TextureDecoder.cpp: new SSE2 optimized GX_TF_I8 decoder. Probably not ultimately optimal SSE2 code, but provably better (on my machine) than the memset version. Tested with __rdtsc counts in an independent project. I get about 6-7 FPS more on average during the intro movie playback in Mario Kart Wii. Hope this compiles for GCC okay.

TextureDecoder.cpp: merged two functionally identical decode5A3RGBA and decode5A3rgba methods.
OpcodeDecoding.cpp and DLCache.cpp: optimization for GX_LOAD_XF_REG. The PSUHFB solution sounds better for SSSE3, but this is a small win for the default case.

git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@6692 8ced0084-cf51-0410-be5f-012b33b47a6e

1 files changed, 26 insertions, 8 deletions

diff --git a/Source/Core/VideoCommon/Src/OpcodeDecoding.cpp b/Source/Core/VideoCommon/Src/OpcodeDecoding.cpp
index 4fb9574478..fbb81fb40d 100644
--- a/Source/Core/VideoCommon/Src/OpcodeDecoding.cpp
+++ b/Source/Core/VideoCommon/Src/OpcodeDecoding.cpp
@@ -49,6 +49,24 @@
 #endif
 
 u8* g_pVideoData = 0;
+DataReadU32xNfunc DataReadU32xFuncs[16] = {
+	DataReadU32xN<1>,
+	DataReadU32xN<2>,
+	DataReadU32xN<3>,
+	DataReadU32xN<4>,
+	DataReadU32xN<5>,
+	DataReadU32xN<6>,
+	DataReadU32xN<7>,
+	DataReadU32xN<8>,
+	DataReadU32xN<9>,
+	DataReadU32xN<10>,
+	DataReadU32xN<11>,
+	DataReadU32xN<12>,
+	DataReadU32xN<13>,
+	DataReadU32xN<14>,
+	DataReadU32xN<15>,
+	DataReadU32xN<16>
+};
 
 extern u8* FAKE_GetFifoStartPtr();
 extern u8* FAKE_GetFifoEndPtr();
@@ -233,12 +251,13 @@ static void Decode()
         {
             u32 Cmd2 = DataReadU32();
             int transfer_size = ((Cmd2 >> 16) & 15) + 1;
-            u32 xf_address = Cmd2 & 0xFFFF;
+			u32 xf_address = Cmd2 & 0xFFFF;
 			// TODO - speed this up. pshufb?
-            u32 data_buffer[16];
-            for (int i = 0; i < transfer_size; i++)
-                data_buffer[i] = DataReadU32();
+			u32 data_buffer[16];
+			DataReadU32xFuncs[transfer_size-1](data_buffer);
+
             LoadXFReg(transfer_size, xf_address, data_buffer);
+
 			INCSTAT(stats.thisFrame.numXFLoads);
         }
         break;
@@ -317,7 +336,7 @@ static void DecodeSemiNop()
             u8 sub_cmd = DataReadU8();
             u32 value = DataReadU32();
             LoadCPReg(sub_cmd, value);
-			INCSTAT(stats.thisFrame.numCPLoads);
+            INCSTAT(stats.thisFrame.numCPLoads);
         }
         break;
 
@@ -328,10 +347,9 @@ static void DecodeSemiNop()
             u32 address = Cmd2 & 0xFFFF;
 			// TODO - speed this up. pshufb?
             u32 data_buffer[16];
-            for (int i = 0; i < transfer_size; i++)
-                data_buffer[i] = DataReadU32();
+            DataReadU32xFuncs[transfer_size-1](data_buffer);
             LoadXFReg(transfer_size, address, data_buffer);
-			INCSTAT(stats.thisFrame.numXFLoads);
+            INCSTAT(stats.thisFrame.numXFLoads);
         }
         break;