src/core/SkBlitRow_D32.cpp

/*
 * Copyright 2011 Google Inc.
 *
 * Use of this source code is governed by a BSD-style license that can be
 * found in the LICENSE file.
 */

#include "SkBlitRow.h"
#include "SkBlitMask.h"
#include "SkColorPriv.h"
#include "SkUtils.h"

#define UNROLL

static void S32_Opaque_BlitRow32(SkPMColor* SK_RESTRICT dst,
                                 const SkPMColor* SK_RESTRICT src,
                                 int count, U8CPU alpha) {
    SkASSERT(255 == alpha);
    memcpy(dst, src, count * 4);
}

static void S32_Blend_BlitRow32(SkPMColor* SK_RESTRICT dst,
                                const SkPMColor* SK_RESTRICT src,
                                int count, U8CPU alpha) {
    SkASSERT(alpha <= 255);
    if (count > 0) {
        unsigned src_scale = SkAlpha255To256(alpha);
        unsigned dst_scale = 256 - src_scale;

#ifdef UNROLL
        if (count & 1) {
            *dst = SkAlphaMulQ(*(src++), src_scale) + SkAlphaMulQ(*dst, dst_scale);
            dst += 1;
            count -= 1;
        }

        const SkPMColor* SK_RESTRICT srcEnd = src + count;
        while (src != srcEnd) {
            *dst = SkAlphaMulQ(*(src++), src_scale) + SkAlphaMulQ(*dst, dst_scale);
            dst += 1;
            *dst = SkAlphaMulQ(*(src++), src_scale) + SkAlphaMulQ(*dst, dst_scale);
            dst += 1;
        }
#else
        do {
            *dst = SkAlphaMulQ(*src, src_scale) + SkAlphaMulQ(*dst, dst_scale);
            src += 1;
            dst += 1;
        } while (--count > 0);
#endif
    }
}

static void S32A_Opaque_BlitRow32(SkPMColor* SK_RESTRICT dst,
                                  const SkPMColor* SK_RESTRICT src,
                                  int count, U8CPU alpha) {
    SkASSERT(255 == alpha);
    if (count > 0) {
#ifdef UNROLL
        if (count & 1) {
            *dst = SkPMSrcOver(*(src++), *dst);
            dst += 1;
            count -= 1;
        }

        const SkPMColor* SK_RESTRICT srcEnd = src + count;
        while (src != srcEnd) {
            *dst = SkPMSrcOver(*(src++), *dst);
            dst += 1;
            *dst = SkPMSrcOver(*(src++), *dst);
            dst += 1;
        }
#else
        do {
            *dst = SkPMSrcOver(*src, *dst);
            src += 1;
            dst += 1;
        } while (--count > 0);
#endif
    }
}

static void S32A_Blend_BlitRow32(SkPMColor* SK_RESTRICT dst,
                                 const SkPMColor* SK_RESTRICT src,
                                 int count, U8CPU alpha) {
    SkASSERT(alpha <= 255);
    if (count > 0) {
#ifdef UNROLL
        if (count & 1) {
            *dst = SkBlendARGB32(*(src++), *dst, alpha);
            dst += 1;
            count -= 1;
        }

        const SkPMColor* SK_RESTRICT srcEnd = src + count;
        while (src != srcEnd) {
            *dst = SkBlendARGB32(*(src++), *dst, alpha);
            dst += 1;
            *dst = SkBlendARGB32(*(src++), *dst, alpha);
            dst += 1;
        }
#else
        do {
            *dst = SkBlendARGB32(*src, *dst, alpha);
            src += 1;
            dst += 1;
        } while (--count > 0);
#endif
    }
}

///////////////////////////////////////////////////////////////////////////////

static const SkBlitRow::Proc32 gDefault_Procs32[] = {
    S32_Opaque_BlitRow32,
    S32_Blend_BlitRow32,
    S32A_Opaque_BlitRow32,
    S32A_Blend_BlitRow32
};

SkBlitRow::Proc32 SkBlitRow::Factory32(unsigned flags) {
    SkASSERT(flags < SK_ARRAY_COUNT(gDefault_Procs32));
    // just so we don't crash
    flags &= kFlags32_Mask;

    SkBlitRow::Proc32 proc = PlatformProcs32(flags);
    if (nullptr == proc) {
        proc = gDefault_Procs32[flags];
    }
    SkASSERT(proc);
    return proc;
}

#include "Sk4px.h"

// Color32 uses the blend_256_round_alt algorithm from tests/BlendTest.cpp.
// It's not quite perfect, but it's never wrong in the interesting edge cases,
// and it's quite a bit faster than blend_perfect.
//
// blend_256_round_alt is our currently blessed algorithm.  Please use it or an analogous one.
void SkBlitRow::Color32(SkPMColor dst[], const SkPMColor src[], int count, SkPMColor color) {
    switch (SkGetPackedA32(color)) {
        case   0: memmove(dst, src, count * sizeof(SkPMColor)); return;
        case 255: sk_memset32(dst, color, count);               return;
    }

    // This Sk4px impl works great on other platforms or when we have NEON.
#if defined(SK_CPU_ARM32) && !defined(SK_ARM_HAS_NEON)
    if (auto proc = PlatformColor32Proc()) { return proc(dst, src, count, color); }
#endif

    unsigned invA = 255 - SkGetPackedA32(color);
    invA += invA >> 7;
    SkASSERT(invA < 256);  // We've already handled alpha == 0 above.

    Sk16h colorHighAndRound = Sk4px::DupPMColor(color).widenHi() + Sk16h(128);
    Sk16b invA_16x(invA);

    Sk4px::MapSrc(count, dst, src, [&](const Sk4px& src4) -> Sk4px {
        return (src4 * invA_16x).addNarrowHi(colorHighAndRound);
    });
}