/* * Copyright 2006 The Android Open Source Project * * Use of this source code is governed by a BSD-style license that can be * found in the LICENSE file. */ #include "SkScanPriv.h" #include "SkPath.h" #include "SkMatrix.h" #include "SkBlitter.h" #include "SkRegion.h" #include "SkAntiRun.h" #define SHIFT 2 #define SCALE (1 << SHIFT) #define MASK (SCALE - 1) /** @file We have two techniques for capturing the output of the supersampler: - SUPERMASK, which records a large mask-bitmap this is often faster for small, complex objects - RLE, which records a rle-encoded scanline this is often faster for large objects with big spans These blitters use two coordinate systems: - destination coordinates, scale equal to the output - often abbreviated with 'i' or 'I' in variable names - supersampled coordinates, scale equal to the output * SCALE */ //#define FORCE_SUPERMASK //#define FORCE_RLE /////////////////////////////////////////////////////////////////////////////// /// Base class for a single-pass supersampled blitter. class BaseSuperBlitter : public SkBlitter { public: BaseSuperBlitter(SkBlitter* realBlitter, const SkIRect& ir, const SkRegion& clip, bool isInverse); /// Must be explicitly defined on subclasses. virtual void blitAntiH(int x, int y, const SkAlpha antialias[], const int16_t runs[]) override { SkDEBUGFAIL("How did I get here?"); } /// May not be called on BaseSuperBlitter because it blits out of order. void blitV(int x, int y, int height, SkAlpha alpha) override { SkDEBUGFAIL("How did I get here?"); } protected: SkBlitter* fRealBlitter; /// Current y coordinate, in destination coordinates. int fCurrIY; /// Widest row of region to be blitted, in destination coordinates. int fWidth; /// Leftmost x coordinate in any row, in destination coordinates. int fLeft; /// Leftmost x coordinate in any row, in supersampled coordinates. int fSuperLeft; SkDEBUGCODE(int fCurrX;) /// Current y coordinate in supersampled coordinates. int fCurrY; /// Initial y coordinate (top of bounds). int fTop; SkIRect fSectBounds; }; BaseSuperBlitter::BaseSuperBlitter(SkBlitter* realBlit, const SkIRect& ir, const SkRegion& clip, bool isInverse) { fRealBlitter = realBlit; SkIRect sectBounds; if (isInverse) { // We use the clip bounds instead of the ir, since we may be asked to //draw outside of the rect when we're a inverse filltype sectBounds = clip.getBounds(); } else { if (!sectBounds.intersect(ir, clip.getBounds())) { sectBounds.setEmpty(); } } const int left = sectBounds.left(); const int right = sectBounds.right(); fLeft = left; fSuperLeft = left << SHIFT; fWidth = right - left; fTop = sectBounds.top(); fCurrIY = fTop - 1; fCurrY = (fTop << SHIFT) - 1; SkDEBUGCODE(fCurrX = -1;) } /// Run-length-encoded supersampling antialiased blitter. class SuperBlitter : public BaseSuperBlitter { public: SuperBlitter(SkBlitter* realBlitter, const SkIRect& ir, const SkRegion& clip, bool isInverse); virtual ~SuperBlitter() { this->flush(); } /// Once fRuns contains a complete supersampled row, flush() blits /// it out through the wrapped blitter. void flush(); /// Blits a row of pixels, with location and width specified /// in supersampled coordinates. void blitH(int x, int y, int width) override; /// Blits a rectangle of pixels, with location and size specified /// in supersampled coordinates. void blitRect(int x, int y, int width, int height) override; private: // The next three variables are used to track a circular buffer that // contains the values used in SkAlphaRuns. These variables should only // ever be updated in advanceRuns(), and fRuns should always point to // a valid SkAlphaRuns... int fRunsToBuffer; void* fRunsBuffer; int fCurrentRun; SkAlphaRuns fRuns; // extra one to store the zero at the end int getRunsSz() const { return (fWidth + 1 + (fWidth + 2)/2) * sizeof(int16_t); } // This function updates the fRuns variable to point to the next buffer space // with adequate storage for a SkAlphaRuns. It mostly just advances fCurrentRun // and resets fRuns to point to an empty scanline. void advanceRuns() { const size_t kRunsSz = this->getRunsSz(); fCurrentRun = (fCurrentRun + 1) % fRunsToBuffer; fRuns.fRuns = reinterpret_cast<int16_t*>( reinterpret_cast<uint8_t*>(fRunsBuffer) + fCurrentRun * kRunsSz); fRuns.fAlpha = reinterpret_cast<SkAlpha*>(fRuns.fRuns + fWidth + 1); fRuns.reset(fWidth); } int fOffsetX; }; SuperBlitter::SuperBlitter(SkBlitter* realBlitter, const SkIRect& ir, const SkRegion& clip, bool isInverse) : BaseSuperBlitter(realBlitter, ir, clip, isInverse) { fRunsToBuffer = realBlitter->requestRowsPreserved(); fRunsBuffer = realBlitter->allocBlitMemory(fRunsToBuffer * this->getRunsSz()); fCurrentRun = -1; this->advanceRuns(); fOffsetX = 0; } void SuperBlitter::flush() { if (fCurrIY >= fTop) { SkASSERT(fCurrentRun < fRunsToBuffer); if (!fRuns.empty()) { // SkDEBUGCODE(fRuns.dump();) fRealBlitter->blitAntiH(fLeft, fCurrIY, fRuns.fAlpha, fRuns.fRuns); this->advanceRuns(); fOffsetX = 0; } fCurrIY = fTop - 1; SkDEBUGCODE(fCurrX = -1;) } } /** coverage_to_partial_alpha() is being used by SkAlphaRuns, which *accumulates* SCALE pixels worth of "alpha" in [0,(256/SCALE)] to produce a final value in [0, 255] and handles clamping 256->255 itself, with the same (alpha - (alpha >> 8)) correction as coverage_to_exact_alpha(). */ static inline int coverage_to_partial_alpha(int aa) { aa <<= 8 - 2*SHIFT; return aa; } /** coverage_to_exact_alpha() is being used by our blitter, which wants a final value in [0, 255]. */ static inline int coverage_to_exact_alpha(int aa) { int alpha = (256 >> SHIFT) * aa; // clamp 256->255 return alpha - (alpha >> 8); } void SuperBlitter::blitH(int x, int y, int width) { SkASSERT(width > 0); int iy = y >> SHIFT; SkASSERT(iy >= fCurrIY); x -= fSuperLeft; // hack, until I figure out why my cubics (I think) go beyond the bounds if (x < 0) { width += x; x = 0; } #ifdef SK_DEBUG SkASSERT(y != fCurrY || x >= fCurrX); #endif SkASSERT(y >= fCurrY); if (fCurrY != y) { fOffsetX = 0; fCurrY = y; } if (iy != fCurrIY) { // new scanline this->flush(); fCurrIY = iy; } int start = x; int stop = x + width; SkASSERT(start >= 0 && stop > start); // integer-pixel-aligned ends of blit, rounded out int fb = start & MASK; int fe = stop & MASK; int n = (stop >> SHIFT) - (start >> SHIFT) - 1; if (n < 0) { fb = fe - fb; n = 0; fe = 0; } else { if (fb == 0) { n += 1; } else { fb = SCALE - fb; } } fOffsetX = fRuns.add(x >> SHIFT, coverage_to_partial_alpha(fb), n, coverage_to_partial_alpha(fe), (1 << (8 - SHIFT)) - (((y & MASK) + 1) >> SHIFT), fOffsetX); #ifdef SK_DEBUG fRuns.assertValid(y & MASK, (1 << (8 - SHIFT))); fCurrX = x + width; #endif } #if 0 // UNUSED static void set_left_rite_runs(SkAlphaRuns& runs, int ileft, U8CPU leftA, int n, U8CPU riteA) { SkASSERT(leftA <= 0xFF); SkASSERT(riteA <= 0xFF); int16_t* run = runs.fRuns; uint8_t* aa = runs.fAlpha; if (ileft > 0) { run[0] = ileft; aa[0] = 0; run += ileft; aa += ileft; } SkASSERT(leftA < 0xFF); if (leftA > 0) { *run++ = 1; *aa++ = leftA; } if (n > 0) { run[0] = n; aa[0] = 0xFF; run += n; aa += n; } SkASSERT(riteA < 0xFF); if (riteA > 0) { *run++ = 1; *aa++ = riteA; } run[0] = 0; } #endif void SuperBlitter::blitRect(int x, int y, int width, int height) { SkASSERT(width > 0); SkASSERT(height > 0); // blit leading rows while ((y & MASK)) { this->blitH(x, y++, width); if (--height <= 0) { return; } } SkASSERT(height > 0); // Since this is a rect, instead of blitting supersampled rows one at a // time and then resolving to the destination canvas, we can blit // directly to the destintion canvas one row per SCALE supersampled rows. int start_y = y >> SHIFT; int stop_y = (y + height) >> SHIFT; int count = stop_y - start_y; if (count > 0) { y += count << SHIFT; height -= count << SHIFT; // save original X for our tail blitH() loop at the bottom int origX = x; x -= fSuperLeft; // hack, until I figure out why my cubics (I think) go beyond the bounds if (x < 0) { width += x; x = 0; } // There is always a left column, a middle, and a right column. // ileft is the destination x of the first pixel of the entire rect. // xleft is (SCALE - # of covered supersampled pixels) in that // destination pixel. int ileft = x >> SHIFT; int xleft = x & MASK; // irite is the destination x of the last pixel of the OPAQUE section. // xrite is the number of supersampled pixels extending beyond irite; // xrite/SCALE should give us alpha. int irite = (x + width) >> SHIFT; int xrite = (x + width) & MASK; if (!xrite) { xrite = SCALE; irite--; } // Need to call flush() to clean up pending draws before we // even consider blitV(), since otherwise it can look nonmonotonic. SkASSERT(start_y > fCurrIY); this->flush(); int n = irite - ileft - 1; if (n < 0) { // If n < 0, we'll only have a single partially-transparent column // of pixels to render. xleft = xrite - xleft; SkASSERT(xleft <= SCALE); SkASSERT(xleft > 0); xrite = 0; fRealBlitter->blitV(ileft + fLeft, start_y, count, coverage_to_exact_alpha(xleft)); } else { // With n = 0, we have two possibly-transparent columns of pixels // to render; with n > 0, we have opaque columns between them. xleft = SCALE - xleft; // Using coverage_to_exact_alpha is not consistent with blitH() const int coverageL = coverage_to_exact_alpha(xleft); const int coverageR = coverage_to_exact_alpha(xrite); SkASSERT(coverageL > 0 || n > 0 || coverageR > 0); SkASSERT((coverageL != 0) + n + (coverageR != 0) <= fWidth); fRealBlitter->blitAntiRect(ileft + fLeft, start_y, n, count, coverageL, coverageR); } // preamble for our next call to blitH() fCurrIY = stop_y - 1; fOffsetX = 0; fCurrY = y - 1; fRuns.reset(fWidth); x = origX; } // catch any remaining few rows SkASSERT(height <= MASK); while (--height >= 0) { this->blitH(x, y++, width); } } /////////////////////////////////////////////////////////////////////////////// /// Masked supersampling antialiased blitter. class MaskSuperBlitter : public BaseSuperBlitter { public: MaskSuperBlitter(SkBlitter* realBlitter, const SkIRect& ir, const SkRegion&, bool isInverse); virtual ~MaskSuperBlitter() { fRealBlitter->blitMask(fMask, fClipRect); } void blitH(int x, int y, int width) override; static bool CanHandleRect(const SkIRect& bounds) { #ifdef FORCE_RLE return false; #endif int width = bounds.width(); int64_t rb = SkAlign4(width); // use 64bits to detect overflow int64_t storage = rb * bounds.height(); return (width <= MaskSuperBlitter::kMAX_WIDTH) && (storage <= MaskSuperBlitter::kMAX_STORAGE); } private: enum { #ifdef FORCE_SUPERMASK kMAX_WIDTH = 2048, kMAX_STORAGE = 1024 * 1024 * 2 #else kMAX_WIDTH = 32, // so we don't try to do very wide things, where the RLE blitter would be faster kMAX_STORAGE = 1024 #endif }; SkMask fMask; SkIRect fClipRect; // we add 1 because add_aa_span can write (unchanged) 1 extra byte at the end, rather than // perform a test to see if stopAlpha != 0 uint32_t fStorage[(kMAX_STORAGE >> 2) + 1]; }; MaskSuperBlitter::MaskSuperBlitter(SkBlitter* realBlitter, const SkIRect& ir, const SkRegion& clip, bool isInverse) : BaseSuperBlitter(realBlitter, ir, clip, isInverse) { SkASSERT(CanHandleRect(ir)); SkASSERT(!isInverse); fMask.fImage = (uint8_t*)fStorage; fMask.fBounds = ir; fMask.fRowBytes = ir.width(); fMask.fFormat = SkMask::kA8_Format; fClipRect = ir; if (!fClipRect.intersect(clip.getBounds())) { SkASSERT(0); fClipRect.setEmpty(); } // For valgrind, write 1 extra byte at the end so we don't read // uninitialized memory. See comment in add_aa_span and fStorage[]. memset(fStorage, 0, fMask.fBounds.height() * fMask.fRowBytes + 1); } static void add_aa_span(uint8_t* alpha, U8CPU startAlpha) { /* I should be able to just add alpha[x] + startAlpha. However, if the trailing edge of the previous span and the leading edge of the current span round to the same super-sampled x value, I might overflow to 256 with this add, hence the funny subtract. */ unsigned tmp = *alpha + startAlpha; SkASSERT(tmp <= 256); *alpha = SkToU8(tmp - (tmp >> 8)); } static inline uint32_t quadplicate_byte(U8CPU value) { uint32_t pair = (value << 8) | value; return (pair << 16) | pair; } // Perform this tricky subtract, to avoid overflowing to 256. Our caller should // only ever call us with at most enough to hit 256 (never larger), so it is // enough to just subtract the high-bit. Actually clamping with a branch would // be slower (e.g. if (tmp > 255) tmp = 255;) // static inline void saturated_add(uint8_t* ptr, U8CPU add) { unsigned tmp = *ptr + add; SkASSERT(tmp <= 256); *ptr = SkToU8(tmp - (tmp >> 8)); } // minimum count before we want to setup an inner loop, adding 4-at-a-time #define MIN_COUNT_FOR_QUAD_LOOP 16 static void add_aa_span(uint8_t* alpha, U8CPU startAlpha, int middleCount, U8CPU stopAlpha, U8CPU maxValue) { SkASSERT(middleCount >= 0); saturated_add(alpha, startAlpha); alpha += 1; if (middleCount >= MIN_COUNT_FOR_QUAD_LOOP) { // loop until we're quad-byte aligned while (SkTCast<intptr_t>(alpha) & 0x3) { alpha[0] = SkToU8(alpha[0] + maxValue); alpha += 1; middleCount -= 1; } int bigCount = middleCount >> 2; uint32_t* qptr = reinterpret_cast<uint32_t*>(alpha); uint32_t qval = quadplicate_byte(maxValue); do { *qptr++ += qval; } while (--bigCount > 0); middleCount &= 3; alpha = reinterpret_cast<uint8_t*> (qptr); // fall through to the following while-loop } while (--middleCount >= 0) { alpha[0] = SkToU8(alpha[0] + maxValue); alpha += 1; } // potentially this can be off the end of our "legal" alpha values, but that // only happens if stopAlpha is also 0. Rather than test for stopAlpha != 0 // every time (slow), we just do it, and ensure that we've allocated extra space // (see the + 1 comment in fStorage[] saturated_add(alpha, stopAlpha); } void MaskSuperBlitter::blitH(int x, int y, int width) { int iy = (y >> SHIFT); SkASSERT(iy >= fMask.fBounds.fTop && iy < fMask.fBounds.fBottom); iy -= fMask.fBounds.fTop; // make it relative to 0 // This should never happen, but it does. Until the true cause is // discovered, let's skip this span instead of crashing. // See http://crbug.com/17569. if (iy < 0) { return; } #ifdef SK_DEBUG { int ix = x >> SHIFT; SkASSERT(ix >= fMask.fBounds.fLeft && ix < fMask.fBounds.fRight); } #endif x -= (fMask.fBounds.fLeft << SHIFT); // hack, until I figure out why my cubics (I think) go beyond the bounds if (x < 0) { width += x; x = 0; } uint8_t* row = fMask.fImage + iy * fMask.fRowBytes + (x >> SHIFT); int start = x; int stop = x + width; SkASSERT(start >= 0 && stop > start); int fb = start & MASK; int fe = stop & MASK; int n = (stop >> SHIFT) - (start >> SHIFT) - 1; if (n < 0) { SkASSERT(row >= fMask.fImage); SkASSERT(row < fMask.fImage + kMAX_STORAGE + 1); add_aa_span(row, coverage_to_partial_alpha(fe - fb)); } else { fb = SCALE - fb; SkASSERT(row >= fMask.fImage); SkASSERT(row + n + 1 < fMask.fImage + kMAX_STORAGE + 1); add_aa_span(row, coverage_to_partial_alpha(fb), n, coverage_to_partial_alpha(fe), (1 << (8 - SHIFT)) - (((y & MASK) + 1) >> SHIFT)); } #ifdef SK_DEBUG fCurrX = x + width; #endif } /////////////////////////////////////////////////////////////////////////////// static bool fitsInsideLimit(const SkRect& r, SkScalar max) { const SkScalar min = -max; return r.fLeft > min && r.fTop > min && r.fRight < max && r.fBottom < max; } static int overflows_short_shift(int value, int shift) { const int s = 16 + shift; return (value << s >> s) - value; } /** Would any of the coordinates of this rectangle not fit in a short, when left-shifted by shift? */ static int rect_overflows_short_shift(SkIRect rect, int shift) { SkASSERT(!overflows_short_shift(8191, SHIFT)); SkASSERT(overflows_short_shift(8192, SHIFT)); SkASSERT(!overflows_short_shift(32767, 0)); SkASSERT(overflows_short_shift(32768, 0)); // Since we expect these to succeed, we bit-or together // for a tiny extra bit of speed. return overflows_short_shift(rect.fLeft, SHIFT) | overflows_short_shift(rect.fRight, SHIFT) | overflows_short_shift(rect.fTop, SHIFT) | overflows_short_shift(rect.fBottom, SHIFT); } static bool safeRoundOut(const SkRect& src, SkIRect* dst, int32_t maxInt) { const SkScalar maxScalar = SkIntToScalar(maxInt); if (fitsInsideLimit(src, maxScalar)) { src.roundOut(dst); return true; } return false; } void SkScan::AntiFillPath(const SkPath& path, const SkRegion& origClip, SkBlitter* blitter, bool forceRLE) { if (origClip.isEmpty()) { return; } const bool isInverse = path.isInverseFillType(); SkIRect ir; if (!safeRoundOut(path.getBounds(), &ir, SK_MaxS32 >> SHIFT)) { #if 0 const SkRect& r = path.getBounds(); SkDebugf("--- bounds can't fit in SkIRect\n", r.fLeft, r.fTop, r.fRight, r.fBottom); #endif return; } if (ir.isEmpty()) { if (isInverse) { blitter->blitRegion(origClip); } return; } // If the intersection of the path bounds and the clip bounds // will overflow 32767 when << by SHIFT, we can't supersample, // so draw without antialiasing. SkIRect clippedIR; if (isInverse) { // If the path is an inverse fill, it's going to fill the entire // clip, and we care whether the entire clip exceeds our limits. clippedIR = origClip.getBounds(); } else { if (!clippedIR.intersect(ir, origClip.getBounds())) { return; } } if (rect_overflows_short_shift(clippedIR, SHIFT)) { SkScan::FillPath(path, origClip, blitter); return; } // Our antialiasing can't handle a clip larger than 32767, so we restrict // the clip to that limit here. (the runs[] uses int16_t for its index). // // A more general solution (one that could also eliminate the need to // disable aa based on ir bounds (see overflows_short_shift) would be // to tile the clip/target... SkRegion tmpClipStorage; const SkRegion* clipRgn = &origClip; { static const int32_t kMaxClipCoord = 32767; const SkIRect& bounds = origClip.getBounds(); if (bounds.fRight > kMaxClipCoord || bounds.fBottom > kMaxClipCoord) { SkIRect limit = { 0, 0, kMaxClipCoord, kMaxClipCoord }; tmpClipStorage.op(origClip, limit, SkRegion::kIntersect_Op); clipRgn = &tmpClipStorage; } } // for here down, use clipRgn, not origClip SkScanClipper clipper(blitter, clipRgn, ir); const SkIRect* clipRect = clipper.getClipRect(); if (clipper.getBlitter() == NULL) { // clipped out if (isInverse) { blitter->blitRegion(*clipRgn); } return; } // now use the (possibly wrapped) blitter blitter = clipper.getBlitter(); if (isInverse) { sk_blit_above(blitter, ir, *clipRgn); } SkIRect superRect, *superClipRect = NULL; if (clipRect) { superRect.set( clipRect->fLeft << SHIFT, clipRect->fTop << SHIFT, clipRect->fRight << SHIFT, clipRect->fBottom << SHIFT); superClipRect = &superRect; } SkASSERT(SkIntToScalar(ir.fTop) <= path.getBounds().fTop); // MaskSuperBlitter can't handle drawing outside of ir, so we can't use it // if we're an inverse filltype if (!isInverse && MaskSuperBlitter::CanHandleRect(ir) && !forceRLE) { MaskSuperBlitter superBlit(blitter, ir, *clipRgn, isInverse); SkASSERT(SkIntToScalar(ir.fTop) <= path.getBounds().fTop); sk_fill_path(path, superClipRect, &superBlit, ir.fTop, ir.fBottom, SHIFT, *clipRgn); } else { SuperBlitter superBlit(blitter, ir, *clipRgn, isInverse); sk_fill_path(path, superClipRect, &superBlit, ir.fTop, ir.fBottom, SHIFT, *clipRgn); } if (isInverse) { sk_blit_below(blitter, ir, *clipRgn); } } /////////////////////////////////////////////////////////////////////////////// #include "SkRasterClip.h" void SkScan::FillPath(const SkPath& path, const SkRasterClip& clip, SkBlitter* blitter) { if (clip.isEmpty()) { return; } if (clip.isBW()) { FillPath(path, clip.bwRgn(), blitter); } else { SkRegion tmp; SkAAClipBlitter aaBlitter; tmp.setRect(clip.getBounds()); aaBlitter.init(blitter, &clip.aaRgn()); SkScan::FillPath(path, tmp, &aaBlitter); } } void SkScan::AntiFillPath(const SkPath& path, const SkRasterClip& clip, SkBlitter* blitter) { if (clip.isEmpty()) { return; } if (clip.isBW()) { AntiFillPath(path, clip.bwRgn(), blitter); } else { SkRegion tmp; SkAAClipBlitter aaBlitter; tmp.setRect(clip.getBounds()); aaBlitter.init(blitter, &clip.aaRgn()); SkScan::AntiFillPath(path, tmp, &aaBlitter, true); } }