Let's try that again...

git-svn-id: http://pcsx2.googlecode.com/svn/trunk@1202 96395faa-99c1-11dd-bbfe-3dabce05a288

pcsx2/Elfheader.cpp

@@ -19,7 +19,7 @@
 #include "PrecompiledHeader.h"
 
 #include "Common.h"
-#include "CDVDisodrv.h"
+#include "CDVD/CDVDisodrv.h"
 
 using namespace std;
 

pcsx2/IopCommon.h

@@ -23,8 +23,8 @@
 #include "R3000A.h"
 #include "Common.h"
 
-#include "CdRom.h"
+#include "CDVD/CdRom.h"
-#include "CDVD.h"
+#include "CDVD/CDVD.h"
 
 #include "Sio.h"
 #include "Sif.h"

pcsx2/Misc.cpp

@@ -29,7 +29,7 @@
 #include "IopCommon.h"
 #include "HostGui.h"
 
-#include "CDVDisodrv.h"
+#include "CDVD/CDVDisodrv.h"
 #include "VUmicro.h"
 #include "VU.h"
 #include "iCore.h"

pcsx2/SaveState.cpp

@@ -21,7 +21,7 @@
 #include "IopCommon.h"
 #include "SaveState.h"
 
-#include "CDVDisodrv.h"
+#include "CDVD/CDVDisodrv.h"
 #include "VUmicro.h"
 #include "VU.h"
 #include "iCore.h"

pcsx2/configure.ac

@@ -134,4 +134,3 @@ echo " local plugin inis? $localinis"
 echo " custom cflags? $customcflags"
 echo " memcpy_fast? $memcpyfast"
 #echo " microVU? $microVU"
-echo " pcsx2_dir = '$pcsx2_dir'"

pcsx2/windows/WinMisc.cpp

@@ -17,7 +17,7 @@
  */
  
 #include "Win32.h"
-#include "cdvd.h"
+#include "CDVD/CDVD.h"
 
 static LARGE_INTEGER lfreq;
 

pcsx2/x86/iVif.cpp

@@ -25,34 +25,37 @@
 #include <xmmintrin.h>
 #include <emmintrin.h>
 
+//#define USE_OLD_IVIF_CODE
+
 // sse2 highly optimized vif (~200 separate functions are built) zerofrog(@gmail.com)
 extern u32 g_vif1Masks[48], g_vif0Masks[48];
 extern u32 g_vif1HasMask3[4], g_vif0HasMask3[4];
 
 // arranged in writearr, rowarr, colarr, updatearr
 static PCSX2_ALIGNED16(u32 s_maskarr[16][4]) = {
-	0xffffffff, 0x00000000, 0x00000000, 0xffffffff,
+	{0xffffffff, 0x00000000, 0x00000000, 0xffffffff},
-	0xffff0000, 0x0000ffff, 0x00000000, 0xffffffff,
+	{0xffff0000, 0x0000ffff, 0x00000000, 0xffffffff},
-	0xffff0000, 0x00000000, 0x0000ffff, 0xffffffff,
+	{0xffff0000, 0x00000000, 0x0000ffff, 0xffffffff},
-	0xffff0000, 0x00000000, 0x00000000, 0xffff0000,
+	{0xffff0000, 0x00000000, 0x00000000, 0xffff0000},
-	0x0000ffff, 0xffff0000, 0x00000000, 0xffffffff,
+	{0x0000ffff, 0xffff0000, 0x00000000, 0xffffffff},
-	0x00000000, 0xffffffff, 0x00000000, 0xffffffff,
+	{0x00000000, 0xffffffff, 0x00000000, 0xffffffff},
-	0x00000000, 0xffff0000, 0x0000ffff, 0xffffffff,
+	{0x00000000, 0xffff0000, 0x0000ffff, 0xffffffff},
-	0x00000000, 0xffff0000, 0x00000000, 0xffff0000,
+	{0x00000000, 0xffff0000, 0x00000000, 0xffff0000},
-	0x0000ffff, 0x00000000, 0xffff0000, 0xffffffff,
+	{0x0000ffff, 0x00000000, 0xffff0000, 0xffffffff},
-	0x00000000, 0x0000ffff, 0xffff0000, 0xffffffff,
+	{0x00000000, 0x0000ffff, 0xffff0000, 0xffffffff},
-	0x00000000, 0x00000000, 0xffffffff, 0xffffffff,
+	{0x00000000, 0x00000000, 0xffffffff, 0xffffffff},
-	0x00000000, 0x00000000, 0xffff0000, 0xffff0000,
+	{0x00000000, 0x00000000, 0xffff0000, 0xffff0000},
-	0x0000ffff, 0x00000000, 0x00000000, 0x0000ffff,
+	{0x0000ffff, 0x00000000, 0x00000000, 0x0000ffff},
-	0x00000000, 0x0000ffff, 0x00000000, 0x0000ffff,
+	{0x00000000, 0x0000ffff, 0x00000000, 0x0000ffff},
-	0x00000000, 0x00000000, 0x0000ffff, 0x0000ffff,
+	{0x00000000, 0x00000000, 0x0000ffff, 0x0000ffff},
-	0x00000000, 0x00000000, 0x00000000, 0x00000000
+	{0x00000000, 0x00000000, 0x00000000, 0x00000000}
 };
 
 extern u8 s_maskwrite[256];
 
 extern "C" PCSX2_ALIGNED16(u32 s_TempDecompress[4]) = {0};
 
+#if defined(_MSC_VER) || !defined(USE_OLD_IVIF_CODE)
 void __fastcall SetNewMask(u32* vif1masks, u32* hasmask, u32 mask, u32 oldmask)
 {
 	u32 i;
@@ -86,3 +89,48 @@ void __fastcall SetNewMask(u32* vif1masks, u32* hasmask, u32 mask, u32 oldmask)
 	}
 	FreezeXMMRegs(0);
 }
+
+#else // gcc
+//  After some experimentation, I'm putting the old code back in for now for testing purposes, as the 
+// other version reliably SegFaults when loading YuGiOh: Duelist of the Roses on Linux (when setting
+// r0 to _mm_load_si128, when vif1masks=0x846f670, hasmask=0x846f454, mask=0, and oldmask=5).
+// Seems to work everywhere else. It'll stay disabled for now, but it's easier for me to fiddle with if its in here. -arcum42
+void __fastcall SetNewMask(u32* vif1masks, u32* hasmask, u32 mask, u32 oldmask)
+{
+    u32 i;
+        u32 prev = 0;
+        FreezeXMMRegs(1);
+
+        for(i = 0; i < 4; ++i, mask >>= 8, oldmask >>= 8, vif1masks += 16) {
+
+                prev |= s_maskwrite[mask&0xff];//((mask&3)==3)||((mask&0xc)==0xc)||((mask&0x30)==0x30)||((mask&0xc0)==0xc0);
+                hasmask[i] = prev;
+
+                if( (mask&0xff) != (oldmask&0xff) ) {
+            u8* p0 = (u8*)&s_maskarr[mask&15][0];
+            u8* p1 = (u8*)&s_maskarr[(mask>>4)&15][0];
+
+            __asm__(".intel_syntax noprefix\n"
+                "movaps xmm0, [%0]\n"
+                "movaps xmm1, [%1]\n"
+                "movaps xmm2, xmm0\n"
+                "punpcklwd xmm0, xmm0\n"
+                "punpckhwd xmm2, xmm2\n"
+                "movaps xmm3, xmm1\n"
+                "punpcklwd xmm1, xmm1\n"
+                "punpckhwd xmm3, xmm3\n"
+                "movq [%2], xmm0\n"
+                "movq [%2+8], xmm1\n"
+                "movhps [%2+16], xmm0\n"
+                "movhps [%2+24], xmm1\n"
+                "movq [%2+32], xmm2\n"
+                "movq [%2+40], xmm3\n"
+                "movhps [%2+48], xmm2\n"
+                "movhps [%2+56], xmm3\n"
+                    ".att_syntax\n" : : "r"(p0), "r"(p1), "r"(vif1masks) );
+                }
+        }
+        FreezeXMMRegs(0);
+}
+
+#endif