/******************************Module*Header*******************************\ * Module Name: stretch.c * * Copyright (c) 1993-1996 Microsoft Corporation \**************************************************************************/ #include "precomp.h" #define STRETCH_MAX_EXTENT 32767 typedef DWORDLONG ULONGLONG; /******************************Public*Routine******************************\ * VOID vMgaDirectStretch8Narrow * * Hardware assisted stretchblt at 8bpp when the width is 7 or less, for * old MGAs. * \**************************************************************************/ VOID vMgaDirectStretch8Narrow( STR_BLT* pStrBlt) { BYTE* pjSrc; BYTE* pjDstEnd; ULONG ulDst; ULONG xAccum; ULONG xTmp; ULONG yTmp; ULONG ulDstScan; BYTE* pjDstScan; PDEV* ppdev = pStrBlt->ppdev; BYTE* pjBase = ppdev->pjBase; LONG yDst = pStrBlt->YDstStart; LONG xDst = pStrBlt->XDstStart; LONG xSrc = pStrBlt->XSrcStart; BYTE* pjSrcScan = pStrBlt->pjSrcScan + xSrc; LONG yCount = pStrBlt->YDstCount; LONG WidthX = pStrBlt->XDstEnd - xDst; ULONG xInt = pStrBlt->ulXDstToSrcIntCeil; ULONG xFrac = pStrBlt->ulXDstToSrcFracCeil; ULONG yAccum = pStrBlt->ulYFracAccumulator; ULONG yFrac = pStrBlt->ulYDstToSrcFracCeil; ULONG yInt = 0; yInt = pStrBlt->lDeltaSrc * pStrBlt->ulYDstToSrcIntCeil; ulDstScan = ppdev->ulYDstOrg + (yDst + ppdev->yOffset) * ppdev->cxMemory + (xDst + ppdev->xOffset); pjDstScan = ppdev->pjBase + SRCWND + (ulDstScan & 31); // We can't touch the frame buffer while the accelerator is doing // any drawing: CHECK_FIFO_SPACE(pjBase, FIFOSIZE); WAIT_NOT_BUSY(pjBase); do { ULONG yTmp = yAccum + yFrac; BYTE jSrc0; BYTE* pjDst; BYTE* pjDstEndNarrow; CP_WRITE_REGISTER(pjBase + HST_DSTPAGE, ulDstScan); ulDstScan += ppdev->cxMemory; // Increment to next scan pjDst = pjDstScan; pjSrc = pjSrcScan; xAccum = pStrBlt->ulXFracAccumulator; pjDstEndNarrow = pjDst + WidthX; do { jSrc0 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt + (xTmp < xAccum); *pjDst++ = jSrc0; xAccum = xTmp; } while (pjDst != pjDstEndNarrow); pjSrcScan += yInt; if (yTmp < yAccum) { pjSrcScan += pStrBlt->lDeltaSrc; } yAccum = yTmp; } while (--yCount); } /******************************Public*Routine******************************\ * VOID vMilDirectStretch8Narrow * * Hardware assisted stretchblt at 8bpp when the width is 7 or less, for * Millenniums. * \**************************************************************************/ VOID vMilDirectStretch8Narrow( STR_BLT* pStrBlt) { BYTE* pjSrc; BYTE* pjDstEnd; ULONG ulDst; ULONG xAccum; ULONG xTmp; ULONG yTmp; ULONG ulDstScan; BYTE* pjDstScan; PDEV* ppdev = pStrBlt->ppdev; BYTE* pjBase = ppdev->pjBase; LONG yDst = pStrBlt->YDstStart; LONG xDst = pStrBlt->XDstStart; LONG xSrc = pStrBlt->XSrcStart; BYTE* pjSrcScan = pStrBlt->pjSrcScan + xSrc; LONG yCount = pStrBlt->YDstCount; LONG WidthX = pStrBlt->XDstEnd - xDst; ULONG xInt = pStrBlt->ulXDstToSrcIntCeil; ULONG xFrac = pStrBlt->ulXDstToSrcFracCeil; ULONG yAccum = pStrBlt->ulYFracAccumulator; ULONG yFrac = pStrBlt->ulYDstToSrcFracCeil; ULONG yInt = 0; yInt = pStrBlt->lDeltaSrc * pStrBlt->ulYDstToSrcIntCeil; ulDstScan = ppdev->ulYDstOrg + (yDst + ppdev->yOffset) * ppdev->cxMemory + (xDst + ppdev->xOffset); pjDstScan = ppdev->pjScreen; // We can't touch the frame buffer while the accelerator is doing // any drawing: CHECK_FIFO_SPACE(pjBase, FIFOSIZE); WAIT_NOT_BUSY(pjBase); do { ULONG yTmp = yAccum + yFrac; BYTE jSrc0; BYTE* pjDst; BYTE* pjDstEndNarrow; pjDst = pjDstScan + ulDstScan; ulDstScan += ppdev->cxMemory; // Increment to next scan pjSrc = pjSrcScan; xAccum = pStrBlt->ulXFracAccumulator; pjDstEndNarrow = pjDst + WidthX; do { jSrc0 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt + (xTmp < xAccum); *pjDst++ = jSrc0; xAccum = xTmp; } while (pjDst != pjDstEndNarrow); pjSrcScan += yInt; if (yTmp < yAccum) { pjSrcScan += pStrBlt->lDeltaSrc; } yAccum = yTmp; } while (--yCount); } /******************************Public*Routine******************************\ * VOID vMilDirectStretch8 * * Hardware assisted stretchblt at 8bpp when the width is 8 or more, * for Millenniums. * \**************************************************************************/ VOID vMilDirectStretch8( STR_BLT* pStrBlt) { BYTE* pjSrc; BYTE* pjDstEnd; LONG WidthXAln; ULONG ulDst; ULONG xAccum; ULONG xTmp; ULONG yTmp; BYTE* pjOldScan; LONG cyDuplicate; LONG xDuplicate; LONG yDuplicate; LONG lDuplicate; ULONG yInt; PDEV* ppdev = pStrBlt->ppdev; BYTE* pjBase = ppdev->pjBase; LONG cxMemory = ppdev->cxMemory; LONG xDst = pStrBlt->XDstStart; LONG xSrc = pStrBlt->XSrcStart; BYTE* pjSrcScan = pStrBlt->pjSrcScan + xSrc; BYTE* pjDst = pStrBlt->pjDstScan + xDst; LONG yDst = pStrBlt->YDstStart; LONG yCount = pStrBlt->YDstCount; ULONG StartAln = (ULONG)((ULONG_PTR)pjDst & 0x03); LONG WidthX = pStrBlt->XDstEnd - xDst; ULONG EndAln = (ULONG)((ULONG_PTR)(pjDst + WidthX) & 0x03); ULONG xInt = pStrBlt->ulXDstToSrcIntCeil; ULONG xFrac = pStrBlt->ulXDstToSrcFracCeil; ULONG yAccum = pStrBlt->ulYFracAccumulator; ULONG yFrac = pStrBlt->ulYDstToSrcFracCeil; LONG lDstStride = pStrBlt->lDeltaDst - WidthX; LONG lDeltaDst = pStrBlt->lDeltaDst; LONG lDeltaSrc = pStrBlt->lDeltaSrc; WidthXAln = WidthX - EndAln - ((- (LONG) StartAln) & 0x03); yInt = pStrBlt->lDeltaSrc * pStrBlt->ulYDstToSrcIntCeil; yDuplicate = yDst + ppdev->yOffset - 1; xDuplicate = xDst + ppdev->xOffset + ppdev->ulYDstOrg; START_DIRECT_ACCESS_STORM(ppdev, pjBase); do { BYTE jSrc0,jSrc1,jSrc2,jSrc3; ULONG yTmp; pjSrc = pjSrcScan; xAccum = pStrBlt->ulXFracAccumulator; switch (StartAln) { case 1: jSrc0 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt; if (xTmp < xAccum) pjSrc++; *pjDst++ = jSrc0; xAccum = xTmp; case 2: jSrc0 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt; if (xTmp < xAccum) pjSrc++; *pjDst++ = jSrc0; xAccum = xTmp; case 3: jSrc0 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt; if (xTmp < xAccum) pjSrc++; *pjDst++ = jSrc0; xAccum = xTmp; } pjDstEnd = pjDst + WidthXAln; while (pjDst != pjDstEnd) { jSrc0 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt; if (xTmp < xAccum) pjSrc++; jSrc1 = *pjSrc; xAccum = xTmp + xFrac; pjSrc = pjSrc + xInt; if (xAccum < xTmp) pjSrc++; jSrc2 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt; if (xTmp < xAccum) pjSrc++; jSrc3 = *pjSrc; xAccum = xTmp + xFrac; pjSrc = pjSrc + xInt; if (xAccum < xTmp) pjSrc++; ulDst = (jSrc3 << 24) | (jSrc2 << 16) | (jSrc1 << 8) | jSrc0; *(PULONG)pjDst = ulDst; pjDst += 4; } switch (EndAln) { case 3: jSrc0 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt; if (xTmp < xAccum) pjSrc++; *pjDst++ = jSrc0; xAccum = xTmp; case 2: jSrc0 = *pjSrc; xTmp = xAccum + xFrac; pjSrc = pjSrc + xInt; if (xTmp < xAccum) pjSrc++; *pjDst++ = jSrc0; xAccum = xTmp; case 1: jSrc0 = *pjSrc; *pjDst++ = jSrc0; } // Now count the number of duplicate scans: pjDst += lDstStride; yDuplicate++; cyDuplicate = -1; pjOldScan = pjSrcScan; do { cyDuplicate++; pjSrcScan += yInt; yTmp = yAccum + yFrac; if (yTmp < yAccum) { pjSrcScan += lDeltaSrc; } yAccum = yTmp; yCount--; } while ((yCount != 0) && (pjSrcScan == pjOldScan)); // Duplicate the scan 'cyDuplicate' times with one blt: if (cyDuplicate != 0) { lDuplicate = yDuplicate * cxMemory + xDuplicate; CHECK_FIFO_SPACE(pjBase, 4); CP_WRITE(pjBase, DWG_AR3, lDuplicate); CP_WRITE(pjBase, DWG_AR0, lDuplicate + WidthX - 1); CP_WRITE(pjBase, DWG_DWGCTL, opcode_BITBLT | atype_RPL | blockm_OFF | bltmod_BFCOL | pattern_OFF | transc_BG_OPAQUE | bop_SRCCOPY | shftzero_ZERO | sgnzero_ZERO); CP_START(pjBase, DWG_YDSTLEN, ((yDuplicate + 1) << yval_SHIFT) | cyDuplicate); yDuplicate += cyDuplicate; pjDst += cyDuplicate * lDeltaDst; } } while (yCount != 0); END_DIRECT_ACCESS_STORM(ppdev, pjBase); } /******************************Public*Routine******************************\ * VOID vMilDirectStretch16 * * Hardware assisted stretchblt at 16bpp, for Millenniums. * \**************************************************************************/ VOID vMilDirectStretch16( STR_BLT* pStrBlt) { BYTE* pjOldScan; USHORT* pusSrc; USHORT* pusDstEnd; LONG WidthXAln; ULONG ulDst; ULONG xAccum; ULONG xTmp; ULONG yTmp; LONG cyDuplicate; LONG xDuplicate; LONG yDuplicate; LONG lDuplicate; LONG yInt; PDEV* ppdev = pStrBlt->ppdev; BYTE* pjBase = ppdev->pjBase; LONG cxMemory = ppdev->cxMemory; LONG xDst = pStrBlt->XDstStart; LONG xSrc = pStrBlt->XSrcStart; BYTE* pjSrcScan = (pStrBlt->pjSrcScan) + xSrc * 2; USHORT* pusDst = (USHORT*)(pStrBlt->pjDstScan) + xDst; LONG yDst = pStrBlt->YDstStart; LONG yCount = pStrBlt->YDstCount; ULONG StartAln = ((ULONG)((ULONG_PTR)pusDst & 0x02)) >> 1; LONG WidthX = pStrBlt->XDstEnd - xDst; ULONG EndAln = (ULONG)(((ULONG_PTR)(pusDst + WidthX) & 0x02) >> 1); ULONG xInt = pStrBlt->ulXDstToSrcIntCeil; ULONG xFrac = pStrBlt->ulXDstToSrcFracCeil; ULONG yAccum = pStrBlt->ulYFracAccumulator; ULONG yFrac = pStrBlt->ulYDstToSrcFracCeil; LONG lDstStride = pStrBlt->lDeltaDst - 2 * WidthX; LONG lDeltaDst = pStrBlt->lDeltaDst; LONG lDeltaSrc = pStrBlt->lDeltaSrc; WidthXAln = WidthX - EndAln - StartAln; yInt = pStrBlt->lDeltaSrc * (LONG)(pStrBlt->ulYDstToSrcIntCeil); yDuplicate = yDst + ppdev->yOffset - 1; xDuplicate = xDst + ppdev->xOffset + ppdev->ulYDstOrg; START_DIRECT_ACCESS_STORM(ppdev, pjBase); // Loop stretching each scan line do { USHORT usSrc0,usSrc1; ULONG yTmp; pusSrc = (USHORT*) pjSrcScan; xAccum = pStrBlt->ulXFracAccumulator; if (StartAln) { usSrc0 = *pusSrc; xTmp = xAccum + xFrac; pusSrc = pusSrc + xInt; if (xTmp < xAccum) pusSrc++; *pusDst++ = usSrc0; xAccum = xTmp; } pusDstEnd = pusDst + WidthXAln; while (pusDst != pusDstEnd) { usSrc0 = *pusSrc; xTmp = xAccum + xFrac; pusSrc = pusSrc + xInt; if (xTmp < xAccum) pusSrc++; usSrc1 = *pusSrc; xAccum = xTmp + xFrac; pusSrc = pusSrc + xInt; if (xAccum < xTmp) pusSrc++; ulDst = (ULONG)((usSrc1 << 16) | usSrc0); *(ULONG*)pusDst = ulDst; pusDst+=2; } if (EndAln) { usSrc0 = *pusSrc; *pusDst++ = usSrc0; } // Now count the number of duplicate scans: pusDst = (USHORT*) ((BYTE*) pusDst + lDstStride); yDuplicate++; cyDuplicate = -1; pjOldScan = pjSrcScan; do { cyDuplicate++; pjSrcScan += yInt; yTmp = yAccum + yFrac; if (yTmp < yAccum) { pjSrcScan += lDeltaSrc; } yAccum = yTmp; yCount--; } while ((yCount != 0) && (pjSrcScan == pjOldScan)); // Duplicate the scan 'cyDuplicate' times with one blt: if (cyDuplicate != 0) { lDuplicate = yDuplicate * cxMemory + xDuplicate; CHECK_FIFO_SPACE(pjBase, 4); CP_WRITE(pjBase, DWG_AR3, lDuplicate); CP_WRITE(pjBase, DWG_AR0, lDuplicate + WidthX - 1); CP_WRITE(pjBase, DWG_DWGCTL, opcode_BITBLT | atype_RPL | blockm_OFF | bltmod_BFCOL | pattern_OFF | transc_BG_OPAQUE | bop_SRCCOPY | shftzero_ZERO | sgnzero_ZERO); CP_START(pjBase, DWG_YDSTLEN, ((yDuplicate + 1) << yval_SHIFT) | cyDuplicate); yDuplicate += cyDuplicate; pusDst = (USHORT*) ((BYTE*) pusDst + cyDuplicate * lDeltaDst); } } while (yCount != 0); END_DIRECT_ACCESS_STORM(ppdev, pjBase); } /******************************Public*Routine******************************\ * VOID vMilDirectStretch24 * * Hardware assisted stretchblt at 24bpp, for Millenniums. * * We use the data-transfer register so that we don't have to worry about * funky alignments. * \**************************************************************************/ VOID vMilDirectStretch24( STR_BLT* pStrBlt) { BYTE* pjSrc; LONG lAddress; ULONG ulDst; ULONG xAccum; ULONG xTmp; ULONG yTmp; LONG i; LONG cyDuplicate; LONG xDuplicate; LONG yDuplicate; LONG lDuplicate; BYTE* pjOldScan; LONG xDstLeft; LONG xDstRight; LONG xDstRightFast; LONG yDstTop; LONG cyBreak; LONG iBreak; PDEV* ppdev = pStrBlt->ppdev; BYTE* pjBase = ppdev->pjBase; LONG cxMemory = ppdev->cxMemory; BYTE* pjSrcScan = pStrBlt->pjSrcScan + pStrBlt->XSrcStart * 3; LONG yCount = pStrBlt->YDstCount; LONG WidthX = pStrBlt->XDstEnd - pStrBlt->XDstStart; ULONG xInt = pStrBlt->ulXDstToSrcIntCeil * 3; ULONG xFrac = pStrBlt->ulXDstToSrcFracCeil; ULONG yAccum = pStrBlt->ulYFracAccumulator; ULONG yFrac = pStrBlt->ulYDstToSrcFracCeil; LONG lDeltaSrc = pStrBlt->lDeltaSrc; ULONG yInt = pStrBlt->lDeltaSrc * pStrBlt->ulYDstToSrcIntCeil; ULONG ulXFracAccumulator = pStrBlt->ulXFracAccumulator; yDstTop = ppdev->yOffset + pStrBlt->YDstStart; xDstLeft = ppdev->xOffset + pStrBlt->XDstStart; xDstRight = ppdev->xOffset + pStrBlt->XDstEnd - 1; // Note inclusive xDstRightFast = (((xDstRight * 3) + 2) | 0x40) / 3; // Figure out how many scans we can duplicate before we hit the first // WRAM boundary: cyBreak = 0xffff; for (iBreak = 0; iBreak < ppdev->cyBreak; iBreak++) { if (ppdev->ayBreak[iBreak] >= yDstTop) { cyBreak = ppdev->ayBreak[iBreak] - yDstTop; break; } } ppdev->HopeFlags = SIGN_CACHE; CHECK_FIFO_SPACE(pjBase, 6); CP_WRITE(pjBase, DWG_FXBNDRY, (xDstRight << bfxright_SHIFT) | xDstLeft); CP_WRITE(pjBase, DWG_AR5, cxMemory); CP_WRITE(pjBase, DWG_AR3, 0); CP_WRITE(pjBase, DWG_AR0, WidthX - 1); CP_WRITE(pjBase, DWG_YDST, yDstTop); CP_WRITE(pjBase, DWG_CXRIGHT, xDstRight); // For fast-blt work-around yDuplicate = yDstTop - 1; xDuplicate = xDstLeft + ppdev->ulYDstOrg; do { CHECK_FIFO_SPACE(pjBase, 2); CP_WRITE(pjBase, DWG_DWGCTL, opcode_ILOAD | atype_RPL | blockm_OFF | pattern_OFF | transc_BG_OPAQUE | bop_SRCCOPY | bltmod_BFCOL | shftzero_ZERO | sgnzero_ZERO); CP_START(pjBase, DWG_LEN, 1); // Make sure the MGA is ready to take the data: CHECK_FIFO_SPACE(pjBase, FIFOSIZE); pjSrc = pjSrcScan; xAccum = ulXFracAccumulator; i = WidthX; while (TRUE) { ulDst = *(pjSrc); // Pixel 0 ulDst |= *(pjSrc + 1) << 8; ulDst |= *(pjSrc + 2) << 16; if (--i == 0) break; pjSrc += xInt; xTmp = xAccum + xFrac; if (xTmp < xAccum) pjSrc += 3; ulDst |= *(pjSrc) << 24; // Pixel 1 CP_WRITE_SRC(pjBase, ulDst); ulDst = *(pjSrc + 1); ulDst |= *(pjSrc + 2) << 8; if (--i == 0) break; pjSrc += xInt; xAccum = xTmp + xFrac; if (xAccum < xTmp) pjSrc += 3; ulDst |= *(pjSrc) << 16; // Pixel 2 ulDst |= *(pjSrc + 1) << 24; CP_WRITE_SRC(pjBase, ulDst); ulDst = *(pjSrc + 2); if (--i == 0) break; pjSrc += xInt; xTmp = xAccum + xFrac; if (xTmp < xAccum) pjSrc += 3; ulDst |= *(pjSrc) << 8; // Pixel 3 ulDst |= *(pjSrc + 1) << 16; ulDst |= *(pjSrc + 2) << 24; if (--i == 0) break; CP_WRITE_SRC(pjBase, ulDst); pjSrc += xInt; xAccum = xTmp + xFrac; if (xAccum < xTmp) pjSrc += 3; } // Write out the remainder of the scan: CP_WRITE_SRC(pjBase, ulDst); // Now count the number of duplicate scans: cyBreak--; yDuplicate++; cyDuplicate = -1; pjOldScan = pjSrcScan; do { cyDuplicate++; pjSrcScan += yInt; yTmp = yAccum + yFrac; if (yTmp < yAccum) { pjSrcScan += lDeltaSrc; } yAccum = yTmp; yCount--; } while ((yCount != 0) && (pjSrcScan == pjOldScan)); // Duplicate the scan 'cyDuplicate' times with one blt: if (cyDuplicate != 0) { lDuplicate = (yDuplicate * cxMemory) + xDuplicate; yDuplicate += cyDuplicate; CHECK_FIFO_SPACE(pjBase, 8); CP_WRITE(pjBase, DWG_AR3, lDuplicate); CP_WRITE(pjBase, DWG_AR0, lDuplicate + WidthX - 1); cyBreak -= cyDuplicate; if (cyBreak >= 0) { // We haven't crossed a WRAM boundary, so we can use a // WRAM-WRAM blt to duplicate the scan: CP_WRITE(pjBase, DWG_DWGCTL, opcode_FBITBLT | atype_RPL | blockm_OFF | bltmod_BFCOL | pattern_OFF | transc_BG_OPAQUE | bop_NOP | shftzero_ZERO | sgnzero_ZERO); CP_WRITE(pjBase, DWG_FXRIGHT, xDstRightFast); CP_START(pjBase, DWG_LEN, cyDuplicate); CP_WRITE(pjBase, DWG_FXRIGHT, xDstRight); } else { CP_WRITE(pjBase, DWG_DWGCTL, opcode_BITBLT | atype_RPL | blockm_OFF | bltmod_BFCOL | pattern_OFF | transc_BG_OPAQUE | bop_SRCCOPY | shftzero_ZERO | sgnzero_ZERO); CP_START(pjBase, DWG_LEN, cyDuplicate); iBreak++; if (iBreak >= ppdev->cyBreak) { // That was the last break we have to worry about: cyBreak = 0xffff; } else { cyBreak += ppdev->ayBreak[iBreak] - ppdev->ayBreak[iBreak - 1]; } } CP_WRITE(pjBase, DWG_AR3, 0); CP_WRITE(pjBase, DWG_AR0, WidthX - 1); } } while (yCount != 0); CHECK_FIFO_SPACE(pjBase, 1); CP_WRITE(pjBase, DWG_CXRIGHT, ppdev->cxMemory - 1); } /******************************Public*Routine******************************\ * VOID vMilDirectStretch32 * * Hardware assisted stretchblt at 32bpp, for Millenniums. * \**************************************************************************/ VOID vMilDirectStretch32( STR_BLT* pStrBlt) { BYTE* pjOldScan; ULONG* pulSrc; ULONG* pulDstEnd; ULONG ulDst; ULONG xAccum; ULONG xTmp; ULONG yTmp; LONG cyDuplicate; LONG xDuplicate; LONG yDuplicate; LONG lDuplicate; ULONG yInt; PDEV* ppdev = pStrBlt->ppdev; BYTE* pjBase = ppdev->pjBase; LONG cxMemory = ppdev->cxMemory; LONG xDst = pStrBlt->XDstStart; LONG xSrc = pStrBlt->XSrcStart; BYTE* pjSrcScan = pStrBlt->pjSrcScan + xSrc * 4; ULONG* pulDst = (ULONG*)(pStrBlt->pjDstScan) + xDst; LONG yDst = pStrBlt->YDstStart; LONG yCount = pStrBlt->YDstCount; LONG WidthX = pStrBlt->XDstEnd - xDst; ULONG xInt = pStrBlt->ulXDstToSrcIntCeil; ULONG xFrac = pStrBlt->ulXDstToSrcFracCeil; ULONG yAccum = pStrBlt->ulYFracAccumulator; ULONG yFrac = pStrBlt->ulYDstToSrcFracCeil; LONG lDstStride = pStrBlt->lDeltaDst - 4*WidthX; LONG lDeltaDst = pStrBlt->lDeltaDst; LONG lDeltaSrc = pStrBlt->lDeltaSrc; yInt = pStrBlt->lDeltaSrc * pStrBlt->ulYDstToSrcIntCeil; yDuplicate = yDst + ppdev->yOffset - 1; xDuplicate = xDst + ppdev->xOffset + ppdev->ulYDstOrg; START_DIRECT_ACCESS_STORM(ppdev, pjBase); do { ULONG ulSrc; ULONG yTmp; pulSrc = (ULONG*) pjSrcScan; xAccum = pStrBlt->ulXFracAccumulator; pulDstEnd = pulDst + WidthX; while (pulDst != pulDstEnd) { ulSrc = *pulSrc; xTmp = xAccum + xFrac; pulSrc = pulSrc + xInt; if (xTmp < xAccum) pulSrc++; *(ULONG*)pulDst = ulSrc; pulDst++; xAccum = xTmp; } // Now count the number of duplicate scans: pulDst = (ULONG*) ((BYTE*) pulDst + lDstStride); yDuplicate++; cyDuplicate = -1; pjOldScan = pjSrcScan; do { cyDuplicate++; pjSrcScan += yInt; yTmp = yAccum + yFrac; if (yTmp < yAccum) { pjSrcScan += lDeltaSrc; } yAccum = yTmp; yCount--; } while ((yCount != 0) && (pjSrcScan == pjOldScan)); // Duplicate the scan 'cyDuplicate' times with one blt: if (cyDuplicate != 0) { lDuplicate = yDuplicate * cxMemory + xDuplicate; CHECK_FIFO_SPACE(pjBase, 4); CP_WRITE(pjBase, DWG_AR3, lDuplicate); CP_WRITE(pjBase, DWG_AR0, lDuplicate + WidthX - 1); CP_WRITE(pjBase, DWG_DWGCTL, opcode_BITBLT | atype_RPL | blockm_OFF | bltmod_BFCOL | pattern_OFF | transc_BG_OPAQUE | bop_SRCCOPY | shftzero_ZERO | sgnzero_ZERO); CP_START(pjBase, DWG_YDSTLEN, ((yDuplicate + 1) << yval_SHIFT) | cyDuplicate); yDuplicate += cyDuplicate; pulDst = (ULONG*) ((BYTE*) pulDst + cyDuplicate * lDeltaDst); } } while (yCount != 0); END_DIRECT_ACCESS_STORM(ppdev, pjBase); } /******************************Public*Routine******************************\ * * Routine Description: * * StretchBlt using integer math. Must be from one surface to another * surface of the same format. * * Arguments: * * ppdev - PDEV for device * prclDst - Pointer to rectangle of Dst extents * pvSrc - Pointer to start of Src bitmap * lDeltaSrc - Bytes from start of Src scan line to start of next * prclSrc - Pointer to rectangle of Src extents * prclClip - Clip Dest to this rect * * Return Value: * * Status * \**************************************************************************/ VOID vStretchDIB( PDEV* ppdev, RECTL* prclDst, VOID* pvSrc, LONG lDeltaSrc, RECTL* prclSrc, RECTL* prclClip) { STR_BLT StrBlt; ULONG XSrcToDstIntFloor; ULONG XSrcToDstFracFloor; ULONG ulXDstToSrcIntCeil; ULONG ulXDstToSrcFracCeil; ULONG YSrcToDstIntFloor; ULONG YSrcToDstFracFloor; ULONG ulYDstToSrcIntCeil; ULONG ulYDstToSrcFracCeil; LONG SrcIntScan; LONG DstDeltaScanEnd; ULONG ulXFracAccumulator; ULONG ulYFracAccumulator; LONG LeftClipDistance; LONG TopClipDistance; BOOL bStretch; union { LARGE_INTEGER large; ULONGLONG li; } liInit; PFN_DIRSTRETCH pfnStr; // // Calculate exclusive start and end points: // LONG WidthDst = prclDst->right - prclDst->left; LONG HeightDst = prclDst->bottom - prclDst->top; LONG WidthSrc = prclSrc->right - prclSrc->left; LONG HeightSrc = prclSrc->bottom - prclSrc->top; LONG XSrcStart = prclSrc->left; LONG XSrcEnd = prclSrc->right; LONG XDstStart = prclDst->left; LONG XDstEnd = prclDst->right; LONG YSrcStart = prclSrc->top; LONG YSrcEnd = prclSrc->bottom; LONG YDstStart = prclDst->top; LONG YDstEnd = prclDst->bottom; // // Validate parameters: // ASSERTDD(pvSrc != (VOID*)NULL, "Bad source bitmap pointer"); ASSERTDD(prclDst != (RECTL*)NULL, "Bad destination rectangle"); ASSERTDD(prclSrc != (RECTL*)NULL, "Bad source rectangle"); ASSERTDD((WidthDst > 0) && (HeightDst > 0) && (WidthSrc > 0) && (HeightSrc > 0), "Can't do mirroring or empty rectangles here"); ASSERTDD((WidthDst <= STRETCH_MAX_EXTENT) && (HeightDst <= STRETCH_MAX_EXTENT) && (WidthSrc <= STRETCH_MAX_EXTENT) && (HeightSrc <= STRETCH_MAX_EXTENT), "Stretch exceeds limits"); ASSERTDD(prclClip != NULL, "Bad clip rectangle"); // // Calculate X Dst to Src mapping // // // dst->src = ( CEIL( (2k*WidthSrc)/WidthDst) ) / 2k // // = ( FLOOR( (2k*WidthSrc -1) / WidthDst) + 1) / 2k // // where 2k = 2 ^ 32 // { ULONGLONG liWidthSrc; ULONGLONG liQuo; ULONG ulTemp; // // Work around a compiler bug dealing with the assignment // 'liHeightSrc = (((LONGLONG)HeightSrc) << 32) - 1': // liInit.large.LowPart = (ULONG) -1; liInit.large.HighPart = WidthSrc - 1; liWidthSrc = liInit.li; liQuo = liWidthSrc / (ULONGLONG) WidthDst; ulXDstToSrcIntCeil = (ULONG)(liQuo >> 32); ulXDstToSrcFracCeil = (ULONG)liQuo; // // Now add 1, use fake carry: // ulTemp = ulXDstToSrcFracCeil + 1; ulXDstToSrcIntCeil += (ulTemp < ulXDstToSrcFracCeil); ulXDstToSrcFracCeil = ulTemp; } // // Calculate Y Dst to Src mapping // // // dst->src = ( CEIL( (2k*HeightSrc)/HeightDst) ) / 2k // // = ( FLOOR( (2k*HeightSrc -1) / HeightDst) + 1) / 2k // // where 2k = 2 ^ 32 // { ULONGLONG liHeightSrc; ULONGLONG liQuo; ULONG ulTemp; // // Work around a compiler bug dealing with the assignment // 'liHeightSrc = (((LONGLONG)HeightSrc) << 32) - 1': // liInit.large.LowPart = (ULONG) -1; liInit.large.HighPart = HeightSrc - 1; liHeightSrc = liInit.li; liQuo = liHeightSrc / (ULONGLONG) HeightDst; ulYDstToSrcIntCeil = (ULONG)(liQuo >> 32); ulYDstToSrcFracCeil = (ULONG)liQuo; // // Now add 1, use fake carry: // ulTemp = ulYDstToSrcFracCeil + 1; ulYDstToSrcIntCeil += (ulTemp < ulYDstToSrcFracCeil); ulYDstToSrcFracCeil = ulTemp; } // // Now clip Dst in X, and/or calc src clipping effect on dst // // adjust left and right edges if needed, record // distance adjusted for fixing the src // if (XDstStart < prclClip->left) { XDstStart = prclClip->left; } if (XDstEnd > prclClip->right) { XDstEnd = prclClip->right; } // // Check for totally clipped out destination: // if (XDstEnd <= XDstStart) { return; } LeftClipDistance = XDstStart - prclDst->left; { ULONG ulTempInt; ULONG ulTempFrac; // // Calculate displacement for .5 in destination and add: // ulTempFrac = (ulXDstToSrcFracCeil >> 1) | (ulXDstToSrcIntCeil << 31); ulTempInt = (ulXDstToSrcIntCeil >> 1); XSrcStart += ulTempInt; ulXFracAccumulator = ulTempFrac; if (LeftClipDistance != 0) { ULONGLONG ullFraction; ULONG ulTmp; ullFraction = UInt32x32To64(ulXDstToSrcFracCeil, LeftClipDistance); ulTmp = ulXFracAccumulator; ulXFracAccumulator += (ULONG) (ullFraction); if (ulXFracAccumulator < ulTmp) XSrcStart++; XSrcStart += (ulXDstToSrcIntCeil * LeftClipDistance) + (ULONG) (ullFraction >> 32); } } // // Now clip Dst in Y, and/or calc src clipping effect on dst // // adjust top and bottom edges if needed, record // distance adjusted for fixing the src // if (YDstStart < prclClip->top) { YDstStart = prclClip->top; } if (YDstEnd > prclClip->bottom) { YDstEnd = prclClip->bottom; } // // Check for totally clipped out destination: // if (YDstEnd <= YDstStart) { return; } TopClipDistance = YDstStart - prclDst->top; { ULONG ulTempInt; ULONG ulTempFrac; // // Calculate displacement for .5 in destination and add: // ulTempFrac = (ulYDstToSrcFracCeil >> 1) | (ulYDstToSrcIntCeil << 31); ulTempInt = ulYDstToSrcIntCeil >> 1; YSrcStart += (LONG)ulTempInt; ulYFracAccumulator = ulTempFrac; if (TopClipDistance != 0) { ULONGLONG ullFraction; ULONG ulTmp; ullFraction = UInt32x32To64(ulYDstToSrcFracCeil, TopClipDistance); ulTmp = ulYFracAccumulator; ulYFracAccumulator += (ULONG) (ullFraction); if (ulYFracAccumulator < ulTmp) YSrcStart++; YSrcStart += (ulYDstToSrcIntCeil * TopClipDistance) + (ULONG) (ullFraction >> 32); } } // // Warm up the hardware if doing an expanding stretch in 'y': // bStretch = (HeightDst > HeightSrc); if (bStretch) { BYTE* pjBase = ppdev->pjBase; LONG xOffset = ppdev->xOffset; CHECK_FIFO_SPACE(pjBase, 6); CP_WRITE(pjBase, DWG_DWGCTL, opcode_BITBLT + atype_RPL + blockm_OFF + bltmod_BFCOL + pattern_OFF + transc_BG_OPAQUE + bop_SRCCOPY); CP_WRITE(pjBase, DWG_SHIFT, 0); CP_WRITE(pjBase, DWG_SGN, 0); CP_WRITE(pjBase, DWG_AR5, ppdev->cxMemory); CP_WRITE(pjBase, DWG_FXLEFT, XDstStart + xOffset); CP_WRITE(pjBase, DWG_FXRIGHT, XDstEnd + xOffset - 1); ppdev->HopeFlags = SIGN_CACHE; } // // Fill out blt structure, then call format-specific stretch code // StrBlt.ppdev = ppdev; StrBlt.XDstEnd = XDstEnd; StrBlt.YDstStart = YDstStart; StrBlt.YDstCount = YDstEnd - YDstStart; if (StrBlt.YDstCount > 0) { // // Caclulate starting scan line address. Since the inner loop // routines are format dependent, they must add XDstStart/XSrcStart // to pjDstScan/pjSrcScan to get the actual starting pixel address. // StrBlt.pjSrcScan = (BYTE*) pvSrc + (YSrcStart * lDeltaSrc); StrBlt.pjDstScan = ppdev->pjScreen + (ppdev->yOffset + YDstStart) * ppdev->lDelta + (ppdev->xOffset + ppdev->ulYDstOrg) * ppdev->cjPelSize; StrBlt.lDeltaSrc = lDeltaSrc; StrBlt.XSrcStart = XSrcStart; StrBlt.XDstStart = XDstStart; StrBlt.lDeltaDst = ppdev->lDelta; StrBlt.ulXDstToSrcIntCeil = ulXDstToSrcIntCeil; StrBlt.ulXDstToSrcFracCeil = ulXDstToSrcFracCeil; StrBlt.ulYDstToSrcIntCeil = ulYDstToSrcIntCeil; StrBlt.ulYDstToSrcFracCeil = ulYDstToSrcFracCeil; StrBlt.ulXFracAccumulator = ulXFracAccumulator; StrBlt.ulYFracAccumulator = ulYFracAccumulator; if (ppdev->ulBoardId == MGA_STORM) { if (ppdev->iBitmapFormat == BMF_8BPP) { if ((XDstEnd - XDstStart) < 7) pfnStr = vMilDirectStretch8Narrow; else pfnStr = vMilDirectStretch8; } else if (ppdev->iBitmapFormat == BMF_16BPP) { pfnStr = vMilDirectStretch16; } else if (ppdev->iBitmapFormat == BMF_24BPP) { pfnStr = vMilDirectStretch24; } else { ASSERTDD(ppdev->iBitmapFormat == BMF_32BPP, "Expected 32bpp"); pfnStr = vMilDirectStretch32; } } else { #if defined(_X86_) { if (ppdev->iBitmapFormat == BMF_8BPP) { if ((XDstEnd - XDstStart) < 7) pfnStr = vMgaDirectStretch8Narrow; else pfnStr = vMgaDirectStretch8; } else { ASSERTDD(ppdev->iBitmapFormat == BMF_16BPP, "Expected 16bpp"); pfnStr = vMgaDirectStretch16; } } #endif } (*pfnStr)(&StrBlt); } } /******************************Public*Routine******************************\ * BOOL DrvStretchBlt * \**************************************************************************/ BOOL DrvStretchBlt( SURFOBJ* psoDst, SURFOBJ* psoSrc, SURFOBJ* psoMsk, CLIPOBJ* pco, XLATEOBJ* pxlo, COLORADJUSTMENT* pca, POINTL* pptlHTOrg, RECTL* prclDst, RECTL* prclSrc, POINTL* pptlMsk, ULONG iMode) { DSURF* pdsurfSrc; DSURF* pdsurfDst; PDEV* ppdev; OH* poh; SURFOBJ* psoDstNew; SURFOBJ* psoSrcNew; BOOL bPunt = FALSE; BOOL bRet; // GDI guarantees us that for a StretchBlt the destination surface // will always be a device surface, and not a DIB: ppdev = (PDEV*) psoDst->dhpdev; pdsurfDst = (DSURF*) psoDst->dhsurf; pdsurfSrc = (DSURF*) psoSrc->dhsurf; poh = pdsurfDst->poh; ppdev->xOffset = poh->x; ppdev->yOffset = poh->y; // It's quicker for GDI to do a StretchBlt when the source surface // is not a device-managed surface, because then it can directly // read the source bits without having to allocate a temporary // buffer and call DrvCopyBits to get a copy that it can use. if (ppdev->ulBoardId == MGA_STORM) { psoDstNew = psoDst; if (psoSrc->iType != STYPE_BITMAP) { pdsurfSrc = (DSURF*) psoSrc->dhsurf; if (pdsurfSrc->dt == DT_SCREEN) { // The source is a device bitmap that is currently stored // in device memory. psoSrcNew = psoSrc; bPunt = TRUE; } else { ASSERTDD(pdsurfSrc->dt == DT_DIB, "Can only handle DIB DFBs here"); // The source was a device bitmap that we just converted // to a DIB: psoSrcNew = pdsurfSrc->pso; } } else { psoSrcNew = psoSrc; } } else { psoDstNew = psoDst; if (psoSrc->iType == STYPE_DEVBITMAP) { pdsurfSrc = (DSURF*) psoSrc->dhsurf; if (pdsurfSrc->dt == DT_SCREEN) { // The source is a device bitmap that is currently stored // in device memory. psoSrcNew = psoSrc; bPunt = TRUE; } else { ASSERTDD(pdsurfSrc->dt == DT_DIB, "Can only handle DIB DFBs here"); // The source was a device bitmap that we just converted // to a DIB: psoSrcNew = pdsurfSrc->pso; } } else { psoSrcNew = psoSrc; } // With the old MGA's, we have fast stretch code only for 8bpp and // 16bpp: if (ppdev->iBitmapFormat > BMF_16BPP) { bPunt = TRUE; } // With the old MGA's, we have fast stretch code only for x86: #if !defined(_X86_) { bPunt = TRUE; } #endif } if (pdsurfDst->dt == DT_DIB) { // The destination was a device bitmap that we just converted // to a DIB: psoDstNew = pdsurfDst->pso; bPunt = TRUE; } if (!bPunt) { RECTL rclClip; RECTL* prclClip; ULONG cxDst; ULONG cyDst; ULONG cxSrc; ULONG cySrc; BOOL bMore; CLIPENUM ce; LONG c; LONG i; if ((psoSrcNew->iType == STYPE_BITMAP) && (psoMsk == NULL) && ((pxlo == NULL) || (pxlo->flXlate & XO_TRIVIAL)) && ((psoSrcNew->iBitmapFormat == ppdev->iBitmapFormat))) { cxDst = prclDst->right - prclDst->left; cyDst = prclDst->bottom - prclDst->top; cxSrc = prclSrc->right - prclSrc->left; cySrc = prclSrc->bottom - prclSrc->top; // Our 'vStretchDIB' routine requires that the stretch be // non-inverting, within a certain size, to have no source // clipping, and to have no empty rectangles (the latter is the // reason for the '- 1' on the unsigned compare here): if (((cxSrc - 1) < STRETCH_MAX_EXTENT) && ((cySrc - 1) < STRETCH_MAX_EXTENT) && ((cxDst - 1) < STRETCH_MAX_EXTENT) && ((cyDst - 1) < STRETCH_MAX_EXTENT) && (prclSrc->left >= 0) && (prclSrc->top >= 0) && (prclSrc->right <= psoSrcNew->sizlBitmap.cx) && (prclSrc->bottom <= psoSrcNew->sizlBitmap.cy)) { // Our snazzy routine only does COLORONCOLOR. But for // stretching blts, BLACKONWHITE and WHITEONBLACK are also // equivalent to COLORONCOLOR: if ((iMode == COLORONCOLOR) || ((iMode < COLORONCOLOR) && (cxSrc <= cxDst) && (cySrc <= cyDst))) { if ((pco == NULL) || (pco->iDComplexity == DC_TRIVIAL)) { rclClip.left = LONG_MIN; rclClip.top = LONG_MIN; rclClip.right = LONG_MAX; rclClip.bottom = LONG_MAX; prclClip = &rclClip; StretchSingleClipRect: vStretchDIB(ppdev, prclDst, psoSrcNew->pvScan0, psoSrcNew->lDelta, prclSrc, prclClip); return(TRUE); } else if (pco->iDComplexity == DC_RECT) { prclClip = &pco->rclBounds; goto StretchSingleClipRect; } else { CLIPOBJ_cEnumStart(pco, FALSE, CT_RECTANGLES, CD_ANY, 0); do { bMore = CLIPOBJ_bEnum(pco, sizeof(ce), (ULONG*) &ce); c = cIntersect(prclDst, ce.arcl, ce.c); if (c != 0) { for (i = 0; i < c; i++) { vStretchDIB(ppdev, prclDst, psoSrcNew->pvScan0, psoSrcNew->lDelta, prclSrc, &ce.arcl[i]); } } } while (bMore); return(TRUE); } } } } } // GDI is nice enough to handle the cases where 'psoDst' and/or 'psoSrc' // are device-managed surfaces, but it ain't gonna be fast... if (ppdev->ulBoardId == MGA_STORM) { START_DIRECT_ACCESS_STORM(ppdev, ppdev->pjBase); } else { START_DIRECT_ACCESS_MGA(ppdev, ppdev->pjBase); } bRet = EngStretchBlt(psoDstNew, psoSrcNew, psoMsk, pco, pxlo, pca, pptlHTOrg, prclDst, prclSrc, pptlMsk, iMode); if (ppdev->ulBoardId == MGA_STORM) { END_DIRECT_ACCESS_STORM(ppdev, ppdev->pjBase); } else { END_DIRECT_ACCESS_MGA(ppdev, ppdev->pjBase); } return(bRet); }