Make the SSE3.1 VideoCommon code available in GCC builds.

The GCC model for extended instructions like these is that you compile
with -msse3 etc. These affect code generation for whole compilation units,
so the idea is that you have a separate .c file for each instruction set
class and then indirect to the desired one at runtime.

Without e.g. -msse4.1, the GCC built-ins used by <foointrin.h> are not
available. However, in our specific case of compiling with -msse2 and
wanting to use SSE3.1 code, enough built-ins are available that we only
need to provide a little hack for pshufb.

Upgrading this to also use SSE4.1 instructions doesn't appear feasible
without a lot of undesirable duplication of GCC built-in functions and
headers, so we'd probably have to move to the GCC model of separate
source files for that.


git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@6014 8ced0084-cf51-0410-be5f-012b33b47a6e

Source/Core/Common/Src/Common.h

@@ -145,4 +145,8 @@ extern const char *netplay_dolphin_ver;
 #define __chdir chdir
 #endif
 
+#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 3) || defined __APPLE__
+#define _M_SSE 0x301
+#endif
+
 #endif // _COMMON_H_

Source/Core/Common/Src/CommonFuncs.h

@@ -27,6 +27,19 @@
 template <bool> struct CompileTimeAssert;
 template<> struct CompileTimeAssert<true> {};
 
+#if defined __GNUC__ && !defined __SSSE3__
+#include <emmintrin.h>
+static __inline __m128i __attribute__((__always_inline__))
+_mm_shuffle_epi8(__m128i a, __m128i mask)
+{
+	__m128i result;
+	__asm__("pshufb %1, %0"
+		: "=x" (result)
+		: "xm" (mask), "0" (a));
+	return result;
+}
+#endif
+
 #ifndef _WIN32
 
 #include <errno.h>
@@ -44,6 +57,7 @@ size_t strnlen(const char *s, size_t n);
 		#define Crash() {asm ("int $3");}
 	#endif
 	#define ARRAYSIZE(A) (sizeof(A)/sizeof((A)[0]))
+
 inline u32 _rotl(u32 x, int shift) {
     shift &= 31;
     if (!shift) return x;
@@ -65,7 +79,7 @@ inline u64 _rotr64(u64 x, unsigned int shift){
 	unsigned int n = shift % 64;
 	return (x >> n) | (x << (64 - n));
 }
-	#define SLEEP(x) usleep(x*1000)
+
 #else // WIN32
 // Function Cross-Compatibility
 	#define strcasecmp _stricmp
@@ -80,7 +94,6 @@ char* strndup (char const *s, size_t n);
 	#define ftell _ftelli64
 	#define atoll _atoi64
 	#define stat64 _stat64
-	#define SLEEP(x) Sleep(x)
 
 	#if _M_IX86
 		#define Crash() {__asm int 3}

Source/Core/Core/Src/CoreParameter.cpp

@@ -45,9 +45,9 @@ SCoreStartupParameter::SCoreStartupParameter()
   bSkipIdle(true), bNTSC(false),
   bHLE_BS2(true), bUseFastMem(false),
   bLockThreads(false),
-  bEnableCheats(false), bSMC(false),
+  bEnableCheats(false),
   bRunCompareServer(false), bRunCompareClient(false),
-  bMMU(false), iTLBHack(0), SelectedLanguage(0), bWii(false),
+  bMMU(false), bSMC(false), iTLBHack(0), SelectedLanguage(0), bWii(false),
   bConfirmStop(false), bHideCursor(false), 
   bAutoHideCursor(false), bUsePanicHandlers(true),
   iRenderWindowXPos(0), iRenderWindowYPos(0),

Source/Core/VideoCommon/Src/TextureDecoder.cpp

@@ -15,14 +15,6 @@
 // Official SVN repository and contact information can be found at
 // http://code.google.com/p/dolphin-emu/
 
-#include <cmath>
-
-#if _M_SSE >= 0x401
-#include <smmintrin.h>
-#elif _M_SSE >= 0x301
-#include <tmmintrin.h>
-#endif
-
 #include "Common.h"
 //#include "VideoCommon.h" // to get debug logs
 
@@ -37,6 +29,14 @@
 
 #include "LookUpTables.h"
 
+#include <cmath>
+
+#if _M_SSE >= 0x401
+#include <smmintrin.h>
+#elif _M_SSE >= 0x301
+#include <tmmintrin.h>
+#endif
+
 bool TexFmt_Overlay_Enable=false;
 bool TexFmt_Overlay_Center=false;