I missed some cases in r5901:

Because we only ever call Pos_ReadDirect (and through that, DataRead<T>)
from JIT generated code, the compiler may not get the heads-up to properly
prepare for run-time instantiation of those template functions.

Explicitly instantiating Pos_ReadDirect gets around that issue.

Also force DataRead* inline as gcc didn't always do that itself when the
DataRead functions in turn were called from (other) template functions.


git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@5902 8ced0084-cf51-0410-be5f-012b33b47a6e

Source/Core/Common/Src/CommonFuncs.h

@@ -121,9 +121,12 @@ inline u16 swap16(u16 _data) {return bswap_16(_data);}
 inline u32 swap32(u32 _data) {return bswap_32(_data);}
 inline u64 swap64(u64 _data) {return bswap_64(_data);}
 #elif __APPLE__
-inline u16 swap16(u16 _data) {return (_data >> 8) | (_data << 8);}
+inline __attribute__((always_inline)) u16 swap16(u16 _data)
-inline u32 swap32(u32 _data) {return __builtin_bswap32(_data);}
+	{return (_data >> 8) | (_data << 8);}
-inline u64 swap64(u64 _data) {return __builtin_bswap64(_data);}
+inline __attribute__((always_inline)) u32 swap32(u32 _data)
+	{return __builtin_bswap32(_data);}
+inline __attribute__((always_inline)) u64 swap64(u64 _data)
+	{return __builtin_bswap64(_data);}
 #else
 // Slow generic implementation.
 inline u16 swap16(u16 data) {return (data >> 8) | (data << 8);}

Source/Core/VideoCommon/Src/VertexLoader_Color.cpp

@@ -33,14 +33,14 @@
 extern int colIndex;
 extern int colElements[2];
 
-inline void _SetCol(u32 val)
+__forceinline void _SetCol(u32 val)
 {
 	*(u32*)VertexManager::s_pCurBufferPointer = val;
 	VertexManager::s_pCurBufferPointer += 4;
 	colIndex++;
 }
 
-void _SetCol4444(u16 val)
+__forceinline void _SetCol4444(u16 val)
 {
 	u32 col = Convert4To8(val & 0xF) << ASHIFT;
 	col |= Convert4To8((val >> 12) & 0xF) << RSHIFT;
@@ -49,7 +49,7 @@ void _SetCol4444(u16 val)
 	_SetCol(col);
 }
 
-void _SetCol6666(u32 val)
+__forceinline void _SetCol6666(u32 val)
 {
 	u32 col = Convert6To8((val >> 18) & 0x3F) << RSHIFT;
 	col |= Convert6To8((val >> 12) & 0x3F) << GSHIFT;
@@ -58,7 +58,7 @@ void _SetCol6666(u32 val)
 	_SetCol(col);
 }
 
-void _SetCol565(u16 val)
+__forceinline void _SetCol565(u16 val)
 {
 	u32 col = Convert5To8((val >> 11) & 0x1F) << RSHIFT;
 	col |= Convert6To8((val >> 5) & 0x3F) << GSHIFT;
@@ -67,12 +67,12 @@ void _SetCol565(u16 val)
 }
 
 
-inline u32 _Read24(const u8 *addr)
+__forceinline u32 _Read24(const u8 *addr)
 {
 	return addr[0] | (addr[1] << 8) | (addr[2] << 16) | 0xFF000000;
 }
 
-inline u32 _Read32(const u8 *addr)
+__forceinline u32 _Read32(const u8 *addr)
 {
 	return *(const u32 *)addr;
 }

Source/Core/VideoCommon/Src/VertexLoader_Position.cpp

@@ -92,17 +92,6 @@ void Pos_ReadDirect()
 	VertexManager::s_pCurBufferPointer += 12;
 }
 
-// Explicitly instantiate these functions to decrease the possibility of
-// symbol binding problems when (only) calling them from JIT compiled code.
-template void Pos_ReadDirect<u8,  true>();
-template void Pos_ReadDirect<s8,  true>();
-template void Pos_ReadDirect<u16, true>();
-template void Pos_ReadDirect<s16, true>();
-template void Pos_ReadDirect<u8,  false>();
-template void Pos_ReadDirect<s8,  false>();
-template void Pos_ReadDirect<u16, false>();
-template void Pos_ReadDirect<s16, false>();
-
 void LOADERDECL Pos_ReadDirect_UByte3()  { Pos_ReadDirect<u8,  true>(); }
 void LOADERDECL Pos_ReadDirect_Byte3()   { Pos_ReadDirect<s8,  true>(); }
 void LOADERDECL Pos_ReadDirect_UShort3() { Pos_ReadDirect<u16, true>(); }
@@ -162,7 +151,7 @@ inline void Pos_ReadIndex_Short(int Index)
 }
 
 template<bool three>
-inline void Pos_ReadIndex_Float(int Index)
+void Pos_ReadIndex_Float(int Index)
 {
 	const u32* pData = (const u32 *)(cached_arraybases[ARRAY_POSITION] + (Index * arraystrides[ARRAY_POSITION]));
 	((u32*)VertexManager::s_pCurBufferPointer)[0] = Common::swap32(pData[0]);
@@ -180,7 +169,7 @@ static const __m128i kMaskSwap32_3 = _mm_set_epi32(0xFFFFFFFFL, 0x08090A0BL, 0x0
 static const __m128i kMaskSwap32_2 = _mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0x04050607L, 0x00010203L);
 
 template<bool three>
-inline void Pos_ReadIndex_Float_SSSE3(int Index)
+void Pos_ReadIndex_Float_SSSE3(int Index)
 {
 	const u32* pData = (const u32 *)(cached_arraybases[ARRAY_POSITION] + (Index * arraystrides[ARRAY_POSITION]));
 	const __m128i a = _mm_loadu_si128((__m128i*)pData);
@@ -191,6 +180,27 @@ inline void Pos_ReadIndex_Float_SSSE3(int Index)
 }
 #endif
 
+// Explicitly instantiate these functions to decrease the possibility of
+// symbol binding problems when (only) calling them from JIT compiled code.
+template void Pos_ReadDirect<u8,  true>();
+template void Pos_ReadDirect<s8,  true>();
+template void Pos_ReadDirect<u16, true>();
+template void Pos_ReadDirect<s16, true>();
+template void Pos_ReadDirect<u8,  false>();
+template void Pos_ReadDirect<s8,  false>();
+template void Pos_ReadDirect<u16, false>();
+template void Pos_ReadDirect<s16, false>();
+template void Pos_ReadIndex_Byte<u8, true>(int Index);
+template void Pos_ReadIndex_Byte<s8, true>(int Index);
+template void Pos_ReadIndex_Short<u16, true>(int Index);
+template void Pos_ReadIndex_Short<s16, true>(int Index);
+template void Pos_ReadIndex_Float<true>(int Index);
+template void Pos_ReadIndex_Byte<u8, false>(int Index);
+template void Pos_ReadIndex_Byte<s8, false>(int Index);
+template void Pos_ReadIndex_Short<u16, false>(int Index);
+template void Pos_ReadIndex_Short<s16, false>(int Index);
+template void Pos_ReadIndex_Float<false>(int Index);
+
 // ==============================================================================
 // Index 8
 // ==============================================================================