From 855d16ffb3ebc54be005cd06100f40440fad7c4c Mon Sep 17 00:00:00 2001
From: "Unknown W. Brackets" <checkins@unknownbrackets.org>
Date: Mon, 7 Nov 2022 20:05:51 -0800
Subject: [PATCH] GPU: Prefer scaling depth to 16-bit if using 24.

In most cases, this will be a lot faster than pixel depth rounding, and
may avoid more issues in vertex rounding.
---
 GPU/D3D11/GPU_D3D11.cpp   |  3 +--
 GPU/GLES/GPU_GLES.cpp     | 11 ++++++-----
 GPU/Vulkan/GPU_Vulkan.cpp | 25 ++++++++++++++++---------
 3 files changed, 23 insertions(+), 16 deletions(-)

diff --git a/GPU/D3D11/GPU_D3D11.cpp b/GPU/D3D11/GPU_D3D11.cpp
index d1c5edcce3..90fb480155 100644
--- a/GPU/D3D11/GPU_D3D11.cpp
+++ b/GPU/D3D11/GPU_D3D11.cpp
@@ -125,8 +125,7 @@ u32 GPU_D3D11::CheckGPUFeatures() const {
 	if (!g_Config.bHighQualityDepth && (features & GPU_USE_ACCURATE_DEPTH) != 0) {
 		features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
 	} else if (PSP_CoreParameter().compat.flags().PixelDepthRounding) {
-		// Use fragment rounding on desktop and GLES3, most accurate.
-		features |= GPU_ROUND_FRAGMENT_DEPTH_TO_16BIT;
+		features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
 	} else if (PSP_CoreParameter().compat.flags().VertexDepthRounding) {
 		features |= GPU_ROUND_DEPTH_TO_16BIT;
 	}
diff --git a/GPU/GLES/GPU_GLES.cpp b/GPU/GLES/GPU_GLES.cpp
index 791fa22869..8b71b2dce1 100644
--- a/GPU/GLES/GPU_GLES.cpp
+++ b/GPU/GLES/GPU_GLES.cpp
@@ -181,17 +181,18 @@ u32 GPU_GLES::CheckGPUFeatures() const {
 
 	// If we already have a 16-bit depth buffer, we don't need to round.
 	bool prefer24 = draw_->GetDeviceCaps().preferredDepthBufferFormat == Draw::DataFormat::D24_S8;
-	if (prefer24) {
+	bool prefer16 = draw_->GetDeviceCaps().preferredDepthBufferFormat == Draw::DataFormat::D16;
+	if (!prefer16) {
 		if (!g_Config.bHighQualityDepth && (features & GPU_USE_ACCURATE_DEPTH) != 0) {
 			features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
 		} else if (PSP_CoreParameter().compat.flags().PixelDepthRounding) {
-			if (!gl_extensions.IsGLES || gl_extensions.GLES3) {
-				// Use fragment rounding on desktop and GLES3, most accurate.
-				features |= GPU_ROUND_FRAGMENT_DEPTH_TO_16BIT;
-			} else if (prefer24 && (features & GPU_USE_ACCURATE_DEPTH) != 0) {
+			if (prefer24 && (features & GPU_USE_ACCURATE_DEPTH) != 0) {
 				// Here we can simulate a 16 bit depth buffer by scaling.
 				// Note that the depth buffer is fixed point, not floating, so dividing by 256 is pretty good.
 				features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
+			} else if (!gl_extensions.IsGLES || gl_extensions.GLES3) {
+				// Use fragment rounding on desktop and GLES3, most accurate.
+				features |= GPU_ROUND_FRAGMENT_DEPTH_TO_16BIT;
 			} else {
 				// At least do vertex rounding if nothing else.
 				features |= GPU_ROUND_DEPTH_TO_16BIT;
diff --git a/GPU/Vulkan/GPU_Vulkan.cpp b/GPU/Vulkan/GPU_Vulkan.cpp
index 754ef6ffce..17387c3e8b 100644
--- a/GPU/Vulkan/GPU_Vulkan.cpp
+++ b/GPU/Vulkan/GPU_Vulkan.cpp
@@ -257,15 +257,22 @@ u32 GPU_Vulkan::CheckGPUFeatures() const {
 		INFO_LOG(G3D, "Deficient texture format support: 4444: %d  1555: %d  565: %d", fmt4444, fmt1555, fmt565);
 	}
 
-	if (!g_Config.bHighQualityDepth && (features & GPU_USE_ACCURATE_DEPTH) != 0) {
-		features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
-	}
-	else if (PSP_CoreParameter().compat.flags().PixelDepthRounding) {
-		// Use fragment rounding on desktop and GLES3, most accurate.
-		features |= GPU_ROUND_FRAGMENT_DEPTH_TO_16BIT;
-	}
-	else if (PSP_CoreParameter().compat.flags().VertexDepthRounding) {
-		features |= GPU_ROUND_DEPTH_TO_16BIT;
+	bool prefer24 = draw_->GetDeviceCaps().preferredDepthBufferFormat == Draw::DataFormat::D24_S8;
+	bool prefer16 = draw_->GetDeviceCaps().preferredDepthBufferFormat == Draw::DataFormat::D16;
+	if (!prefer16) {
+		if (!g_Config.bHighQualityDepth && (features & GPU_USE_ACCURATE_DEPTH) != 0) {
+			features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
+		} else if (PSP_CoreParameter().compat.flags().PixelDepthRounding) {
+			if (prefer24 && (features & GPU_USE_ACCURATE_DEPTH) != 0) {
+				// Here we can simulate a 16 bit depth buffer by scaling.
+				// Note that the depth buffer is fixed point, not floating, so dividing by 256 is pretty good.
+				features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
+			} else {
+				features |= GPU_ROUND_FRAGMENT_DEPTH_TO_16BIT;
+			}
+		} else if (PSP_CoreParameter().compat.flags().VertexDepthRounding) {
+			features |= GPU_ROUND_DEPTH_TO_16BIT;
+		}
 	}
 
 	if (g_Config.bStereoRendering && draw_->GetDeviceCaps().multiViewSupported) {