diff --git a/GPU/Common/FragmentShaderGenerator.cpp b/GPU/Common/FragmentShaderGenerator.cpp
index 15f4067b8b..7722e769fd 100644
--- a/GPU/Common/FragmentShaderGenerator.cpp
+++ b/GPU/Common/FragmentShaderGenerator.cpp
@@ -260,7 +260,7 @@ bool GenerateFragmentShader(const FShaderID &id, char *buffer, const ShaderLangu
 				if (texFunc == GE_TEXFUNC_BLEND) {
 					WRITE(p, "float3 u_texenv : register(c%i);\n", CONST_PS_TEXENV);
 				}
-				WRITE(p, "bool u_texAlpha : register(c%i);\n", CONST_PS_TEXALPHA);
+				WRITE(p, "bool u_texAlpha : register(b%i);\n", CONST_PS_TEXALPHA);  // NOTE! "b" register, not "c"!
 			}
 			WRITE(p, "float3 u_fogcolor : register(c%i);\n", CONST_PS_FOGCOLOR);
 			if (texture3D) {
diff --git a/GPU/Common/FragmentShaderGenerator.h b/GPU/Common/FragmentShaderGenerator.h
index d69583452c..c2842f7ae4 100644
--- a/GPU/Common/FragmentShaderGenerator.h
+++ b/GPU/Common/FragmentShaderGenerator.h
@@ -23,7 +23,7 @@
 
 struct FShaderID;
 
-// D3D9 constants
+// D3D9 float constants
 
 #define CONST_PS_TEXENV 0
 #define CONST_PS_ALPHACOLORREF 1
@@ -36,10 +36,13 @@ struct FShaderID;
 #define CONST_PS_TEXCLAMP 8
 #define CONST_PS_TEXCLAMPOFF 9
 #define CONST_PS_MIPBIAS 10
-#define CONST_PS_TEXALPHA 11
 
 // For stencil upload
-#define CONST_PS_STENCILVALUE 12
+#define BCONST_PS_STENCILVALUE 11
+
+// D3D9 bool constants, they have their own register space.
+#define CONST_PS_TEXALPHA 0
+
 
 // Can technically be deduced from the fragment shader ID, but this is safer.
 enum class FragmentShaderFlags : u32 {
diff --git a/GPU/Common/ShaderId.h b/GPU/Common/ShaderId.h
index beab743309..62adaaf38a 100644
--- a/GPU/Common/ShaderId.h
+++ b/GPU/Common/ShaderId.h
@@ -6,9 +6,8 @@
 
 #include "Common/CommonFuncs.h"
 
-// TODO: There will be additional bits, indicating that groups of these will be
-// sent to the shader and processed there. This will cut down the number of shaders ("ubershader approach")
-// This is probably only really worth doing for lighting and bones.
+// VS_BIT_LIGHT_UBERSHADER indicates that some groups of these will be
+// sent to the shader and processed there. This cuts down the number of shaders ("ubershader approach").
 enum VShaderBit : uint8_t {
 	VS_BIT_LMODE = 0,
 	VS_BIT_IS_THROUGH = 1,