desc.height;
-	fbo->colorDepth = desc.colorDepth;
 	fbo->depthstenciltex = nullptr;
 
 	HRESULT rtResult = device_->CreateTexture(fbo->width, fbo->height, 1, D3DUSAGE_RENDERTARGET, D3DFMT_A8R8G8B8, D3DPOOL_DEFAULT, &fbo->tex, NULL);

Common/GPU/OpenGL/GLQueueRunner.cpp

@@ -1434,7 +1434,6 @@ void GLQueueRunner::PerformReadback(const GLRStep &pass) {
 	if (convert && tempBuffer_ && readbackBuffer_) {
 		ConvertFromRGBA8888(readbackBuffer_, tempBuffer_, pixelStride, pixelStride, rect.w, rect.h, pass.readback.dstFormat);
 	}
-
 	CHECK_GL_ERROR_IF_DEBUG();
 }
 

Common/GPU/OpenGL/thin3d_gl.cpp

@@ -742,8 +742,6 @@ public:
 
 	GLRenderManager *render_;
 	GLRFramebuffer *framebuffer = nullptr;
-
-	FBColorDepth colorDepth = FBO_8888;
 };
 
 // TODO: SSE/NEON optimize, and move to ColorConv.cpp.

Common/GPU/thin3d.h

@@ -244,13 +244,6 @@ enum class NativeObject {
 	NULL_IMAGEVIEW,
 };
 
-enum FBColorDepth {
-	FBO_8888,
-	FBO_565,
-	FBO_4444,
-	FBO_5551,
-};
-
 enum FBChannel {
 	FB_COLOR_BIT = 1,
 	FB_DEPTH_BIT = 2,
@@ -292,7 +285,6 @@ struct FramebufferDesc {
 	int depth;
 	int numColorAttachments;
 	bool z_stencil;
-	FBColorDepth colorDepth;
 	const char *tag;  // For graphics debuggers
 };
 

GPU/Common/FramebufferManagerCommon.cpp

@@ -1072,7 +1072,6 @@ void FramebufferManagerCommon::ResizeFramebufFBO(VirtualFramebuffer *vfb, int w,
 	// During hardware rendering, we always render at full color depth even if the game wouldn't on real hardware.
 	// It's not worth the trouble trying to support lower bit-depth rendering, just
 	// more cases to test that nobody will ever use.
-	vfb->colorDepth = Draw::FBO_8888;
 
 	textureCache_->ForgetLastTexture();
 
@@ -1091,7 +1090,7 @@ void FramebufferManagerCommon::ResizeFramebufFBO(VirtualFramebuffer *vfb, int w,
 	shaderManager_->DirtyLastShader();
 	char tag[256];
 	snprintf(tag, sizeof(tag), "%08x_%08x_%dx%d_%s", vfb->fb_address, vfb->z_address, w, h, GeBufferFormatToString(vfb->format));
-	vfb->fbo = draw_->CreateFramebuffer({ vfb->renderWidth, vfb->renderHeight, 1, 1, true, (Draw::FBColorDepth)vfb->colorDepth, tag });
+	vfb->fbo = draw_->CreateFramebuffer({ vfb->renderWidth, vfb->renderHeight, 1, 1, true, tag });
 	if (old.fbo) {
 		INFO_LOG(FRAMEBUF, "Resizing FBO for %08x : %dx%dx%s", vfb->fb_address, w, h, GeBufferFormatToString(vfb->format));
 		if (vfb->fbo) {
@@ -1380,7 +1379,7 @@ VirtualFramebuffer *FramebufferManagerCommon::CreateRAMFramebuffer(uint32_t fbAd
 	char name[64];
 	snprintf(name, sizeof(name), "%08x_color_RAM", vfb->fb_address);
 	textureCache_->NotifyFramebuffer(vfb, NOTIFY_FB_CREATED);
-	vfb->fbo = draw_->CreateFramebuffer({ vfb->renderWidth, vfb->renderHeight, 1, 1, true, (Draw::FBColorDepth)vfb->colorDepth, name });
+	vfb->fbo = draw_->CreateFramebuffer({ vfb->renderWidth, vfb->renderHeight, 1, 1, true, name });
 	vfbs_.push_back(vfb);
 
 	u32 byteSize = ColorBufferByteSize(vfb);
@@ -1429,16 +1428,17 @@ VirtualFramebuffer *FramebufferManagerCommon::FindDownloadTempBuffer(VirtualFram
 		nvfb->drawnWidth = vfb->drawnWidth;
 		nvfb->drawnHeight = vfb->drawnHeight;
 		nvfb->drawnFormat = vfb->format;
-		nvfb->colorDepth = vfb->colorDepth;
 
-		if (!CreateDownloadTempBuffer(nvfb)) {
+		char name[64];
+		snprintf(name, sizeof(name), "download_temp");
+		nvfb->fbo = draw_->CreateFramebuffer({ nvfb->bufferWidth, nvfb->bufferHeight, 1, 1, false, name });
+		if (!nvfb->fbo) {
 			delete nvfb;
+			ERROR_LOG(FRAMEBUF, "Error creating download temp FBO! %d x %d", nvfb->renderWidth, nvfb->renderHeight);
 			return nullptr;
 		}
 
 		bvfbs_.push_back(nvfb);
-	} else {
-		UpdateDownloadTempBuffer(nvfb);
 	}
 
 	nvfb->usageFlags |= FB_USAGE_RENDERTARGET;
@@ -1448,36 +1448,6 @@ VirtualFramebuffer *FramebufferManagerCommon::FindDownloadTempBuffer(VirtualFram
 	return nvfb;
 }
 
-bool FramebufferManagerCommon::CreateDownloadTempBuffer(VirtualFramebuffer *nvfb) {
-	// When updating VRAM, it need to be exact format.
-	if (!gstate_c.Supports(GPU_PREFER_CPU_DOWNLOAD)) {
-		switch (nvfb->format) {
-		case GE_FORMAT_4444:
-			nvfb->colorDepth = Draw::FBO_4444;
-			break;
-		case GE_FORMAT_5551:
-			nvfb->colorDepth = Draw::FBO_5551;
-			break;
-		case GE_FORMAT_565:
-			nvfb->colorDepth = Draw::FBO_565;
-			break;
-		case GE_FORMAT_8888:
-		default:
-			nvfb->colorDepth = Draw::FBO_8888;
-			break;
-		}
-	}
-
-	char name[64];
-	snprintf(name, sizeof(name), "download_temp");
-	nvfb->fbo = draw_->CreateFramebuffer({ nvfb->bufferWidth, nvfb->bufferHeight, 1, 1, false, (Draw::FBColorDepth)nvfb->colorDepth, name });
-	if (!nvfb->fbo) {
-		ERROR_LOG(FRAMEBUF, "Error creating FBO! %d x %d", nvfb->renderWidth, nvfb->renderHeight);
-		return false;
-	}
-	return true;
-}
-
 void FramebufferManagerCommon::ApplyClearToMemory(int x1, int y1, int x2, int y2, u32 clearColor) {
 	if (currentRenderVfb_) {
 		if ((currentRenderVfb_->usageFlags & FB_USAGE_DOWNLOAD_CLEAR) != 0) {
@@ -1767,8 +1737,8 @@ void FramebufferManagerCommon::DestroyAllFBOs() {
 	tempFBOs_.clear();
 }
 
-Draw::Framebuffer *FramebufferManagerCommon::GetTempFBO(TempFBO reason, u16 w, u16 h, Draw::FBColorDepth color_depth) {
-	u64 key = ((u64)reason << 48) | ((u64)color_depth << 32) | ((u32)w << 16) | h;
+Draw::Framebuffer *FramebufferManagerCommon::GetTempFBO(TempFBO reason, u16 w, u16 h) {
+	u64 key = ((u64)reason << 48) | ((u32)w << 16) | h;
 	auto it = tempFBOs_.find(key);
 	if (it != tempFBOs_.end()) {
 		it->second.last_frame_used = gpuStats.numFlips;
@@ -1778,7 +1748,7 @@ Draw::Framebuffer *FramebufferManagerCommon::GetTempFBO(TempFBO reason, u16 w, u
 	bool z_stencil = reason == TempFBO::STENCIL;
 	char name[128];
 	snprintf(name, sizeof(name), "temp_fbo_%dx%d%s", w, h, z_stencil ? "_depth" : "");
-	Draw::Framebuffer *fbo = draw_->CreateFramebuffer({ w, h, 1, 1, z_stencil, color_depth, name });
+	Draw::Framebuffer *fbo = draw_->CreateFramebuffer({ w, h, 1, 1, z_stencil, name });
 	if (!fbo) {
 		return nullptr;
 	}

GPU/Common/FramebufferManagerCommon.h

@@ -78,8 +78,6 @@ struct VirtualFramebuffer {
 	u16 newHeight;
 	int lastFrameNewSize;
 
-	// TODO: Handle fbo and colorDepth better.
-	u8 colorDepth;
 	Draw::Framebuffer *fbo;
 
 	u16 drawnWidth;
@@ -304,7 +302,7 @@ public:
 	virtual void Resized();
 	virtual void DestroyAllFBOs();
 
-	Draw::Framebuffer *GetTempFBO(TempFBO reason, u16 w, u16 h, Draw::FBColorDepth colorDepth = Draw::FBO_8888);
+	Draw::Framebuffer *GetTempFBO(TempFBO reason, u16 w, u16 h);
 
 	// Debug features
 	virtual bool GetFramebuffer(u32 fb_address, int fb_stride, GEBufferFormat format, GPUDebugBuffer &buffer, int maxRes);
@@ -349,8 +347,6 @@ protected:
 	void DownloadFramebufferOnSwitch(VirtualFramebuffer *vfb);
 	void FindTransferFramebuffers(VirtualFramebuffer *&dstBuffer, VirtualFramebuffer *&srcBuffer, u32 dstBasePtr, int dstStride, int &dstX, int &dstY, u32 srcBasePtr, int srcStride, int &srcX, int &srcY, int &srcWidth, int &srcHeight, int &dstWidth, int &dstHeight, int bpp);
 	VirtualFramebuffer *FindDownloadTempBuffer(VirtualFramebuffer *vfb);
-	virtual bool CreateDownloadTempBuffer(VirtualFramebuffer *nvfb);
-	virtual void UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) = 0;
 
 	VirtualFramebuffer *CreateRAMFramebuffer(uint32_t fbAddress, int width, int height, int stride, GEBufferFormat format);
 

GPU/Common/PresentationCommon.cpp

@@ -317,7 +317,7 @@ bool PresentationCommon::AllocateFramebuffer(int w, int h) {
 	}
 
 	// No depth/stencil for post processing
-	Draw::Framebuffer *fbo = draw_->CreateFramebuffer({ w, h, 1, 1, false, Draw::FBO_8888, "presentation" });
+	Draw::Framebuffer *fbo = draw_->CreateFramebuffer({ w, h, 1, 1, false, "presentation" });
 	if (!fbo) {
 		return false;
 	}

GPU/D3D11/FramebufferManagerD3D11.cpp

@@ -331,7 +331,7 @@ void FramebufferManagerD3D11::BindFramebufferAsColorTexture(int stage, VirtualFr
 	// Currently rendering to this framebuffer. Need to make a copy.
 	if (!skipCopy && framebuffer == currentRenderVfb_) {
 		// TODO: Maybe merge with bvfbs_?  Not sure if those could be packing, and they're created at a different size.
-		Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight, (Draw::FBColorDepth)framebuffer->colorDepth);
+		Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight);
 		if (renderCopy) {
 			VirtualFramebuffer copyInfo = *framebuffer;
 			copyInfo.fbo = renderCopy;
@@ -353,10 +353,6 @@ void FramebufferManagerD3D11::BindFramebufferAsColorTexture(int stage, VirtualFr
 	}
 }
 
-void FramebufferManagerD3D11::UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) {
-	// Nothing to do here.
-}
-
 void FramebufferManagerD3D11::SimpleBlit(
 	Draw::Framebuffer *dest, float destX1, float destY1, float destX2, float destY2,
 	Draw::Framebuffer *src, float srcX1, float srcY1, float srcX2, float srcY2, bool linearFilter) {
@@ -445,7 +441,7 @@ void FramebufferManagerD3D11::BlitFramebuffer(VirtualFramebuffer *dst, int dstX,
 	// Direct3D doesn't support rect -> self.
 	Draw::Framebuffer *srcFBO = src->fbo;
 	if (src == dst) {
-		Draw::Framebuffer *tempFBO = GetTempFBO(TempFBO::BLIT, src->renderWidth, src->renderHeight, (Draw::FBColorDepth)src->colorDepth);
+		Draw::Framebuffer *tempFBO = GetTempFBO(TempFBO::BLIT, src->renderWidth, src->renderHeight);
 		SimpleBlit(tempFBO, dstX1, dstY1, dstX2, dstY2, src->fbo, srcX1, srcY1, srcX2, srcY2, false);
 		srcFBO = tempFBO;
 	}

GPU/D3D11/FramebufferManagerD3D11.h

@@ -58,8 +58,6 @@ protected:
 	// Used by ReadFramebufferToMemory and later framebuffer block copies
 	void BlitFramebuffer(VirtualFramebuffer *dst, int dstX, int dstY, VirtualFramebuffer *src, int srcX, int srcY, int w, int h, int bpp) override;
 
-	void UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) override;
-
 private:
 	void Bind2DShader() override;
 	void PackDepthbuffer(VirtualFramebuffer *vfb, int x, int y, int w, int h);

GPU/D3D11/GPU_D3D11.cpp

@@ -129,7 +129,6 @@ void GPU_D3D11::CheckGPUFeatures() {
 		features |= GPU_SUPPORTS_VS_RANGE_CULLING;
 	}
 	features |= GPU_SUPPORTS_BLEND_MINMAX;
-	features |= GPU_PREFER_CPU_DOWNLOAD;
 
 	// Accurate depth is required because the Direct3D API does not support inverse Z.
 	// So we cannot incorrectly use the viewport transform as the depth range on Direct3D.

GPU/D3D11/TextureCacheD3D11.cpp

@@ -377,7 +377,7 @@ void TextureCacheD3D11::ApplyTextureFramebuffer(VirtualFramebuffer *framebuffer,
 		const GEPaletteFormat clutFormat = gstate.getClutPaletteFormat();
 		ID3D11ShaderResourceView *clutTexture = depalShaderCache_->GetClutTexture(clutFormat, clutHash_, clutBuf_, expand32);
 
-		Draw::Framebuffer *depalFBO = framebufferManagerD3D11_->GetTempFBO(TempFBO::DEPAL, framebuffer->renderWidth, framebuffer->renderHeight, Draw::FBO_8888);
+		Draw::Framebuffer *depalFBO = framebufferManagerD3D11_->GetTempFBO(TempFBO::DEPAL, framebuffer->renderWidth, framebuffer->renderHeight);
 		shaderManager_->DirtyLastShader();
 
 		// Not sure why or if we need this here - we're not about to actually draw using draw_, just use its framebuffer binds.

GPU/Directx9/FramebufferManagerDX9.cpp

@@ -355,7 +355,7 @@ static const D3DVERTEXELEMENT9 g_FramebufferVertexElements[] = {
 		}
 		if (!skipCopy && framebuffer == currentRenderVfb_) {
 			// TODO: Maybe merge with bvfbs_?  Not sure if those could be packing, and they're created at a different size.
-			Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight, (Draw::FBColorDepth)framebuffer->colorDepth);
+			Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight);
 			if (renderCopy) {
 				VirtualFramebuffer copyInfo = *framebuffer;
 				copyInfo.fbo = renderCopy;
@@ -371,10 +371,6 @@ static const D3DVERTEXELEMENT9 g_FramebufferVertexElements[] = {
 		}
 	}
 
-	void FramebufferManagerDX9::UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) {
-		// Nothing to do here.
-	}
-
 	void FramebufferManagerDX9::BlitFramebuffer(VirtualFramebuffer *dst, int dstX, int dstY, VirtualFramebuffer *src, int srcX, int srcY, int w, int h, int bpp) {
 		if (!dst->fbo || !src->fbo || !useBufferedRendering_) {
 			// This can happen if we recently switched from non-buffered.
@@ -408,7 +404,7 @@ static const D3DVERTEXELEMENT9 g_FramebufferVertexElements[] = {
 		// Direct3D 9 doesn't support rect -> self.
 		Draw::Framebuffer *srcFBO = src->fbo;
 		if (src == dst) {
-			Draw::Framebuffer *tempFBO = GetTempFBO(TempFBO::BLIT, src->renderWidth, src->renderHeight, (Draw::FBColorDepth)src->colorDepth);
+			Draw::Framebuffer *tempFBO = GetTempFBO(TempFBO::BLIT, src->renderWidth, src->renderHeight);
 			bool result = draw_->BlitFramebuffer(
 				src->fbo, srcX1, srcY1, srcX2, srcY2,
 				tempFBO, dstX1, dstY1, dstX2, dstY2,
@@ -606,7 +602,7 @@ static const D3DVERTEXELEMENT9 g_FramebufferVertexElements[] = {
 				// Let's resize.  We must stretch to a render target first.
 				w = vfb->width * maxRes;
 				h = vfb->height * maxRes;
-				tempFBO = draw_->CreateFramebuffer({ w, h, 1, 1, false, Draw::FBO_8888 });
+				tempFBO = draw_->CreateFramebuffer({ w, h, 1, 1, false });
 				if (draw_->BlitFramebuffer(vfb->fbo, 0, 0, vfb->renderWidth, vfb->renderHeight, tempFBO, 0, 0, w, h, Draw::FB_COLOR_BIT, g_Config.iBufFilter == SCALE_LINEAR ? Draw::FB_BLIT_LINEAR : Draw::FB_BLIT_NEAREST, "GetFramebuffer")) {
 					renderTarget = (LPDIRECT3DSURFACE9)draw_->GetFramebufferAPITexture(tempFBO, Draw::FB_COLOR_BIT | Draw::FB_SURFACE_BIT, 0);
 				}

GPU/Directx9/FramebufferManagerDX9.h

@@ -69,8 +69,6 @@ protected:
 	// Used by ReadFramebufferToMemory and later framebuffer block copies
 	void BlitFramebuffer(VirtualFramebuffer *dst, int dstX, int dstY, VirtualFramebuffer *src, int srcX, int srcY, int w, int h, int bpp) override;
 
-	void UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) override;
-
 private:
 	void PackFramebufferSync_(VirtualFramebuffer *vfb, int x, int y, int w, int h) override;
 	void PackDepthbuffer(VirtualFramebuffer *vfb, int x, int y, int w, int h);

GPU/Directx9/GPU_DX9.cpp

@@ -164,7 +164,6 @@ void GPU_DX9::CheckGPUFeatures() {
 	features |= GPU_SUPPORTS_16BIT_FORMATS;
 	features |= GPU_SUPPORTS_BLEND_MINMAX;
 	features |= GPU_SUPPORTS_TEXTURE_LOD_CONTROL;
-	features |= GPU_PREFER_CPU_DOWNLOAD;
 
 	// Accurate depth is required because the Direct3D API does not support inverse Z.
 	// So we cannot incorrectly use the viewport transform as the depth range on Direct3D.

GPU/Directx9/TextureCacheDX9.cpp

@@ -344,7 +344,7 @@ void TextureCacheDX9::ApplyTextureFramebuffer(VirtualFramebuffer *framebuffer, G
 		const GEPaletteFormat clutFormat = gstate.getClutPaletteFormat();
 		LPDIRECT3DTEXTURE9 clutTexture = depalShaderCache_->GetClutTexture(clutFormat, clutHash_, clutBuf_);
 
-		Draw::Framebuffer *depalFBO = framebufferManagerDX9_->GetTempFBO(TempFBO::DEPAL, framebuffer->renderWidth, framebuffer->renderHeight, Draw::FBO_8888);
+		Draw::Framebuffer *depalFBO = framebufferManagerDX9_->GetTempFBO(TempFBO::DEPAL, framebuffer->renderWidth, framebuffer->renderHeight);
 		draw_->BindFramebufferAsRenderTarget(depalFBO, { Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE }, "Depal");
 		shaderManager_->DirtyLastShader();
 

GPU/GLES/DepthBufferGLES.cpp

@@ -126,7 +126,7 @@ void FramebufferManagerGLES::PackDepthbuffer(VirtualFramebuffer *vfb, int x, int
 		}
 
 		shaderManagerGL_->DirtyLastShader();
-		auto *blitFBO = GetTempFBO(TempFBO::COPY, vfb->renderWidth, vfb->renderHeight, Draw::FBO_8888);
+		auto *blitFBO = GetTempFBO(TempFBO::COPY, vfb->renderWidth, vfb->renderHeight);
 		draw_->BindFramebufferAsRenderTarget(blitFBO, { Draw::RPAction::CLEAR, Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE }, "PackDepthbuffer");
 		render_->SetViewport({ 0, 0, (float)vfb->renderWidth, (float)vfb->renderHeight, 0.0f, 1.0f });
 

GPU/GLES/FramebufferManagerGLES.cpp

@@ -278,7 +278,7 @@ void FramebufferManagerGLES::BindFramebufferAsColorTexture(int stage, VirtualFra
 	}
 	if (!skipCopy && framebuffer == currentRenderVfb_) {
 		// TODO: Maybe merge with bvfbs_?  Not sure if those could be packing, and they're created at a different size.
-		Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight, (Draw::FBColorDepth)framebuffer->colorDepth);
+		Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight);
 		if (renderCopy) {
 			VirtualFramebuffer copyInfo = *framebuffer;
 			copyInfo.fbo = renderCopy;
@@ -293,18 +293,6 @@ void FramebufferManagerGLES::BindFramebufferAsColorTexture(int stage, VirtualFra
 	}
 }
 
-void FramebufferManagerGLES::UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) {
-	_assert_msg_(nvfb->fbo, "Expecting a valid nvfb in UpdateDownloadTempBuffer");
-
-	// Discard the previous contents of this buffer where possible.
-	if (gl_extensions.GLES3) {
-		draw_->BindFramebufferAsRenderTarget(nvfb->fbo, { Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE }, "UpdateDownloadTempBuffer");
-	} else if (gl_extensions.IsGLES) {
-		draw_->BindFramebufferAsRenderTarget(nvfb->fbo, { Draw::RPAction::CLEAR, Draw::RPAction::CLEAR, Draw::RPAction::CLEAR }, "UpdateDownloadTempBuffer");
-		gstate_c.Dirty(DIRTY_BLEND_STATE);
-	}
-}
-
 void FramebufferManagerGLES::BlitFramebuffer(VirtualFramebuffer *dst, int dstX, int dstY, VirtualFramebuffer *src, int srcX, int srcY, int w, int h, int bpp) {
 	if (!dst->fbo || !src->fbo || !useBufferedRendering_) {
 		// This can happen if they recently switched from non-buffered.
@@ -347,12 +335,11 @@ void FramebufferManagerGLES::BlitFramebuffer(VirtualFramebuffer *dst, int dstX,
 		// glBlitFramebuffer can clip, but glCopyImageSubData is more restricted.
 		// In case the src goes outside, we just skip the optimization in that case.
 		const bool sameSize = dstX2 - dstX1 == srcX2 - srcX1 && dstY2 - dstY1 == srcY2 - srcY1;
-		const bool sameDepth = dst->colorDepth == src->colorDepth;
 		const bool srcInsideBounds = srcX2 <= src->renderWidth && srcY2 <= src->renderHeight;
 		const bool dstInsideBounds = dstX2 <= dst->renderWidth && dstY2 <= dst->renderHeight;
 		const bool xOverlap = src == dst && srcX2 > dstX1 && srcX1 < dstX2;
 		const bool yOverlap = src == dst && srcY2 > dstY1 && srcY1 < dstY2;
-		if (sameSize && sameDepth && srcInsideBounds && dstInsideBounds && !(xOverlap && yOverlap)) {
+		if (sameSize && srcInsideBounds && dstInsideBounds && !(xOverlap && yOverlap)) {
 			draw_->CopyFramebufferImage(src->fbo, 0, srcX1, srcY1, 0, dst->fbo, 0, dstX1, dstY1, 0, dstX2 - dstX1, dstY2 - dstY1, 1, Draw::FB_COLOR_BIT, "BlitFramebuffer");
 			return;
 		}

GPU/GLES/FramebufferManagerGLES.h

@@ -62,8 +62,6 @@ protected:
 	// Used by ReadFramebufferToMemory and later framebuffer block copies
 	void BlitFramebuffer(VirtualFramebuffer *dst, int dstX, int dstY, VirtualFramebuffer *src, int srcX, int srcY, int w, int h, int bpp) override;
 
-	void UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) override;
-
 private:
 	void CreateDeviceObjects();
 	void DestroyDeviceObjects();

GPU/GLES/GPU_GLES.cpp

@@ -189,21 +189,6 @@ void GPU_GLES::CheckGPUFeatures() {
 		features |= GPU_SUPPORTS_VAO;
 	}
 
-	bool useCPU = false;
-	if (!gl_extensions.IsGLES) {
-		// Some cards or drivers seem to always dither when downloading a framebuffer to 16-bit.
-		// This causes glitches in games that expect the exact values.
-		// It has not been experienced on NVIDIA cards, so those are left using the GPU (which is faster.)
-		if (gl_extensions.gpuVendor != GPU_VENDOR_NVIDIA || !gl_extensions.VersionGEThan(3, 0)) {
-			useCPU = true;
-		}
-	} else {
-		useCPU = true;
-	}
-
-	if (useCPU)
-		features |= GPU_PREFER_CPU_DOWNLOAD;
-
 	if ((gl_extensions.gpuVendor == GPU_VENDOR_NVIDIA) || (gl_extensions.gpuVendor == GPU_VENDOR_AMD))
 		features |= GPU_PREFER_REVERSE_COLOR_ORDER;
 

GPU/GLES/StencilBufferGLES.cpp

@@ -170,7 +170,7 @@ bool FramebufferManagerGLES::NotifyStencilUpload(u32 addr, int size, StencilUplo
 
 	Draw::Framebuffer *blitFBO = nullptr;
 	if (useBlit) {
-		blitFBO = GetTempFBO(TempFBO::STENCIL, w, h, Draw::FBO_8888);
+		blitFBO = GetTempFBO(TempFBO::STENCIL, w, h);
 		draw_->BindFramebufferAsRenderTarget(blitFBO, { Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE }, "NotifyStencilUpload_Blit");
 	} else if (dstBuffer->fbo) {
 		draw_->BindFramebufferAsRenderTarget(dstBuffer->fbo, { Draw::RPAction::KEEP, Draw::RPAction::KEEP, Draw::RPAction::DONT_CARE }, "NotifyStencilUpload_NoBlit");

GPU/GLES/TextureCacheGLES.cpp

@@ -383,7 +383,7 @@ void TextureCacheGLES::ApplyTextureFramebuffer(VirtualFramebuffer *framebuffer,
 
 		const GEPaletteFormat clutFormat = gstate.getClutPaletteFormat();
 		GLRTexture *clutTexture = depalShaderCache_->GetClutTexture(clutFormat, clutHash_, clutBuf_);
-		Draw::Framebuffer *depalFBO = framebufferManagerGL_->GetTempFBO(TempFBO::DEPAL, framebuffer->renderWidth, framebuffer->renderHeight, Draw::FBO_8888);
+		Draw::Framebuffer *depalFBO = framebufferManagerGL_->GetTempFBO(TempFBO::DEPAL, framebuffer->renderWidth, framebuffer->renderHeight);
 		draw_->BindFramebufferAsRenderTarget(depalFBO, { Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE }, "Depal");
 
 		render_->SetScissor(GLRect2D{ 0, 0, (int)framebuffer->renderWidth, (int)framebuffer->renderHeight });

GPU/GPUState.h

@@ -493,7 +493,7 @@ enum {
 	GPU_SUPPORTS_FRAMEBUFFER_BLIT_TO_DEPTH = FLAG_BIT(27),
 	GPU_SUPPORTS_OES_TEXTURE_NPOT = FLAG_BIT(28),
 	GPU_NEEDS_Z_EQUAL_W_HACK = FLAG_BIT(29),
-	GPU_PREFER_CPU_DOWNLOAD = FLAG_BIT(30),
+	// Free bit: 30
 	GPU_PREFER_REVERSE_COLOR_ORDER = FLAG_BIT(31),
 };
 

GPU/Vulkan/FramebufferManagerVulkan.cpp

@@ -284,7 +284,7 @@ VkImageView FramebufferManagerVulkan::BindFramebufferAsColorTexture(int stage, V
 	// Currently rendering to this framebuffer. Need to make a copy.
 	if (!skipCopy && framebuffer == currentRenderVfb_) {
 		// TODO: Maybe merge with bvfbs_?  Not sure if those could be packing, and they're created at a different size.
-		Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight, (Draw::FBColorDepth)framebuffer->colorDepth);
+		Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight);
 		if (renderCopy) {
 			VirtualFramebuffer copyInfo = *framebuffer;
 			copyInfo.fbo = renderCopy;
@@ -305,10 +305,6 @@ VkImageView FramebufferManagerVulkan::BindFramebufferAsColorTexture(int stage, V
 	}
 }
 
-void FramebufferManagerVulkan::UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) {
-	// Nothing to do here.
-}
-
 void FramebufferManagerVulkan::BlitFramebuffer(VirtualFramebuffer *dst, int dstX, int dstY, VirtualFramebuffer *src, int srcX, int srcY, int w, int h, int bpp) {
 	if (!dst->fbo || !src->fbo || !useBufferedRendering_) {
 		// This can happen if they recently switched from non-buffered.
@@ -370,12 +366,11 @@ void FramebufferManagerVulkan::BlitFramebuffer(VirtualFramebuffer *dst, int dstX
 	// BlitFramebuffer can clip, but CopyFramebufferImage is more restricted.
 	// In case the src goes outside, we just skip the optimization in that case.
 	const bool sameSize = dstX2 - dstX1 == srcX2 - srcX1 && dstY2 - dstY1 == srcY2 - srcY1;
-	const bool sameDepth = dst->colorDepth == src->colorDepth;
 	const bool srcInsideBounds = srcX2 <= src->renderWidth && srcY2 <= src->renderHeight;
 	const bool dstInsideBounds = dstX2 <= dst->renderWidth && dstY2 <= dst->renderHeight;
 	const bool xOverlap = src == dst && srcX2 > dstX1 && srcX1 < dstX2;
 	const bool yOverlap = src == dst && srcY2 > dstY1 && srcY1 < dstY2;
-	if (sameSize && sameDepth && srcInsideBounds && dstInsideBounds && !(xOverlap && yOverlap)) {
+	if (sameSize && srcInsideBounds && dstInsideBounds && !(xOverlap && yOverlap)) {
 		draw_->CopyFramebufferImage(src->fbo, 0, srcX1, srcY1, 0, dst->fbo, 0, dstX1, dstY1, 0, dstX2 - dstX1, dstY2 - dstY1, 1, Draw::FB_COLOR_BIT, "BlitFramebuffer_Copy");
 	} else {
 		draw_->BlitFramebuffer(src->fbo, srcX1, srcY1, srcX2, srcY2, dst->fbo, dstX1, dstY1, dstX2, dstY2, Draw::FB_COLOR_BIT, Draw::FB_BLIT_NEAREST, "BlitFramebuffer_Blit");

GPU/Vulkan/FramebufferManagerVulkan.h

@@ -67,7 +67,6 @@ protected:
 
 	// Used by ReadFramebufferToMemory and later framebuffer block copies
 	void BlitFramebuffer(VirtualFramebuffer *dst, int dstX, int dstY, VirtualFramebuffer *src, int srcX, int srcY, int w, int h, int bpp) override;
-	void UpdateDownloadTempBuffer(VirtualFramebuffer *nvfb) override;
 
 private:
 	void InitDeviceObjects();

GPU/Vulkan/GPU_Vulkan.cpp

@@ -231,7 +231,6 @@ void GPU_Vulkan::CheckGPUFeatures() {
 	features |= GPU_SUPPORTS_INSTANCE_RENDERING;
 	features |= GPU_SUPPORTS_VERTEX_TEXTURE_FETCH;
 	features |= GPU_SUPPORTS_TEXTURE_FLOAT;
-	features |= GPU_PREFER_CPU_DOWNLOAD;
 
 	if (vulkan_->GetDeviceInfo().canBlitToPreferredDepthStencilFormat) {
 		features |= GPU_SUPPORTS_FRAMEBUFFER_BLIT_TO_DEPTH;

GPU/Vulkan/TextureCacheVulkan.cpp

@@ -571,7 +571,7 @@ void TextureCacheVulkan::ApplyTextureFramebuffer(VirtualFramebuffer *framebuffer
 		const GEPaletteFormat clutFormat = gstate.getClutPaletteFormat();
 		VulkanTexture *clutTexture = depalShaderCache_->GetClutTexture(clutFormat, clutHash_, clutBuf_, expand32);
 
-		Draw::Framebuffer *depalFBO = framebufferManager_->GetTempFBO(TempFBO::DEPAL, framebuffer->renderWidth, framebuffer->renderHeight, Draw::FBO_8888);
+		Draw::Framebuffer *depalFBO = framebufferManager_->GetTempFBO(TempFBO::DEPAL, framebuffer->renderWidth, framebuffer->renderHeight);
 		draw_->BindFramebufferAsRenderTarget(depalFBO, { Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE, Draw::RPAction::DONT_CARE }, "Depal");
 
 		Vulkan2D::Vertex verts[4] = {