GS/HW: Clean up target download formats

bin/resources/GameIndex.yaml

@@ -12927,7 +12927,7 @@ SLAJ-25041:
   name: "Harry Potter and the Prisoner of Azkaban"
   region: "NTSC-C"
   gsHWFixes:
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Aligns mirror reflections, removes some bloom-related ghosting.
     recommendedBlendingLevel: 4 # Improves lighting effects.
 SLAJ-25042:
@@ -16829,7 +16829,7 @@ SLES-51192:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51193:
@@ -16839,7 +16839,7 @@ SLES-51193:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51194:
@@ -16849,7 +16849,7 @@ SLES-51194:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51195:
@@ -16859,7 +16859,7 @@ SLES-51195:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51196:
@@ -16869,7 +16869,7 @@ SLES-51196:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51197:
@@ -16926,7 +16926,7 @@ SLES-51214:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51215:
@@ -16936,7 +16936,7 @@ SLES-51215:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51216:
@@ -16946,7 +16946,7 @@ SLES-51216:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51217:
@@ -16956,7 +16956,7 @@ SLES-51217:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51218:
@@ -16966,7 +16966,7 @@ SLES-51218:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51219:
@@ -16976,7 +16976,7 @@ SLES-51219:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
     texturePreloading: 0 # Performs better on certain areas like mid-game in and around the castle as the hash goes easily to 200 MB.
 SLES-51220:
@@ -19896,7 +19896,7 @@ SLES-52440:
   region: "PAL-M7"
   compat: 3
   gsHWFixes:
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Aligns mirror reflections, removes some bloom-related ghosting.
     recommendedBlendingLevel: 4 # Improves lighting effects.
 SLES-52444:
@@ -20123,7 +20123,7 @@ SLES-52527:
   name: "Harry Potter og Fangen fra Azkaban"
   region: "PAL-M4"
   gsHWFixes:
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Aligns mirror reflections, removes some bloom-related ghosting.
     recommendedBlendingLevel: 4 # Improves lighting effects.
 SLES-52531:
@@ -20434,7 +20434,7 @@ SLES-52600:
   name: "Harry Potter and the Prisoner of Azkaban"
   region: "PAL-PL"
   gsHWFixes:
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Aligns mirror reflections, removes some bloom-related ghosting.
     recommendedBlendingLevel: 4 # Improves lighting effects.
 SLES-52601:
@@ -31108,7 +31108,7 @@ SLKA-25172:
   name: "Harry Potter and the Prisoner of Azkaban"
   region: "NTSC-K"
   gsHWFixes:
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Aligns mirror reflections, removes some bloom-related ghosting.
     recommendedBlendingLevel: 4 # Improves lighting effects.
 SLKA-25173:
@@ -36957,7 +36957,7 @@ SLPM-62241:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
 SLPM-62242:
   name: "撞球 ビリヤードマスター2"
@@ -38369,7 +38369,7 @@ SLPM-62513:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
 SLPM-62514:
   name: "シムピープル ～お茶の間劇場～ [EA BEST HITS]"
@@ -39800,7 +39800,7 @@ SLPM-64528:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
 SLPM-64532:
   name: "Digital Holmes"
@@ -43355,7 +43355,7 @@ SLPM-65612:
   name-en: "Harry Potter and the Prisoner of Azkaban"
   region: "NTSC-J"
   gsHWFixes:
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Aligns mirror reflections, removes some bloom-related ghosting.
     recommendedBlendingLevel: 4 # Improves lighting effects.
 SLPM-65613:
@@ -52157,7 +52157,7 @@ SLPM-68005:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
 SLPM-68007:
   name: "カラオケレボリューション用マイク同梱お試しディスク"
@@ -54308,7 +54308,7 @@ SLPS-20234:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
 SLPS-20237:
   name: "Saikyou Toudai Shougi 3"
@@ -65273,7 +65273,7 @@ SLUS-20576:
     - EETimingHack # Fixes flickering textures.
   gsHWFixes:
     recommendedBlendingLevel: 4 # Fixes missing lighting effects.
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Fixes misaligned lighting and other effects.
 SLUS-20577:
   name: "Drome Racers"
@@ -67218,7 +67218,7 @@ SLUS-20926:
   name: "Harry Potter and the Prisoner of Azkaban"
   region: "NTSC-U"
   gsHWFixes:
-    cpuFramebufferConversion: 1 # Fixes right side of the screen from garbage textures.
+    textureInsideRT: 1 # Fixes right side shadow lookup.
     halfPixelOffset: 2 # Aligns mirror reflections, removes some bloom-related ghosting.
     recommendedBlendingLevel: 4 # Improves lighting effects.
 SLUS-20927:

bin/resources/shaders/dx11/convert.fx

@@ -313,6 +313,127 @@ float ps_convert_rgb5a1_float16_biln(PS_INPUT input) : SV_Depth
 	SAMPLE_RGBA_DEPTH_BILN(rgb5a1_to_depth16);
 }
 
+PS_OUTPUT ps_convert_rgb5a1_8i(PS_INPUT input)
+{
+	PS_OUTPUT output;
+
+	// Convert a RGB5A1 texture into a 8 bits packed texture
+	// Input column: 16x2 RGB5A1 pixels
+	// 0: 16 RGBA
+	// 1: 16 RGBA
+	// Output column: 16x4 Index pixels
+	// 0: 16 R5G2
+	// 1: 16 R5G2
+	// 2: 16 G2B5A1
+	// 3: 16 G2B5A1
+	uint2 pos = uint2(input.p.xy);
+
+	// Collapse separate R G B A areas into their base pixel
+	uint2 column = (pos & ~uint2(0u, 3u)) / uint2(1,2);
+	uint2 subcolumn = (pos & uint2(0u, 1u));
+	column.x -= (column.x / 128) * 64;
+	column.y += (column.y / 32) * 32;
+	
+	uint PSM = uint(DOFFSET);
+	
+	// Deal with swizzling differences
+	if ((PSM & 0x8) != 0) // PSMCT16S
+	{
+		if ((pos.x & 32) != 0)
+		{
+			column.y += 32; // 4 columns high times 4 to get bottom 4 blocks
+			column.x &= ~32;
+		}
+		
+		if ((pos.x & 64) != 0)
+		{
+			column.x -= 32;
+		}
+		
+		if (((pos.x & 16) != 0) != ((pos.y & 16) != 0))
+		{
+			column.x ^= 16; 
+			column.y ^= 8;
+		}
+		
+		if ((PSM & 0x30) != 0) // PSMZ16S - Untested but hopefully ok if anything uses it.
+		{
+			column.x ^= 32;
+			column.y ^= 16;
+		}
+	}
+	else // PSMCT16
+	{
+		if ((pos.y & 32) != 0)
+		{
+			column.y -= 16;
+			column.x += 32;
+		}
+		
+		if ((pos.x & 96) != 0)
+		{
+			uint multi = (pos.x & 96) / 32;
+			column.y += 16 * multi; // 4 columns high times 4 to get bottom 4 blocks
+			column.x -= (pos.x & 96);
+		}
+		
+		if (((pos.x & 16) != 0) != ((pos.y & 16) != 0))
+		{
+			column.x ^= 16; 
+			column.y ^= 8;
+		}
+		
+		if ((PSM & 0x30) != 0) // PSMZ16 - Untested but hopefully ok if anything uses it.
+		{
+			column.x ^= 32;
+			column.y ^= 32;
+		}
+	}
+	
+	uint2 coord = column | subcolumn;
+
+	// Compensate for potentially differing page pitch.
+	uint SBW = uint(EMODA);
+	uint DBW = uint(EMODC);
+	uint2 block_xy = coord / uint2(64,64);
+	uint block_num = (block_xy.y * (DBW / 128)) + block_xy.x;
+	uint2 block_offset = uint2((block_num % (SBW / 64)) * 64, (block_num / (SBW / 64)) * 64);
+	coord = (coord % uint2(64, 64)) + block_offset;
+
+	// Apply offset to cols 1 and 2
+	uint is_col23 = pos.y & 4u;
+	uint is_col13 = pos.y & 2u;
+	uint is_col12 = is_col23 ^ (is_col13 << 1);
+	coord.x ^= is_col12; // If cols 1 or 2, flip bit 3 of x
+
+	float ScaleFactor = BGColor.x;
+	if (floor(ScaleFactor) != ScaleFactor)
+		coord = uint2(float2(coord) * ScaleFactor);
+	else
+		coord *= uint(ScaleFactor);
+
+	float4 pixel = Texture.Load(int3(int2(coord), 0));
+	uint4 denorm_c = (uint4)(pixel * 255.5f);
+	if ((pos.y & 2u) == 0u)
+	{
+		uint red = (denorm_c.r >> 3) & 0x1Fu;
+		uint green = (denorm_c.g >> 3) & 0x1Fu;
+		float sel0 = (float)(((green << 5) | red) & 0xFF) / 255.0f;
+		
+		output.c = (float4)(sel0);
+	}
+	else
+	{
+		uint green = (denorm_c.g >> 3) & 0x1Fu;
+		uint blue = (denorm_c.b >> 3) & 0x1Fu;
+		uint alpha = denorm_c.a & 0x80u;
+		float sel0 = (float)((alpha | (blue << 2) | (green >> 3)) & 0xFF) / 255.0f;
+
+		output.c = (float4)(sel0);
+	}
+	return output;
+}
+
 PS_OUTPUT ps_convert_rgba_8i(PS_INPUT input)
 {
 	PS_OUTPUT output;

bin/resources/shaders/opengl/convert.glsl

@@ -237,9 +237,131 @@ void ps_convert_rgb5a1_float16_biln()
 }
 #endif
 
+#ifdef ps_convert_rgb5a1_8i
+uniform uint SBW;
+uniform uint DBW;
+uniform uint PSM;
+uniform float ScaleFactor;
+
+void ps_convert_rgb5a1_8i()
+{
+	// Convert a RGB5A1 texture into a 8 bits packed texture
+	// Input column: 16x2 RGB5A1 pixels
+	// 0: 16 RGBA
+	// 1: 16 RGBA
+	// Output column: 16x4 Index pixels
+	// 0: 16 R5G2
+	// 1: 16 R5G2
+	// 2: 16 G2B5A1
+	// 3: 16 G2B5A1
+	
+	uvec2 pos = uvec2(gl_FragCoord.xy);
+
+	// Collapse separate R G B A areas into their base pixel
+	uvec2 column = (pos & ~uvec2(0u, 3u)) / uvec2(1,2);
+	uvec2 subcolumn = (pos & uvec2(0u, 1u));
+	column.x -= (column.x / 128) * 64;
+	column.y += (column.y / 32) * 32;
+	
+	// Deal with swizzling differences
+	if ((PSM & 0x8) != 0) // PSMCT16S
+	{
+		if ((pos.x & 32) != 0)
+		{
+			column.y += 32; // 4 columns high times 4 to get bottom 4 blocks
+			column.x &= ~32;
+		}
+		
+		if ((pos.x & 64) != 0)
+		{
+			column.x -= 32;
+		}
+		
+		if (((pos.x & 16) != 0) != ((pos.y & 16) != 0))
+		{
+			column.x ^= 16; 
+			column.y ^= 8;
+		}
+		
+		if ((PSM & 0x30) != 0) // PSMZ16S - Untested but hopefully ok if anything uses it.
+		{
+			column.x ^= 32;
+			column.y ^= 16;
+		}
+	}
+	else // PSMCT16
+	{
+		if ((pos.y & 32) != 0)
+		{
+			column.y -= 16;
+			column.x += 32;
+		}
+		
+		if ((pos.x & 96) != 0)
+		{
+			uint multi = (pos.x & 96) / 32;
+			column.y += 16 * multi; // 4 columns high times 4 to get bottom 4 blocks
+			column.x -= (pos.x & 96);
+		}
+		
+		if (((pos.x & 16) != 0) != ((pos.y & 16) != 0))
+		{
+			column.x ^= 16; 
+			column.y ^= 8;
+		}
+		
+		if ((PSM & 0x30) != 0) // PSMZ16 - Untested but hopefully ok if anything uses it.
+		{
+			column.x ^= 32;
+			column.y ^= 32;
+		}
+	}
+	uvec2 coord = column | subcolumn;
+	
+	// Compensate for potentially differing page pitch.
+	uvec2 block_xy = coord / uvec2(64u, 64u);
+	uint block_num = (block_xy.y * (DBW / 128u)) + block_xy.x;
+	uvec2 block_offset = uvec2((block_num % (SBW / 64u)) * 64u, (block_num / (SBW / 64u)) * 64u);
+	coord = (coord % uvec2(64u, 64u)) + block_offset;
+
+	// Apply offset to cols 1 and 2
+	uint is_col23 = pos.y & 4u;
+	uint is_col13 = pos.y & 2u;
+	uint is_col12 = is_col23 ^ (is_col13 << 1);
+	coord.x ^= is_col12; // If cols 1 or 2, flip bit 3 of x
+
+	if (floor(ScaleFactor) != ScaleFactor)
+		coord = uvec2(vec2(coord) * ScaleFactor);
+	else
+		coord *= uvec2(ScaleFactor);
+
+	vec4 pixel = texelFetch(TextureSampler, ivec2(coord), 0);
+	
+	uvec4 denorm_c = uvec4(pixel * 255.5f);
+	if ((pos.y & 2u) == 0u)
+	{
+		uint red = (denorm_c.r >> 3) & 0x1Fu;
+		uint green = (denorm_c.g >> 3) & 0x1Fu;
+		float sel0 = float(((green << 5) | red) & 0xFF) / 255.0f;
+		
+		SV_Target0 = vec4(sel0);
+	}
+	else
+	{
+		uint green = (denorm_c.g >> 3) & 0x1Fu;
+		uint blue = (denorm_c.b >> 3) & 0x1Fu;
+		uint alpha = denorm_c.a & 0x80u;
+		float sel0 = float((alpha | (blue << 2) | (green >> 3)) & 0xFF) / 255.0f;
+
+		SV_Target0 = vec4(sel0);
+	}
+}
+#endif
+
 #ifdef ps_convert_rgba_8i
 uniform uint SBW;
 uniform uint DBW;
+uniform uint PSM;
 uniform float ScaleFactor;
 
 void ps_convert_rgba_8i()

bin/resources/shaders/vulkan/convert.glsl

@@ -307,12 +307,138 @@ void ps_convert_rgb5a1_float16_biln()
 }
 #endif
 
+#ifdef ps_convert_rgb5a1_8i
+layout(push_constant) uniform cb10
+{
+	uint SBW;
+	uint DBW;
+	uint PSM;
+	float cb_pad1;
+	float ScaleFactor;
+	vec3 cb_pad2;
+};
+
+void ps_convert_rgb5a1_8i()
+{
+	// Convert a RGB5A1 texture into a 8 bits packed texture
+	// Input column: 16x2 RGB5A1 pixels
+	// 0: 16 RGBA
+	// 1: 16 RGBA
+	// Output column: 16x4 Index pixels
+	// 0: 16 R5G2
+	// 1: 16 R5G2
+	// 2: 16 G2B5A1
+	// 3: 16 G2B5A1
+	
+	uvec2 pos = uvec2(gl_FragCoord.xy);
+
+	// Collapse separate R G B A areas into their base pixel
+	uvec2 column = (pos & ~uvec2(0u, 3u)) / uvec2(1,2);
+	uvec2 subcolumn = (pos & uvec2(0u, 1u));
+	column.x -= (column.x / 128) * 64;
+	column.y += (column.y / 32) * 32;
+	
+	// Deal with swizzling differences
+	if ((PSM & 0x8) != 0) // PSMCT16S
+	{
+		if ((pos.x & 32) != 0)
+		{
+			column.y += 32; // 4 columns high times 4 to get bottom 4 blocks
+			column.x &= ~32;
+		}
+		
+		if ((pos.x & 64) != 0)
+		{
+			column.x -= 32;
+		}
+		
+		if (((pos.x & 16) != 0) != ((pos.y & 16) != 0))
+		{
+			column.x ^= 16; 
+			column.y ^= 8;
+		}
+		
+		if ((PSM & 0x30) != 0) // PSMZ16S - Untested but hopefully ok if anything uses it.
+		{
+			column.x ^= 32;
+			column.y ^= 16;
+		}
+	}
+	else // PSMCT16
+	{
+		if ((pos.y & 32) != 0)
+		{
+			column.y -= 16;
+			column.x += 32;
+		}
+		
+		if ((pos.x & 96) != 0)
+		{
+			uint multi = (pos.x & 96) / 32;
+			column.y += 16 * multi; // 4 columns high times 4 to get bottom 4 blocks
+			column.x -= (pos.x & 96);
+		}
+		
+		if (((pos.x & 16) != 0) != ((pos.y & 16) != 0))
+		{
+			column.x ^= 16; 
+			column.y ^= 8;
+		}
+		
+		if ((PSM & 0x30) != 0) // PSMZ16 - Untested but hopefully ok if anything uses it.
+		{
+			column.x ^= 32;
+			column.y ^= 32;
+		}
+	}
+	uvec2 coord = column | subcolumn;
+	
+	// Compensate for potentially differing page pitch.
+	uvec2 block_xy = coord / uvec2(64u, 64u);
+	uint block_num = (block_xy.y * (DBW / 128u)) + block_xy.x;
+	uvec2 block_offset = uvec2((block_num % (SBW / 64u)) * 64u, (block_num / (SBW / 64u)) * 64u);
+	coord = (coord % uvec2(64u, 64u)) + block_offset;
+
+	// Apply offset to cols 1 and 2
+	uint is_col23 = pos.y & 4u;
+	uint is_col13 = pos.y & 2u;
+	uint is_col12 = is_col23 ^ (is_col13 << 1);
+	coord.x ^= is_col12; // If cols 1 or 2, flip bit 3 of x
+
+	if (floor(ScaleFactor) != ScaleFactor)
+		coord = uvec2(vec2(coord) * ScaleFactor);
+	else
+		coord *= uvec2(ScaleFactor);
+
+	vec4 pixel = texelFetch(samp0, ivec2(coord), 0);
+	uvec4 denorm_c = uvec4(pixel * 255.5f);
+	if ((pos.y & 2u) == 0u)
+	{
+		uint red = (denorm_c.r >> 3) & 0x1Fu;
+		uint green = (denorm_c.g >> 3) & 0x1Fu;
+		float sel0 = float(((green << 5) | red) & 0xFF) / 255.0f;
+		
+		o_col0 = vec4(sel0);
+	}
+	else
+	{
+		uint green = (denorm_c.g >> 3) & 0x1Fu;
+		uint blue = (denorm_c.b >> 3) & 0x1Fu;
+		uint alpha = denorm_c.a & 0x80u;
+		float sel0 = float((alpha | (blue << 2) | (green >> 3)) & 0xFF) / 255.0f;
+
+		o_col0 = vec4(sel0);
+	}
+}
+#endif
+
 #ifdef ps_convert_rgba_8i
 layout(push_constant) uniform cb10
 {
 	uint SBW;
 	uint DBW;
-	uvec2 cb_pad1;
+	uint PSM;
+	float cb_pad1;
 	float ScaleFactor;
 	vec3 cb_pad2;
 };

pcsx2/GS/Renderers/Common/GSDevice.cpp

@@ -68,6 +68,7 @@ const char* shaderName(ShaderConvert value)
 		case ShaderConvert::DEPTH_COPY:             return "ps_depth_copy";
 		case ShaderConvert::DOWNSAMPLE_COPY:        return "ps_downsample_copy";
 		case ShaderConvert::RGBA_TO_8I:             return "ps_convert_rgba_8i";
+		case ShaderConvert::RGB5A1_TO_8I:           return "ps_convert_rgb5a1_8i";
 		case ShaderConvert::CLUT_4:                 return "ps_convert_clut_4";
 		case ShaderConvert::CLUT_8:                 return "ps_convert_clut_8";
 		case ShaderConvert::YUV:                    return "ps_yuv";

pcsx2/GS/Renderers/Common/GSDevice.h

@@ -44,6 +44,7 @@ enum class ShaderConvert
 	DEPTH_COPY,
 	DOWNSAMPLE_COPY,
 	RGBA_TO_8I,
+	RGB5A1_TO_8I,
 	CLUT_4,
 	CLUT_8,
 	YUV,

pcsx2/GS/Renderers/DX11/GSDevice11.cpp

@@ -1417,14 +1417,14 @@ void GSDevice11::ConvertToIndexedTexture(GSTexture* sTex, float sScale, u32 offs
 	{
 		float scale;
 		float pad1[3];
-		u32 SBW, DBW, pad3;
+		u32 SBW, DBW, SPSM;
 	};
 
-	const Uniforms cb = {sScale, {}, SBW, DBW};
+	const Uniforms cb = {sScale, {}, SBW, DBW, SPSM};
 	m_ctx->UpdateSubresource(m_merge.cb.get(), 0, nullptr, &cb, 0, 0);
 
 	const GSVector4 dRect(0, 0, dTex->GetWidth(), dTex->GetHeight());
-	const ShaderConvert shader = ShaderConvert::RGBA_TO_8I;
+	const ShaderConvert shader = ((SPSM & 0xE) == 0) ? ShaderConvert::RGBA_TO_8I : ShaderConvert::RGB5A1_TO_8I;
 	StretchRect(sTex, GSVector4::zero(), dTex, dRect, m_convert.ps[static_cast<int>(shader)].get(), m_merge.cb.get(), nullptr, false);
 }
 

pcsx2/GS/Renderers/DX12/GSDevice12.cpp

@@ -1486,15 +1486,15 @@ void GSDevice12::ConvertToIndexedTexture(
 	{
 		float scale;
 		float pad1[3];
-		u32 SBW, DBW, pad2;
+		u32 SBW, DBW, SPSM;
 	};
 
-	const Uniforms cb = {sScale, {}, SBW, DBW};
+	const Uniforms cb = {sScale, {}, SBW, DBW, SPSM};
 	SetUtilityRootSignature();
 	SetUtilityPushConstants(&cb, sizeof(cb));
 
 	const GSVector4 dRect(0, 0, dTex->GetWidth(), dTex->GetHeight());
-	const ShaderConvert shader = ShaderConvert::RGBA_TO_8I;
+	const ShaderConvert shader = ((SPSM & 0xE) == 0) ? ShaderConvert::RGBA_TO_8I : ShaderConvert::RGB5A1_TO_8I;
 	DoStretchRect(static_cast<GSTexture12*>(sTex), GSVector4::zero(), static_cast<GSTexture12*>(dTex), dRect,
 		m_convert[static_cast<int>(shader)].get(), false, true);
 }

pcsx2/GS/Renderers/HW/GSTextureCache.cpp

@@ -1617,8 +1617,7 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const bool is_color, const
 						{
 							const bool outside_target = !t->Overlaps(bp, bw, psm, r);
 
-							// We don't have a shader for this.
-							if (!possible_shuffle && TEX0.PSM == PSMT8 && ((GSLocalMemory::m_psm[t->m_TEX0.PSM].bpp != 32) || outside_target))
+							if (!possible_shuffle && TEX0.PSM == PSMT8 && outside_target)
 							{
 								continue;
 							}
@@ -1647,7 +1646,7 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const bool is_color, const
 				// Make sure the texture actually is INSIDE the RT, it's possibly not valid if it isn't.
 				// Also check BP >= TBP, create source isn't equpped to expand it backwards and all data comes from the target. (GH3)
 				else if (GSConfig.UserHacks_TextureInsideRt >= GSTextureInRtMode::InsideTargets &&
-						 (GSLocalMemory::m_psm[color_psm].bpp >= 16 || (/*possible_shuffle &&*/ GSLocalMemory::m_psm[color_psm].bpp == 8 && GSLocalMemory::m_psm[t->m_TEX0.PSM].bpp == 32)) && // Channel shuffles or non indexed lookups.
+						 (GSLocalMemory::m_psm[color_psm].bpp >= 16 || (/*possible_shuffle &&*/ GSLocalMemory::m_psm[color_psm].bpp == 8 && GSLocalMemory::m_psm[t->m_TEX0.PSM].bpp >= 16)) && // Channel shuffles or non indexed lookups.
 					t->m_age <= 1 && (!found_t || t->m_last_draw > dst->m_last_draw) /*&& CanTranslate(bp, bw, psm, block_boundary_rect, t->m_TEX0.TBP0, t->m_TEX0.PSM, t->m_TEX0.TBW)*/)
 				{
 					u32 rt_tbw = std::max(1U, t->m_TEX0.TBW);
@@ -1666,7 +1665,7 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const bool is_color, const
 					// Keep note that 2 bw is basically 1 normal page, as bw is in 64 pixels, and 8bit pages are 128 pixels wide, aka 2 bw.
 					// Also check for 4HH/HL and 8H which use the alpha channel, if the page order is wrong this can cause problems as well (Jak X font).
 					else if (!possible_shuffle && GSLocalMemory::m_psm[psm].trbpp <= 8 && (GSUtil::GetChannelMask(t->m_TEX0.PSM) != 0xF || 
-							(GSLocalMemory::m_psm[t->m_TEX0.PSM].bpp == 32 && (!(block_boundary_rect.w <= GSLocalMemory::m_psm[psm].pgs.y &&
+							((GSLocalMemory::m_psm[t->m_TEX0.PSM].bpp != 16 || GSLocalMemory::m_psm[psm].bpp < 16) && (!(block_boundary_rect.w <= GSLocalMemory::m_psm[psm].pgs.y &&
 							((GSLocalMemory::m_psm[psm].bpp == 32) ? bw : ((bw + 1) / 2)) <= t->m_TEX0.TBW) &&
 							!(((GSLocalMemory::m_psm[psm].bpp == 32) ? bw : ((bw + 1) / 2)) == rt_tbw)))))
 					{
@@ -2007,7 +2006,6 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const bool is_color, const
 			GL_CACHE("TC: src miss (0x%x, 0x%x, %s)", TEX0.TBP0, psm_s.pal > 0 ? TEX0.CBP : 0, psm_str(TEX0.PSM));
 		}
 #endif
-
 		// This is for the condition where the target doesn't exist on a shuffle and it needs to load from memory.
 		// The Godfather clears the depth buffer with a normal clear, so our depth target gets deleted, then because it finds no target
 		// it assumes it really is 16bits, causing the texture to be full of garbage, and our shuffle handling becomes a mess.
@@ -5667,7 +5665,7 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
 		if (is_8bits)
 		{
 			GL_INS("TC: Reading RT as a packed-indexed 8 bits format");
-			shader = ShaderConvert::RGBA_TO_8I;
+			shader = GSLocalMemory::m_psm[dst->m_TEX0.PSM].bpp == 16 ? ShaderConvert::RGB5A1_TO_8I : ShaderConvert::RGBA_TO_8I;
 		}
 
 #ifdef ENABLE_OGL_DEBUG
@@ -5737,7 +5735,11 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
 			{
 				// We're inside the target, so conversion needs to happen on the entire target so we can offset properly.
 				src->m_unscaled_size.x = dst->m_unscaled_size.x * 2;
-				src->m_unscaled_size.y = dst->m_unscaled_size.y * 2;
+				if (GSLocalMemory::m_psm[dst->m_TEX0.PSM].bpp == 32)
+					src->m_unscaled_size.y = dst->m_unscaled_size.y * 2;
+				else
+					src->m_unscaled_size.y = dst->m_unscaled_size.y;
+
 				new_size.x = src->m_unscaled_size.x;
 				new_size.y = src->m_unscaled_size.y;
 			}
@@ -5908,7 +5910,8 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
 
 					// Adjust to match a PSMT8 texture (coordinates are double C32, we shouldn't be converting from anything else).
 					x_offset *= 2;
-					y_offset *= 2;
+					if (GSLocalMemory::m_psm[dst->m_TEX0.PSM].bpp == 32) 
+						y_offset *= 2;
 
 					src->m_region.SetX(x_offset, x_offset + tw);
 					src->m_region.SetY(y_offset, y_offset + th);
@@ -6695,6 +6698,8 @@ void GSTextureCache::Read(Target* t, const GSVector4i& r)
 	{
 		case PSMCT32:
 		case PSMCT24:
+		case PSMZ32:
+		case PSMZ24:
 		{
 			// If we're downloading a depth buffer that's been reinterpreted as a color
 			// format, convert it to integer. The format/swizzle is likely wrong, but it's
@@ -6720,27 +6725,11 @@ void GSTextureCache::Read(Target* t, const GSVector4i& r)
 
 		case PSMCT16:
 		case PSMCT16S:
-		{
-			fmt = GSTexture::Format::UInt16;
-			ps_shader = is_depth ? ShaderConvert::FLOAT32_TO_16_BITS : ShaderConvert::RGBA8_TO_16_BITS;
-			dltex = &m_uint16_download_texture;
-		}
-		break;
-
-		case PSMZ32:
-		case PSMZ24:
-		{
-			fmt = GSTexture::Format::UInt32;
-			ps_shader = ShaderConvert::FLOAT32_TO_32_BITS;
-			dltex = &m_uint32_download_texture;
-		}
-		break;
-
 		case PSMZ16:
 		case PSMZ16S:
 		{
 			fmt = GSTexture::Format::UInt16;
-			ps_shader = ShaderConvert::FLOAT32_TO_16_BITS;
+			ps_shader = is_depth ? ShaderConvert::FLOAT32_TO_16_BITS : ShaderConvert::RGBA8_TO_16_BITS;
 			dltex = &m_uint16_download_texture;
 		}
 		break;

pcsx2/GS/Renderers/Metal/GSDeviceMTL.mm

@@ -1713,12 +1713,12 @@ void GSDeviceMTL::UpdateCLUTTexture(GSTexture* sTex, float sScale, u32 offsetX,
 
 void GSDeviceMTL::ConvertToIndexedTexture(GSTexture* sTex, float sScale, u32 offsetX, u32 offsetY, u32 SBW, u32 SPSM, GSTexture* dTex, u32 DBW, u32 DPSM)
 { @autoreleasepool {
-	const ShaderConvert shader = ShaderConvert::RGBA_TO_8I;
+	const ShaderConvert shader = ((SPSM & 0xE) == 0) ? ShaderConvert::RGBA_TO_8I : ShaderConvert::RGB5A1_TO_8I;
 	id<MTLRenderPipelineState> pipeline = m_convert_pipeline[static_cast<int>(shader)];
 	if (!pipeline)
 		[NSException raise:@"StretchRect Missing Pipeline" format:@"No pipeline for %d", static_cast<int>(shader)];
 
-	GSMTLIndexedConvertPSUniform uniform = { sScale, SBW, DBW };
+	GSMTLIndexedConvertPSUniform uniform = { sScale, SBW, DBW, SPSM };
 
 	const GSVector4 dRect(0, 0, dTex->GetWidth(), dTex->GetHeight());
 	DoStretchRect(sTex, GSVector4::zero(), dTex, dRect, pipeline, false, LoadAction::DontCareIfFull, &uniform, sizeof(uniform));

pcsx2/GS/Renderers/Metal/GSMTLSharedHeader.h

@@ -64,6 +64,7 @@ struct GSMTLIndexedConvertPSUniform
 	float scale;
 	uint sbw;
 	uint dbw;
+	uint psm;
 };
 
 struct GSMTLDownsamplePSUniform

pcsx2/GS/Renderers/Metal/convert.metal

@@ -296,6 +296,121 @@ fragment DepthOut ps_convert_rgb5a1_float16_biln(ConvertShaderData data [[stage_
 	return res.sample_biln<rgb5a1_to_depth16>(data.t);
 }
 
+fragment float4 ps_convert_rgb5a1_8i(ConvertShaderData data [[stage_in]], DirectReadTextureIn<float> res,
+	constant GSMTLIndexedConvertPSUniform& uniform [[buffer(GSMTLBufferIndexUniforms)]])
+{
+	// Convert a RGB5A1 texture into a 8 bits packed texture
+	// Input column: 16x2 RGB5A1 pixels
+	// 0: 16 RGBA
+	// 1: 16 RGBA
+	// Output column: 16x4 Index pixels
+	// 0: 16 R5G2
+	// 1: 16 R5G2
+	// 2: 16 G2B5A1
+	// 3: 16 G2B5A1
+	uint2 pos = uint2(data.p.xy);
+
+	// Collapse separate R G B A areas into their base pixel
+	uint2 column = (pos & ~uint2(0u, 3u)) / uint2(1,2);
+	uint2 subcolumn = (pos & uint2(0u, 1u));
+	column.x -= (column.x / 128) * 64;
+	column.y += (column.y / 32) * 32;
+
+	// Deal with swizzling differences
+	if ((uniform.psm & 0x8) != 0) // PSMCT16S
+	{
+		if ((pos.x & 32) != 0)
+		{
+			column.y += 32; // 4 columns high times 4 to get bottom 4 blocks
+			column.x &= ~32;
+		}
+		
+		if ((pos.x & 64) != 0)
+		{
+			column.x -= 32;
+		}
+		
+		if (((pos.x & 16) != 0) != ((pos.y & 16) != 0))
+		{
+			column.x ^= 16; 
+			column.y ^= 8;
+		}
+		
+		if ((uniform.psm & 0x30) != 0) // PSMZ16S - Untested but hopefully ok if anything uses it.
+		{
+			column.x ^= 32;
+			column.y ^= 16;
+		}
+	}
+	else // PSMCT16
+	{
+		if ((pos.y & 32) != 0)
+		{
+			column.y -= 16;
+			column.x += 32;
+		}
+		
+		if ((pos.x & 96) != 0)
+		{
+			uint multi = (pos.x & 96) / 32;
+			column.y += 16 * multi; // 4 columns high times 4 to get bottom 4 blocks
+			column.x -= (pos.x & 96);
+		}
+		
+		if (((pos.x & 16) != 0) != ((pos.y & 16) != 0))
+		{
+			column.x ^= 16; 
+			column.y ^= 8;
+		}
+		
+		if ((uniform.psm & 0x30) != 0) // PSMZ16 - Untested but hopefully ok if anything uses it.
+		{
+			column.x ^= 32;
+			column.y ^= 32;
+		}
+	}
+	
+	uint2 coord = column | subcolumn;
+
+	// Compensate for potentially differing page pitch.
+	uint2 block_xy = coord / uint2(64, 64);
+	uint block_num = (block_xy.y * (uniform.dbw / 128)) + block_xy.x;
+	uint2 block_offset = uint2((block_num % (uniform.sbw / 64)) * 64, (block_num / (uniform.sbw / 64)) * 64);
+	coord = (coord % uint2(64, 64)) + block_offset;
+
+	// Apply offset to cols 1 and 2
+	uint is_col23 = pos.y & 4;
+	uint is_col13 = pos.y & 2;
+	uint is_col12 = is_col23 ^ (is_col13 << 1);
+	coord.x ^= is_col12; // If cols 1 or 2, flip bit 3 of x
+
+	if (any(floor(uniform.scale) != uniform.scale))
+		coord = uint2(float2(coord) * uniform.scale);
+	else
+		coord = mul24(coord, uint2(uniform.scale));
+
+	float4 pixel = res.tex.read(coord);
+	
+	uint4 denorm_c = (uint4)(pixel * 255.5f);
+	if ((pos.y & 2u) == 0u)
+	{
+		uint red = (denorm_c.r >> 3) & 0x1Fu;
+		uint green = (denorm_c.g >> 3) & 0x1Fu;
+		float sel0 = (float)(((green << 5) | red) & 0xFF) / 255.0f;
+		
+		return float4(sel0);
+	}
+	else
+	{
+		uint green = (denorm_c.g >> 3) & 0x1Fu;
+		uint blue = (denorm_c.b >> 3) & 0x1Fu;
+		uint alpha = denorm_c.a & 0x80u;
+		float sel0 = (float)((alpha | (blue << 2) | (green >> 3)) & 0xFF) / 255.0f;
+
+		return float4(sel0);
+	}
+}
+
 fragment float4 ps_convert_rgba_8i(ConvertShaderData data [[stage_in]], DirectReadTextureIn<float> res,
 	constant GSMTLIndexedConvertPSUniform& uniform [[buffer(GSMTLBufferIndexUniforms)]])
 {

pcsx2/GS/Renderers/OpenGL/GSDeviceOGL.cpp

@@ -350,10 +350,11 @@ bool GSDeviceOGL::Create(GSVSyncMode vsync_mode, bool allow_present_throttle)
 				return false;
 			m_convert.ps[i].SetFormattedName("Convert pipe %s", name);
 
-			if (static_cast<ShaderConvert>(i) == ShaderConvert::RGBA_TO_8I)
+			if (static_cast<ShaderConvert>(i) == ShaderConvert::RGBA_TO_8I || static_cast<ShaderConvert>(i) == ShaderConvert::RGB5A1_TO_8I)
 			{
 				m_convert.ps[i].RegisterUniform("SBW");
 				m_convert.ps[i].RegisterUniform("DBW");
+				m_convert.ps[i].RegisterUniform("PSM");
 				m_convert.ps[i].RegisterUniform("ScaleFactor");
 			}
 			else if (static_cast<ShaderConvert>(i) == ShaderConvert::YUV)
@@ -1594,12 +1595,13 @@ void GSDeviceOGL::ConvertToIndexedTexture(GSTexture* sTex, float sScale, u32 off
 {
 	CommitClear(sTex, false);
 
-	const ShaderConvert shader = ShaderConvert::RGBA_TO_8I;
+	const ShaderConvert shader = ((SPSM & 0xE) == 0) ? ShaderConvert::RGBA_TO_8I : ShaderConvert::RGB5A1_TO_8I;
 	GLProgram& prog = m_convert.ps[static_cast<int>(shader)];
 	prog.Bind();
 	prog.Uniform1ui(0, SBW);
 	prog.Uniform1ui(1, DBW);
-	prog.Uniform1f(2, sScale);
+	prog.Uniform1ui(2, SPSM);
+	prog.Uniform1f(3, sScale);
 
 	OMSetDepthStencilState(m_convert.dss);
 	OMSetBlendState(false);

pcsx2/GS/Renderers/Vulkan/GSDeviceVK.cpp

@@ -3158,15 +3158,16 @@ void GSDeviceVK::ConvertToIndexedTexture(
 	{
 		u32 SBW;
 		u32 DBW;
-		u32 pad1[2];
+		u32 PSM;
+		u32 pad1[1];
 		float ScaleFactor;
 		float pad2[3];
 	};
 
-	const Uniforms uniforms = {SBW, DBW, {}, sScale, {}};
+	const Uniforms uniforms = {SBW, DBW, SPSM, {}, sScale, {}};
 	SetUtilityPushConstants(&uniforms, sizeof(uniforms));
 
-	const ShaderConvert shader = ShaderConvert::RGBA_TO_8I;
+	const ShaderConvert shader = ((SPSM & 0xE) == 0) ? ShaderConvert::RGBA_TO_8I : ShaderConvert::RGB5A1_TO_8I;
 	const GSVector4 dRect(0, 0, dTex->GetWidth(), dTex->GetHeight());
 	DoStretchRect(static_cast<GSTextureVK*>(sTex), GSVector4::zero(), static_cast<GSTextureVK*>(dTex), dRect,
 		m_convert[static_cast<int>(shader)], false, true);

pcsx2/ShaderCacheVersion.h

@@ -3,4 +3,4 @@
 
 /// Version number for GS and other shaders. Increment whenever any of the contents of the
 /// shaders change, to invalidate the cache.
-static constexpr u32 SHADER_CACHE_VERSION = 65;
+static constexpr u32 SHADER_CACHE_VERSION = 66;