Bump minimum architecture to SSE 2 on the x86 platform (a Pentium 4 from 2001!)

CMakeLists.txt

@@ -174,20 +174,23 @@ if( MSVC )
 	# Disable run-time type information
 	set( ALL_C_FLAGS "/GF /Gy /GR-" )
 	
-	if( CMAKE_SIZEOF_VOID_P MATCHES "4")
+	# Use SSE 2 as minimum always as the true color drawers needs it for __vectorcall
-		# SSE2 option (to allow x87 in 32 bit and disallow extended feature sets which have not yet been checked for precision)
+	set( ALL_C_FLAGS "${ALL_C_FLAGS} /arch:SSE2")
-		option (ZDOOM_USE_SSE2	"Use SSE2 instruction set")
+
-		if (ZDOOM_USE_SSE2)
+#	if( CMAKE_SIZEOF_VOID_P MATCHES "4")
-			set( ALL_C_FLAGS "${ALL_C_FLAGS} /arch:SSE2")
+#		# SSE2 option (to allow x87 in 32 bit and disallow extended feature sets which have not yet been checked for precision)
-		else ()
+#		option (ZDOOM_USE_SSE2	"Use SSE2 instruction set")
-			if (MSVC_VERSION GREATER 1699)
+#		if (ZDOOM_USE_SSE2)
-				# On Visual C++ 2012 and later SSE2 is the default, so we need to switch it off explicitly
+#			set( ALL_C_FLAGS "${ALL_C_FLAGS} /arch:SSE2")
-				set( ALL_C_FLAGS "${ALL_C_FLAGS} /arch:IA32")
+#		else ()
-			endif ()
+#			if (MSVC_VERSION GREATER 1699)
-		endif ()
+#				# On Visual C++ 2012 and later SSE2 is the default, so we need to switch it off explicitly
-	else()
+#				set( ALL_C_FLAGS "${ALL_C_FLAGS} /arch:IA32")
-		set( ALL_C_FLAGS "${ALL_C_FLAGS} /arch:SSE2")
+#			endif ()
-	endif()
+#		endif ()
+#	else()
+#		set( ALL_C_FLAGS "${ALL_C_FLAGS} /arch:SSE2")
+#	endif()
 	
 	# Avoid CRT DLL dependancies in release builds, optionally generate assembly output for checking crash locations.
 	option( ZDOOM_GENERATE_ASM "Generate assembly output." OFF )

src/CMakeLists.txt

@@ -791,7 +791,6 @@ file( GLOB HEADER_FILES
 	xlat/*.h
 	swrenderer/*.h
 	swrenderer/drawers/*.h
-	swrenderer/drawers/*.php
 	swrenderer/scene/*.h
 	swrenderer/segments/*.h
 	swrenderer/line/*.h

src/swrenderer/drawers/r_draw_rgba.h

@@ -67,6 +67,13 @@ namespace swrenderer
 	#endif
 	#endif
 
+	// Force the compiler to use a calling convention that works for vector types
+	#if defined(_MSC_VER)
+	#define VECTORCALL __vectorcall
+	#else
+	#define VECTORCALL
+	#endif
+
 	class DrawFuzzColumnRGBACommand : public DrawerCommand
 	{
 		int _x;

src/swrenderer/drawers/r_draw_span32_sse2.h

@@ -152,7 +152,7 @@ namespace swrenderer
 		}
 
 		template<typename ShadeModeT, typename FilterModeT, typename TextureSizeT>
-		void Loop(DrawerThread *thread, TextureData texdata, ShadeConstants shade_constants)
+		FORCEINLINE void VECTORCALL Loop(DrawerThread *thread, TextureData texdata, ShadeConstants shade_constants)
 		{
 			using namespace DrawSpan32TModes;
 
@@ -264,7 +264,7 @@ namespace swrenderer
 		}
 
 		template<typename FilterModeT, typename TextureSizeT>
-		unsigned int Sample(uint32_t xbits, uint32_t ybits, uint32_t xstep, uint32_t ystep, uint32_t xfrac, uint32_t yfrac, uint32_t yshift, uint32_t xshift, uint32_t xmask, const uint32_t *source)
+		FORCEINLINE unsigned int VECTORCALL Sample(uint32_t xbits, uint32_t ybits, uint32_t xstep, uint32_t ystep, uint32_t xfrac, uint32_t yfrac, uint32_t yshift, uint32_t xshift, uint32_t xmask, const uint32_t *source)
 		{
 			using namespace DrawSpan32TModes;
 
@@ -319,7 +319,7 @@ namespace swrenderer
 		}
 
 		template<typename ShadeModeT>
-		__m128i Shade(__m128i fgcolor, __m128i mlight, unsigned int ifgcolor0, unsigned int ifgcolor1, int desaturate, __m128i inv_desaturate, __m128i shade_fade, __m128i shade_light, const DrawerLight *lights, int num_lights, __m128 viewpos_x)
+		FORCEINLINE __m128i VECTORCALL Shade(__m128i fgcolor, __m128i mlight, unsigned int ifgcolor0, unsigned int ifgcolor1, int desaturate, __m128i inv_desaturate, __m128i shade_fade, __m128i shade_light, const DrawerLight *lights, int num_lights, __m128 viewpos_x)
 		{
 			using namespace DrawSpan32TModes;
 
@@ -351,7 +351,7 @@ namespace swrenderer
 			return AddLights(material, fgcolor, lights, num_lights, viewpos_x);
 		}
 
-		__m128i AddLights(__m128i material, __m128i fgcolor, const DrawerLight *lights, int num_lights, __m128 viewpos_x)
+		FORCEINLINE __m128i VECTORCALL AddLights(__m128i material, __m128i fgcolor, const DrawerLight *lights, int num_lights, __m128 viewpos_x)
 		{
 			using namespace DrawSpan32TModes;
 
@@ -398,7 +398,7 @@ namespace swrenderer
 			return fgcolor;
 		}
 
-		__m128i Blend(__m128i fgcolor, __m128i bgcolor, uint32_t srcalpha, uint32_t destalpha, unsigned int ifgcolor0, unsigned int ifgcolor1)
+		FORCEINLINE __m128i VECTORCALL Blend(__m128i fgcolor, __m128i bgcolor, uint32_t srcalpha, uint32_t destalpha, unsigned int ifgcolor0, unsigned int ifgcolor1)
 		{
 			using namespace DrawSpan32TModes;
 

src/swrenderer/drawers/r_draw_sprite32_sse2.h

@@ -100,7 +100,7 @@ namespace swrenderer
 		}
 
 		template<typename ShadeModeT, typename FilterModeT>
-		void Loop(DrawerThread *thread, ShadeConstants shade_constants)
+		FORCEINLINE void VECTORCALL Loop(DrawerThread *thread, ShadeConstants shade_constants)
 		{
 			using namespace DrawSprite32TModes;
 
@@ -254,7 +254,7 @@ namespace swrenderer
 		}
 
 		template<typename FilterModeT>
-		unsigned int Sample(uint32_t frac, const uint32_t *source, const uint32_t *source2, const uint32_t *translation, int textureheight, uint32_t one, uint32_t texturefracx, uint32_t color, uint32_t srccolor)
+		FORCEINLINE unsigned int VECTORCALL Sample(uint32_t frac, const uint32_t *source, const uint32_t *source2, const uint32_t *translation, int textureheight, uint32_t one, uint32_t texturefracx, uint32_t color, uint32_t srccolor)
 		{
 			using namespace DrawSprite32TModes;
 
@@ -303,7 +303,7 @@ namespace swrenderer
 			}
 		}
 
-		unsigned int SampleShade(uint32_t frac, const uint32_t *source, const uint8_t *colormap)
+		FORCEINLINE unsigned int VECTORCALL SampleShade(uint32_t frac, const uint32_t *source, const uint8_t *colormap)
 		{
 			using namespace DrawSprite32TModes;
 
@@ -320,7 +320,7 @@ namespace swrenderer
 		}
 
 		template<typename ShadeModeT>
-		__m128i Shade(__m128i fgcolor, __m128i mlight, unsigned int ifgcolor0, unsigned int ifgcolor1, int desaturate, __m128i inv_desaturate, __m128i shade_fade, __m128i shade_light, __m128i lightcontrib)
+		FORCEINLINE __m128i VECTORCALL Shade(__m128i fgcolor, __m128i mlight, unsigned int ifgcolor0, unsigned int ifgcolor1, int desaturate, __m128i inv_desaturate, __m128i shade_fade, __m128i shade_light, __m128i lightcontrib)
 		{
 			using namespace DrawSprite32TModes;
 
@@ -359,7 +359,7 @@ namespace swrenderer
 			}
 		}
 
-		__m128i Blend(__m128i fgcolor, __m128i bgcolor, unsigned int ifgcolor0, unsigned int ifgcolor1, unsigned int ifgshade0, unsigned int ifgshade1, uint32_t srcalpha, uint32_t destalpha)
+		FORCEINLINE __m128i VECTORCALL Blend(__m128i fgcolor, __m128i bgcolor, unsigned int ifgcolor0, unsigned int ifgcolor1, unsigned int ifgshade0, unsigned int ifgshade1, uint32_t srcalpha, uint32_t destalpha)
 		{
 			using namespace DrawSprite32TModes;
 

src/swrenderer/drawers/r_draw_wall32_sse2.h

@@ -78,7 +78,7 @@ namespace swrenderer
 		}
 
 		template<typename ShadeModeT, typename FilterModeT>
-		void Loop(DrawerThread *thread, ShadeConstants shade_constants)
+		FORCEINLINE void VECTORCALL Loop(DrawerThread *thread, ShadeConstants shade_constants)
 		{
 			using namespace DrawWall32TModes;
 
@@ -205,7 +205,7 @@ namespace swrenderer
 		}
 
 		template<typename FilterModeT>
-		unsigned int Sample(uint32_t frac, const uint32_t *source, const uint32_t *source2, int textureheight, uint32_t one, uint32_t texturefracx)
+		FORCEINLINE unsigned int VECTORCALL Sample(uint32_t frac, const uint32_t *source, const uint32_t *source2, int textureheight, uint32_t one, uint32_t texturefracx)
 		{
 			using namespace DrawWall32TModes;
 
@@ -241,7 +241,7 @@ namespace swrenderer
 		}
 
 		template<typename ShadeModeT>
-		__m128i Shade(__m128i fgcolor, __m128i mlight, unsigned int ifgcolor0, unsigned int ifgcolor1, int desaturate, __m128i inv_desaturate, __m128i shade_fade, __m128i shade_light, const DrawerLight *lights, int num_lights, __m128 viewpos_z)
+		FORCEINLINE __m128i VECTORCALL Shade(__m128i fgcolor, __m128i mlight, unsigned int ifgcolor0, unsigned int ifgcolor1, int desaturate, __m128i inv_desaturate, __m128i shade_fade, __m128i shade_light, const DrawerLight *lights, int num_lights, __m128 viewpos_z)
 		{
 			using namespace DrawWall32TModes;
 
@@ -273,7 +273,7 @@ namespace swrenderer
 			return AddLights(material, fgcolor, lights, num_lights, viewpos_z);
 		}
 
-		__m128i AddLights(__m128i material, __m128i fgcolor, const DrawerLight *lights, int num_lights, __m128 viewpos_z)
+		FORCEINLINE __m128i VECTORCALL AddLights(__m128i material, __m128i fgcolor, const DrawerLight *lights, int num_lights, __m128 viewpos_z)
 		{
 			using namespace DrawWall32TModes;
 
@@ -320,7 +320,7 @@ namespace swrenderer
 			return fgcolor;
 		}
 
-		__m128i Blend(__m128i fgcolor, __m128i bgcolor, unsigned int ifgcolor0, unsigned int ifgcolor1, uint32_t srcalpha, uint32_t destalpha)
+		FORCEINLINE __m128i VECTORCALL Blend(__m128i fgcolor, __m128i bgcolor, unsigned int ifgcolor0, unsigned int ifgcolor1, uint32_t srcalpha, uint32_t destalpha)
 		{
 			using namespace DrawWall32TModes;