zzogl-pg: Fix an issue I noticed in the new Mem Transfer code.

git-svn-id: http://pcsx2.googlecode.com/svn/trunk@2820 96395faa-99c1-11dd-bbfe-3dabce05a288
This commit is contained in:
arcum42 2010-04-07 11:59:03 +00:00
parent 9c398f3d91
commit 6b37b43e96
3 changed files with 192 additions and 148 deletions

View File

@ -92,10 +92,12 @@ int TransferHostLocal##psm(const void* pbyMem, u32 nQWordSize) \
\
if( ((gs.imageEndX-gs.trxpos.dx)%widthlimit) || ((gs.imageEndX-tempX)%widthlimit) ) { \
/* transmit with a width of 1 */ \
if (!TransmitHostLocalY##TransSfx<T>(wp, (1 + (DSTPSM == 0x14)), endY, pbuf)) goto End; \
pbuf = TransmitHostLocalY##TransSfx<T>(wp, (1 + (DSTPSM == 0x14)), endY, pbuf);\
if (pbuf == NULL) goto End; \
} \
else { \
if (!TransmitHostLocalY##TransSfx<T>(wp, widthlimit, endY, pbuf)) goto End; \
pbuf = TransmitHostLocalY##TransSfx<T>(wp, widthlimit, endY, pbuf);\
if (pbuf == NULL) goto End; \
} \
\
if( nSize == 0 || tempY == gs.imageEndY ) \
@ -132,7 +134,8 @@ int TransferHostLocal##psm(const void* pbyMem, u32 nQWordSize) \
\
/* transfer the rest */ \
if( alignedX < gs.imageEndX ) { \
if (!TransmitHostLocalX##TransSfx<T>(wp, widthlimit, blockheight, alignedX, pbuf)) goto End; \
pbuf = TransmitHostLocalX##TransSfx<T>(wp, widthlimit, blockheight, alignedX, pbuf);\
if (pbuf == NULL) goto End; \
pbuf -= TransPitch(alignedX-gs.trxpos.dx, transfersize)/TSize; \
} \
else pbuf += (blockheight-1)* TransPitch(pitch, transfersize)/TSize; \
@ -140,7 +143,8 @@ int TransferHostLocal##psm(const void* pbyMem, u32 nQWordSize) \
} \
\
if( TransPitch(nSize, transfersize)/4 > 0 ) { \
if (!TransmitHostLocalY##TransSfx<T>(wp, widthlimit, gs.imageEndY, pbuf)) goto End; \
pbuf = TransmitHostLocalY##TransSfx<T>(wp, widthlimit, gs.imageEndY, pbuf);\
if (pbuf == NULL) goto End; \
/* sometimes wrong sizes are sent (tekken tag) */ \
assert( gs.imageTransfer == -1 || TransPitch(nSize, transfersize)/4 <= 2 ); \
} \
@ -166,7 +170,7 @@ End: \
// Get ready for the same function 3 times. *sigh*
template <class T>
static __forceinline bool AlignOnBlockBoundry_(TransferData data, TransferFuncts fun, Point alignedPt, int& endY, const T* pbuf)
static __forceinline const T* AlignOnBlockBoundry_(TransferData data, TransferFuncts fun, Point alignedPt, int& endY, const T* pbuf)
{
bool bCanAlign = ((MOD_POW2(gs.trxpos.dx, data.blockwidth) == 0) && (gs.imageX == gs.trxpos.dx) &&
(alignedPt.y > endY) && (alignedPt.x > gs.trxpos.dx));
@ -207,16 +211,17 @@ static __forceinline bool AlignOnBlockBoundry_(TransferData data, TransferFuncts
}
// The only line that's different in these 3 functions.
if (!TransmitHostLocalY_<T>(fun.wp, transwidth, endY, pbuf)) return false;
pbuf = TransmitHostLocalY_<T>(fun.wp, transwidth, endY, pbuf);
if (pbuf == NULL) return NULL;
if( nSize == 0 || tempY == gs.imageEndY ) return false;
if( nSize == 0 || tempY == gs.imageEndY ) return NULL;
}
return true;
return pbuf;
}
template <class T>
static __forceinline bool AlignOnBlockBoundry_4(TransferData data, TransferFuncts fun, Point alignedPt, int& endY, const T* pbuf)
static __forceinline const T* AlignOnBlockBoundry_4(TransferData data, TransferFuncts fun, Point alignedPt, int& endY, const T* pbuf)
{
bool bCanAlign = ((MOD_POW2(gs.trxpos.dx, data.blockwidth) == 0) && (gs.imageX == gs.trxpos.dx) &&
(alignedPt.y > endY) && (alignedPt.x > gs.trxpos.dx));
@ -257,16 +262,17 @@ static __forceinline bool AlignOnBlockBoundry_4(TransferData data, TransferFunct
}
// The only line that's different in these 3 functions.
if (!TransmitHostLocalY_4<T>(fun.wp, transwidth, endY, pbuf)) return false;
pbuf = TransmitHostLocalY_4<T>(fun.wp, transwidth, endY, pbuf);
if (pbuf == NULL) return NULL;
if( nSize == 0 || tempY == gs.imageEndY ) return false;
if( nSize == 0 || tempY == gs.imageEndY ) return NULL;
}
return true;
return pbuf;
}
template <class T>
static __forceinline bool AlignOnBlockBoundry_24(TransferData data, TransferFuncts fun, Point alignedPt, int& endY, const T* pbuf)
static __forceinline const T* AlignOnBlockBoundry_24(TransferData data, TransferFuncts fun, Point alignedPt, int& endY, const T* pbuf)
{
bool bCanAlign = ((MOD_POW2(gs.trxpos.dx, data.blockwidth) == 0) && (gs.imageX == gs.trxpos.dx) &&
(alignedPt.y > endY) && (alignedPt.x > gs.trxpos.dx));
@ -307,16 +313,17 @@ static __forceinline bool AlignOnBlockBoundry_24(TransferData data, TransferFunc
}
// The only line that's different in these 3 functions.
if (!TransmitHostLocalY_24<T>(fun.wp, transwidth, endY, pbuf)) return false;
pbuf = TransmitHostLocalY_24<T>(fun.wp, transwidth, endY, pbuf);
if (pbuf == NULL) return NULL;
if( nSize == 0 || tempY == gs.imageEndY ) return false;
if( nSize == 0 || tempY == gs.imageEndY ) return NULL;
}
return true;
return pbuf;
}
// Here we go again. 3 nearly identical functions.
template <class T>
static __forceinline bool TransferAligningToBlocks_(TransferData data, TransferFuncts fun, Point alignedPt, const T* pbuf)
static __forceinline const T* TransferAligningToBlocks_(TransferData data, TransferFuncts fun, Point alignedPt, const T* pbuf)
{
bool bAligned;
const u32 TSize = sizeof(T);
@ -348,7 +355,8 @@ static __forceinline bool TransferAligningToBlocks_(TransferData data, TransferF
if( alignedPt.x < gs.imageEndX )
{
// The only line that's different in these 3 functions.
if (!TransmitHostLocalX_<T>(fun.wp, data.widthlimit, data.blockheight, alignedPt.x, pbuf)) return false;
pbuf = TransmitHostLocalX_<T>(fun.wp, data.widthlimit, data.blockheight, alignedPt.x, pbuf);
if (pbuf == NULL) return NULL;
pbuf -= TransPitch((alignedPt.x - gs.trxpos.dx), data.transfersize)/TSize;
}
else
@ -358,11 +366,11 @@ static __forceinline bool TransferAligningToBlocks_(TransferData data, TransferF
tempX = gs.trxpos.dx;
}
return true;
return pbuf;
}
template <class T>
static __forceinline bool TransferAligningToBlocks_4(TransferData data, TransferFuncts fun, Point alignedPt, const T* pbuf)
static __forceinline const T* TransferAligningToBlocks_4(TransferData data, TransferFuncts fun, Point alignedPt, const T* pbuf)
{
bool bAligned;
const u32 TSize = sizeof(T);
@ -394,7 +402,8 @@ static __forceinline bool TransferAligningToBlocks_4(TransferData data, Transfer
if( alignedPt.x < gs.imageEndX )
{
// The only line that's different in these 3 functions.
if (!TransmitHostLocalX_4<T>(fun.wp, data.widthlimit, data.blockheight, alignedPt.x, pbuf)) return false;
pbuf = TransmitHostLocalX_4<T>(fun.wp, data.widthlimit, data.blockheight, alignedPt.x, pbuf);
if (pbuf == NULL) return NULL;
pbuf -= TransPitch((alignedPt.x - gs.trxpos.dx), data.transfersize)/TSize;
}
else
@ -404,11 +413,11 @@ static __forceinline bool TransferAligningToBlocks_4(TransferData data, Transfer
tempX = gs.trxpos.dx;
}
return true;
return pbuf;
}
template <class T>
static __forceinline bool TransferAligningToBlocks_24(TransferData data, TransferFuncts fun, Point alignedPt, const T* pbuf)
static __forceinline const T* TransferAligningToBlocks_24(TransferData data, TransferFuncts fun, Point alignedPt, const T* pbuf)
{
bool bAligned;
const u32 TSize = sizeof(T);
@ -440,7 +449,8 @@ static __forceinline bool TransferAligningToBlocks_24(TransferData data, Transfe
if( alignedPt.x < gs.imageEndX )
{
// The only line that's different in these 3 functions.
if (!TransmitHostLocalX_24<T>(fun.wp, data.widthlimit, data.blockheight, alignedPt.x, pbuf)) return false;
pbuf = TransmitHostLocalX_24<T>(fun.wp, data.widthlimit, data.blockheight, alignedPt.x, pbuf);
if (pbuf == NULL) return NULL;
pbuf -= TransPitch((alignedPt.x - gs.trxpos.dx), data.transfersize)/TSize;
}
else
@ -450,7 +460,7 @@ static __forceinline bool TransferAligningToBlocks_24(TransferData data, Transfe
tempX = gs.trxpos.dx;
}
return true;
return pbuf;
}
// Only one of this function, since no TransmitHostLocalX_ or TransmitHostLocalY_'s were involved.
@ -495,9 +505,11 @@ static __forceinline int RealTransfer_(TransferData data, TransferFuncts fun, co
alignedPt.y = ROUND_DOWNPOW2(gs.imageEndY, data.blockheight);
alignedPt.x = ROUND_DOWNPOW2(gs.imageEndX, data.blockwidth);
if (!AlignOnBlockBoundry_<T>(data, fun, alignedPt, endY, pbuf)) return FinishTransfer(data, nLeftOver);
pbuf = AlignOnBlockBoundry_<T>(data, fun, alignedPt, endY, pbuf);
if (pbuf == NULL) return FinishTransfer(data, nLeftOver);
if (!TransferAligningToBlocks_<T>(data, fun, alignedPt, pbuf)) return FinishTransfer(data, nLeftOver);
pbuf = TransferAligningToBlocks_<T>(data, fun, alignedPt, pbuf);
if (pbuf == NULL) return FinishTransfer(data, nLeftOver);
if (TransPitch(nSize, data.transfersize)/4 > 0)
{
@ -529,9 +541,11 @@ static __forceinline int RealTransfer_4(TransferData data, TransferFuncts fun, c
alignedPt.y = ROUND_DOWNPOW2(gs.imageEndY, data.blockheight);
alignedPt.x = ROUND_DOWNPOW2(gs.imageEndX, data.blockwidth);
if (!AlignOnBlockBoundry_4<T>(data, fun, alignedPt, endY, pbuf)) return FinishTransfer(data, nLeftOver);
pbuf = AlignOnBlockBoundry_4<T>(data, fun, alignedPt, endY, pbuf);
if (pbuf == NULL) return FinishTransfer(data, nLeftOver);
if (!TransferAligningToBlocks_4<T>(data, fun, alignedPt, pbuf)) return FinishTransfer(data, nLeftOver);
pbuf = TransferAligningToBlocks_4<T>(data, fun, alignedPt, pbuf);
if (pbuf == NULL) return FinishTransfer(data, nLeftOver);
if (TransPitch(nSize, data.transfersize)/4 > 0)
{
@ -563,9 +577,11 @@ static __forceinline int RealTransfer_24(TransferData data, TransferFuncts fun,
alignedPt.y = ROUND_DOWNPOW2(gs.imageEndY, data.blockheight);
alignedPt.x = ROUND_DOWNPOW2(gs.imageEndX, data.blockwidth);
if (!AlignOnBlockBoundry_24<T>(data, fun, alignedPt, endY, pbuf)) return FinishTransfer(data, nLeftOver);
pbuf = AlignOnBlockBoundry_24<T>(data, fun, alignedPt, endY, pbuf);
if (pbuf == NULL) return FinishTransfer(data, nLeftOver);
if (!TransferAligningToBlocks_24<T>(data, fun, alignedPt, pbuf)) return FinishTransfer(data, nLeftOver);
pbuf = TransferAligningToBlocks_24<T>(data, fun, alignedPt, pbuf);
if (pbuf == NULL) return FinishTransfer(data, nLeftOver);
if (TransPitch(nSize, data.transfersize)/4 > 0)
{

View File

@ -12,7 +12,7 @@ extern u8* pstart;
// transfers whole rows
template <class T>
static __forceinline bool TransmitHostLocalY_(_writePixel_0 wp, u32 widthlimit, u32 endY, const T *pbuf)
static __forceinline const T *TransmitHostLocalY_(_writePixel_0 wp, u32 widthlimit, u32 endY, const T *buf)
{
assert( (nSize%widthlimit) == 0 && widthlimit <= 4 );
if ((gs.imageEndX-gs.trxpos.dx) % widthlimit)
@ -21,34 +21,34 @@ static __forceinline bool TransmitHostLocalY_(_writePixel_0 wp, u32 widthlimit,
for(; tempY < endY; ++tempY)
{
for(; tempX < gs.imageEndX && nSize > 0; tempX += 1, nSize -= 1, pbuf += 1)
for(; tempX < gs.imageEndX && nSize > 0; tempX += 1, nSize -= 1, buf += 1)
{
/* write as many pixel at one time as possible */
wp(pstart, tempX%2048, tempY%2048, pbuf[0], gs.dstbuf.bw);
wp(pstart, tempX%2048, tempY%2048, buf[0], gs.dstbuf.bw);
}
}
}
for(; tempY < endY; ++tempY)
{
for(; tempX < gs.imageEndX && nSize > 0; tempX += widthlimit, nSize -= widthlimit, pbuf += widthlimit)
for(; tempX < gs.imageEndX && nSize > 0; tempX += widthlimit, nSize -= widthlimit, buf += widthlimit)
{
/* write as many pixel at one time as possible */
if( nSize < widthlimit ) return false;
if( nSize < widthlimit ) return NULL;
wp(pstart, tempX%2048, tempY%2048, pbuf[0], gs.dstbuf.bw);
wp(pstart, tempX%2048, tempY%2048, buf[0], gs.dstbuf.bw);
if( widthlimit > 1 )
{
wp(pstart, (tempX+1)%2048, tempY%2048, pbuf[1], gs.dstbuf.bw);
wp(pstart, (tempX+1)%2048, tempY%2048, buf[1], gs.dstbuf.bw);
if( widthlimit > 2 )
{
wp(pstart, (tempX+2)%2048, tempY%2048, pbuf[2], gs.dstbuf.bw);
wp(pstart, (tempX+2)%2048, tempY%2048, buf[2], gs.dstbuf.bw);
if( widthlimit > 3 )
{
wp(pstart, (tempX+3)%2048, tempY%2048, pbuf[3], gs.dstbuf.bw);
wp(pstart, (tempX+3)%2048, tempY%2048, buf[3], gs.dstbuf.bw);
}
}
}
@ -62,38 +62,24 @@ static __forceinline bool TransmitHostLocalY_(_writePixel_0 wp, u32 widthlimit,
else
{
assert( gs.imageTransfer == -1 || nSize*sizeof(T)/4 == 0 );
return false;
return NULL;
}
}
return true;
}
template <class T>
static __forceinline bool TransmitHostLocalX_(_writePixel_0 wp, u32 widthlimit, u32 blockheight, u32 startX, const T *pbuf)
{
for(u32 tempi = 0; tempi < blockheight; ++tempi)
{
for(tempX = startX; tempX < gs.imageEndX; tempX++, pbuf++)
{
wp(pstart, tempX%2048, (tempY+tempi)%2048, pbuf[0], gs.dstbuf.bw);
}
pbuf += pitch - fracX;
}
return true;
return buf;
}
// transfers whole rows
template <class T>
static __forceinline bool TransmitHostLocalY_24(_writePixel_0 wp, u32 widthlimit, u32 endY, const T *pbuf)
static __forceinline const T *TransmitHostLocalY_24(_writePixel_0 wp, u32 widthlimit, u32 endY, const T *buf)
{
if (widthlimit != 8 || ((gs.imageEndX-gs.trxpos.dx)%widthlimit))
{
//GS_LOG("Bad Transmission! %d %d, psm: %d\n", gs.trxpos.dx, gs.imageEndX, DSTPSM);
for(; tempY < endY; ++tempY)
{
for(; tempX < gs.imageEndX && nSize > 0; tempX += 1, nSize -= 1, pbuf += 3)
for(; tempX < gs.imageEndX && nSize > 0; tempX += 1, nSize -= 1, buf += 3)
{
wp(pstart, tempX%2048, tempY%2048, *(u32*)(pbuf), gs.dstbuf.bw);
wp(pstart, tempX%2048, tempY%2048, *(u32*)(buf), gs.dstbuf.bw);
}
if( tempX >= gs.imageEndX )
@ -104,7 +90,7 @@ static __forceinline bool TransmitHostLocalY_24(_writePixel_0 wp, u32 widthlimit
else
{
assert( gs.imageTransfer == -1 || nSize == 0 );
return false;
return NULL;
}
}
}
@ -113,20 +99,20 @@ static __forceinline bool TransmitHostLocalY_24(_writePixel_0 wp, u32 widthlimit
assert( /*(nSize%widthlimit) == 0 &&*/ widthlimit == 8 );
for(; tempY < endY; ++tempY)
{
for(; tempX < gs.imageEndX && nSize > 0; tempX += widthlimit, nSize -= widthlimit, pbuf += 3*widthlimit)
for(; tempX < gs.imageEndX && nSize > 0; tempX += widthlimit, nSize -= widthlimit, buf += 3*widthlimit)
{
if (nSize < widthlimit) return false;
if (nSize < widthlimit) return NULL;
/* write as many pixel at one time as possible */
wp(pstart, tempX%2048, tempY%2048, *(u32*)(pbuf+0), gs.dstbuf.bw);
wp(pstart, (tempX+1)%2048, tempY%2048, *(u32*)(pbuf+3), gs.dstbuf.bw);
wp(pstart, (tempX+2)%2048, tempY%2048, *(u32*)(pbuf+6), gs.dstbuf.bw);
wp(pstart, (tempX+3)%2048, tempY%2048, *(u32*)(pbuf+9), gs.dstbuf.bw);
wp(pstart, (tempX+4)%2048, tempY%2048, *(u32*)(pbuf+12), gs.dstbuf.bw);
wp(pstart, (tempX+5)%2048, tempY%2048, *(u32*)(pbuf+15), gs.dstbuf.bw);
wp(pstart, (tempX+6)%2048, tempY%2048, *(u32*)(pbuf+18), gs.dstbuf.bw);
wp(pstart, (tempX+7)%2048, tempY%2048, *(u32*)(pbuf+21), gs.dstbuf.bw);
wp(pstart, tempX%2048, tempY%2048, *(u32*)(buf+0), gs.dstbuf.bw);
wp(pstart, (tempX+1)%2048, tempY%2048, *(u32*)(buf+3), gs.dstbuf.bw);
wp(pstart, (tempX+2)%2048, tempY%2048, *(u32*)(buf+6), gs.dstbuf.bw);
wp(pstart, (tempX+3)%2048, tempY%2048, *(u32*)(buf+9), gs.dstbuf.bw);
wp(pstart, (tempX+4)%2048, tempY%2048, *(u32*)(buf+12), gs.dstbuf.bw);
wp(pstart, (tempX+5)%2048, tempY%2048, *(u32*)(buf+15), gs.dstbuf.bw);
wp(pstart, (tempX+6)%2048, tempY%2048, *(u32*)(buf+18), gs.dstbuf.bw);
wp(pstart, (tempX+7)%2048, tempY%2048, *(u32*)(buf+21), gs.dstbuf.bw);
}
if (tempX >= gs.imageEndX)
@ -144,57 +130,42 @@ static __forceinline bool TransmitHostLocalY_24(_writePixel_0 wp, u32 widthlimit
nSize = 0;
}
assert( gs.imageTransfer == -1 || nSize == 0 );
return false;
return NULL;
}
}
}
return true;
}
// transmit until endX, don't check size since it has already been prevalidated
template <class T>
static __forceinline bool TransmitHostLocalX_24(_writePixel_0 wp, u32 widthlimit, u32 blockheight, u32 startX, const T *pbuf)
{
for(u32 tempi = 0; tempi < blockheight; ++tempi)
{
for(tempX = startX; tempX < gs.imageEndX; tempX++, pbuf += 3)
{
wp(pstart, tempX%2048, (tempY+tempi)%2048, *(u32*)pbuf, gs.dstbuf.bw);
}
pbuf += 3*(pitch-fracX);
}
return true;
return buf;
}
// meant for 4bit transfers
template <class T>
static __forceinline bool TransmitHostLocalY_4(_writePixel_0 wp, u32 widthlimit, u32 endY, const T *pbuf)
static __forceinline const T *TransmitHostLocalY_4(_writePixel_0 wp, u32 widthlimit, u32 endY, const T *buf)
{
for(; tempY < endY; ++tempY)
{
for(; tempX < gs.imageEndX && nSize > 0; tempX += widthlimit, nSize -= widthlimit)
{
/* write as many pixel at one time as possible */
wp(pstart, tempX%2048, tempY%2048, *pbuf&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+1)%2048, tempY%2048, *pbuf>>4, gs.dstbuf.bw);
pbuf++;
wp(pstart, tempX%2048, tempY%2048, *buf&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+1)%2048, tempY%2048, *buf>>4, gs.dstbuf.bw);
buf++;
if ( widthlimit > 2 )
{
wp(pstart, (tempX+2)%2048, tempY%2048, *pbuf&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+3)%2048, tempY%2048, *pbuf>>4, gs.dstbuf.bw);
pbuf++;
wp(pstart, (tempX+2)%2048, tempY%2048, *buf&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+3)%2048, tempY%2048, *buf>>4, gs.dstbuf.bw);
buf++;
if( widthlimit > 4 )
{
wp(pstart, (tempX+4)%2048, tempY%2048, *pbuf&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+5)%2048, tempY%2048, *pbuf>>4, gs.dstbuf.bw);
pbuf++;
wp(pstart, (tempX+4)%2048, tempY%2048, *buf&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+5)%2048, tempY%2048, *buf>>4, gs.dstbuf.bw);
buf++;
if( widthlimit > 6 )
{
wp(pstart, (tempX+6)%2048, tempY%2048, *pbuf&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+7)%2048, tempY%2048, *pbuf>>4, gs.dstbuf.bw);
pbuf++;
wp(pstart, (tempX+6)%2048, tempY%2048, *buf&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+7)%2048, tempY%2048, *buf>>4, gs.dstbuf.bw);
buf++;
}
}
}
@ -207,26 +178,55 @@ static __forceinline bool TransmitHostLocalY_4(_writePixel_0 wp, u32 widthlimit,
else
{
assert( gs.imageTransfer == -1 || (nSize/32) == 0 );
return false;
return NULL;
}
}
return true;
return buf;
}
template <class T>
static __forceinline const T *TransmitHostLocalX_(_writePixel_0 wp, u32 widthlimit, u32 blockheight, u32 startX, const T *buf)
{
for(u32 tempi = 0; tempi < blockheight; ++tempi)
{
for(tempX = startX; tempX < gs.imageEndX; tempX++, buf++)
{
wp(pstart, tempX%2048, (tempY+tempi)%2048, buf[0], gs.dstbuf.bw);
}
buf += pitch - fracX;
}
return buf;
}
// transmit until endX, don't check size since it has already been prevalidated
template <class T>
static __forceinline bool TransmitHostLocalX_4(_writePixel_0 wp, u32 widthlimit, u32 blockheight, u32 startX, const T *pbuf)
static __forceinline const T *TransmitHostLocalX_24(_writePixel_0 wp, u32 widthlimit, u32 blockheight, u32 startX, const T *buf)
{
for(u32 tempi = 0; tempi < blockheight; ++tempi)
{
for(tempX = startX; tempX < gs.imageEndX; tempX+=2, pbuf++)
for(tempX = startX; tempX < gs.imageEndX; tempX++, buf += 3)
{
wp(pstart, tempX%2048, (tempY+tempi)%2048, pbuf[0]&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+1)%2048, (tempY+tempi)%2048, pbuf[0]>>4, gs.dstbuf.bw);
wp(pstart, tempX%2048, (tempY+tempi)%2048, *(u32*)buf, gs.dstbuf.bw);
}
pbuf += (pitch-fracX)/2;
buf += 3*(pitch-fracX);
}
return true;
return buf;
}
// transmit until endX, don't check size since it has already been prevalidated
template <class T>
static __forceinline const T *TransmitHostLocalX_4(_writePixel_0 wp, u32 widthlimit, u32 blockheight, u32 startX, const T *buf)
{
for(u32 tempi = 0; tempi < blockheight; ++tempi)
{
for(tempX = startX; tempX < gs.imageEndX; tempX+=2, buf++)
{
wp(pstart, tempX%2048, (tempY+tempi)%2048, buf[0]&0x0f, gs.dstbuf.bw);
wp(pstart, (tempX+1)%2048, (tempY+tempi)%2048, buf[0]>>4, gs.dstbuf.bw);
}
buf += (pitch-fracX)/2;
}
return buf;
}
// calculate pitch in source buffer

View File

@ -2835,6 +2835,64 @@ void InitTransferLocalHost()
ResolveInRange(start, end);
}
template <class T>
void TransferLocalHost(void* pbyMem, u32 nQWordSize, int& x, int& y, u8 *pstart, _readPixel_0 rp)
{
int i = x, j = y;
T* pbuf = (T*)pbyMem;
u32 nSize = nQWordSize*16/sizeof(T);
for(; i < gs.imageEndY; ++i)
{
for(; j < gs.imageEndX && nSize > 0; ++j, --nSize)
{
*pbuf++ = rp(pstart, j%2048, i%2048, gs.srcbuf.bw);
}
if( j >= gs.imageEndX )
{
assert( j == gs.imageEndX);
j = gs.trxpos.sx;
}
else
{
assert( nSize == 0 );
break;
}
}
}
void TransferLocalHost_24(void* pbyMem, u32 nQWordSize, int& x, int& y, u8 *pstart, _readPixel_0 rp)
{
int i = x, j = y;
u8* pbuf = (u8*)pbyMem;
u32 nSize = nQWordSize*16/3;
for(; i < gs.imageEndY; ++i)
{
for(; j < gs.imageEndX && nSize > 0; ++j, --nSize)
{
u32 p = rp(pstart, j%2048, i%2048, gs.srcbuf.bw);
pbuf[0] = (u8)p;
pbuf[1] = (u8)(p>>8);
pbuf[2] = (u8)(p>>16);
pbuf += 3;
}
if( j >= gs.imageEndX )
{
assert( j == gs.imageEndX);
j = gs.trxpos.sx;
}
else
{
assert( nSize == 0 );
break;
}
}
}
// left/right, top/down
void TransferLocalHost(void* pbyMem, u32 nQWordSize)
{
@ -2844,47 +2902,17 @@ void TransferLocalHost(void* pbyMem, u32 nQWordSize)
u8* pstart = g_pbyGSMemory + 256*gs.srcbuf.bp;
int i = gs.imageY, j = gs.imageX;
#define TRANSFERLOCALHOST(psm, T) { \
T* pbuf = (T*)pbyMem; \
u32 nSize = nQWordSize*16/sizeof(T); \
for(; i < gs.imageEndY; ++i) { \
for(; j < gs.imageEndX && nSize > 0; ++j, --nSize) { \
*pbuf++ = readPixel##psm##_0(pstart, j%2048, i%2048, gs.srcbuf.bw); \
} \
\
if( j >= gs.imageEndX ) { assert( j == gs.imageEndX); j = gs.trxpos.sx; } \
else { assert( nSize == 0 ); break; } \
} \
} \
#define TRANSFERLOCALHOST_24(psm) { \
u8* pbuf = (u8*)pbyMem; \
u32 nSize = nQWordSize*16/3; \
for(; i < gs.imageEndY; ++i) { \
for(; j < gs.imageEndX && nSize > 0; ++j, --nSize) { \
u32 p = readPixel##psm##_0(pstart, j%2048, i%2048, gs.srcbuf.bw); \
pbuf[0] = (u8)p; \
pbuf[1] = (u8)(p>>8); \
pbuf[2] = (u8)(p>>16); \
pbuf += 3; \
} \
\
if( j >= gs.imageEndX ) { assert( j == gs.imageEndX); j = gs.trxpos.sx; } \
else { assert( nSize == 0 ); break; } \
} \
} \
switch (gs.srcbuf.psm) {
case 0x0: TRANSFERLOCALHOST(32, u32); break;
case 0x1: TRANSFERLOCALHOST_24(24); break;
case 0x2: TRANSFERLOCALHOST(16, u16); break;
case 0xA: TRANSFERLOCALHOST(16S, u16); break;
case 0x13: TRANSFERLOCALHOST(8, u8); break;
case 0x1B: TRANSFERLOCALHOST(8H, u8); break;
case 0x30: TRANSFERLOCALHOST(32Z, u32); break;
case 0x31: TRANSFERLOCALHOST_24(24Z); break;
case 0x32: TRANSFERLOCALHOST(16Z, u16); break;
case 0x3A: TRANSFERLOCALHOST(16SZ, u16); break;
case PSMCT32: TransferLocalHost<u32>(pbyMem, nQWordSize, i, j, pstart, readPixel32_0); break;
case PSMCT24: TransferLocalHost_24(pbyMem, nQWordSize, i, j, pstart, readPixel24_0); break;
case PSMCT16: TransferLocalHost<u16>(pbyMem, nQWordSize, i, j, pstart, readPixel16_0); break;
case PSMCT16S: TransferLocalHost<u16>(pbyMem, nQWordSize, i, j, pstart, readPixel16S_0); break;
case PSMT8: TransferLocalHost<u8>(pbyMem, nQWordSize, i, j, pstart, readPixel8_0); break;
case PSMT8H: TransferLocalHost<u8>(pbyMem, nQWordSize, i, j, pstart, readPixel8H_0); break;
case PSMT32Z: TransferLocalHost<u32>(pbyMem, nQWordSize, i, j, pstart, readPixel32Z_0); break;
case PSMT24Z: TransferLocalHost_24(pbyMem, nQWordSize, i, j, pstart, readPixel24Z_0); break;
case PSMT16Z: TransferLocalHost<u16>(pbyMem, nQWordSize, i, j, pstart, readPixel16Z_0); break;
case PSMT16SZ: TransferLocalHost<u16>(pbyMem, nQWordSize, i, j, pstart, readPixel16SZ_0); break;
default: assert(0);
}