From 81cf76a61968c1618cc48d24633f0b641fb59b89 Mon Sep 17 00:00:00 2001
From: "Unknown W. Brackets" <checkins@unknownbrackets.org>
Date: Sat, 8 Oct 2022 15:33:16 -0700
Subject: [PATCH 1/2] D3D9: Correct sampler states for binding 1+.

---
 Common/GPU/D3D9/thin3d_d3d9.cpp | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/Common/GPU/D3D9/thin3d_d3d9.cpp b/Common/GPU/D3D9/thin3d_d3d9.cpp
index 377a0b4050..04bec49896 100644
--- a/Common/GPU/D3D9/thin3d_d3d9.cpp
+++ b/Common/GPU/D3D9/thin3d_d3d9.cpp
@@ -201,11 +201,19 @@ public:
 	D3DTEXTUREFILTERTYPE magFilt, minFilt, mipFilt;
 
 	void Apply(LPDIRECT3DDEVICE9 device, int index) {
-		dxstate.texAddressU.set(wrapS);
-		dxstate.texAddressV.set(wrapT);
-		dxstate.texMagFilter.set(magFilt);
-		dxstate.texMinFilter.set(minFilt);
-		dxstate.texMipFilter.set(mipFilt);
+		if (index == 0) {
+			dxstate.texAddressU.set(wrapS);
+			dxstate.texAddressV.set(wrapT);
+			dxstate.texMagFilter.set(magFilt);
+			dxstate.texMinFilter.set(minFilt);
+			dxstate.texMipFilter.set(mipFilt);
+		} else {
+			pD3Ddevice9->SetSamplerState(index, D3DSAMP_ADDRESSU, wrapS);
+			pD3Ddevice9->SetSamplerState(index, D3DSAMP_ADDRESSV, wrapT);
+			pD3Ddevice9->SetSamplerState(index, D3DSAMP_MAGFILTER, magFilt);
+			pD3Ddevice9->SetSamplerState(index, D3DSAMP_MINFILTER, minFilt);
+			pD3Ddevice9->SetSamplerState(index, D3DSAMP_MIPFILTER, mipFilt);
+		}
 	}
 };
 

From 157ffed57f30e59c4a66356430ba51622e74f976 Mon Sep 17 00:00:00 2001
From: "Unknown W. Brackets" <checkins@unknownbrackets.org>
Date: Sat, 8 Oct 2022 15:36:36 -0700
Subject: [PATCH 2/2] D3D9: Add simple rendered CLUT handling.

I think there's still a deeper half-pixel offset issue, but this fixes
Brave Story.
---
 GPU/Common/DepalettizeShaderCommon.cpp | 9 ++++++++-
 GPU/Directx9/TextureCacheDX9.cpp       | 4 ++++
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/GPU/Common/DepalettizeShaderCommon.cpp b/GPU/Common/DepalettizeShaderCommon.cpp
index a1e593e626..628efc787f 100644
--- a/GPU/Common/DepalettizeShaderCommon.cpp
+++ b/GPU/Common/DepalettizeShaderCommon.cpp
@@ -174,7 +174,10 @@ void GenerateDepalShaderFloat(ShaderWriter &writer, const DepalConfig &config) {
 	case GE_FORMAT_CLUT8:
 		if (shift == 0 && mask == 0xFF) {
 			// Easy peasy.
-			sprintf(lookupMethod, "index.r");
+			if (writer.Lang().shaderLanguage == HLSL_D3D9)
+				sprintf(lookupMethod, "index.a");
+			else
+				sprintf(lookupMethod, "index.r");
 			formatOK = true;
 		} else {
 			// Deal with this if we find it.
@@ -302,6 +305,10 @@ void GenerateDepalShaderFloat(ShaderWriter &writer, const DepalConfig &config) {
 	// Offset by half a texel (plus clutBase) to turn NEAREST filtering into FLOOR.
 	// Technically, the clutBase should be |'d, not added, but that's hard with floats.
 	float texel_offset = ((float)config.startPos + 0.5f) / texturePixels;
+	if (writer.Lang().shaderLanguage == HLSL_D3D9) {
+		// Seems to need a half-pixel offset fix?  Might mean it was rendered wrong...
+		texel_offset += 0.5f / texturePixels;
+	}
 	char offset[128] = "";
 	sprintf(offset, " + %f", texel_offset);
 
diff --git a/GPU/Directx9/TextureCacheDX9.cpp b/GPU/Directx9/TextureCacheDX9.cpp
index 9edc451a52..0870299ac5 100644
--- a/GPU/Directx9/TextureCacheDX9.cpp
+++ b/GPU/Directx9/TextureCacheDX9.cpp
@@ -47,6 +47,8 @@ Draw::DataFormat FromD3D9Format(u32 fmt) {
 		return Draw::DataFormat::A1R5G5B5_UNORM_PACK16;
 	case D3DFMT_R5G6B5:
 		return Draw::DataFormat::R5G6B5_UNORM_PACK16;
+	case D3DFMT_A8:
+		return Draw::DataFormat::R8_UNORM;
 	case D3DFMT_A8R8G8B8:
 	default:
 		return Draw::DataFormat::R8G8B8A8_UNORM;
@@ -251,6 +253,8 @@ void TextureCacheDX9::BuildTexture(TexCacheEntry *const entry) {
 		dstFmt = ToD3D9Format(plan.replaced->Format(plan.baseLevelSrc));
 	} else if (plan.scaleFactor > 1 || plan.saveTexture) {
 		dstFmt = D3DFMT_A8R8G8B8;
+	} else if (plan.decodeToClut8) {
+		dstFmt = D3DFMT_A8;
 	}
 
 	int levels;