moz-skia/tests/ColorTest.cpp

190 строки
6.5 KiB
C++

/*
* 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 "Test.h"
#include "SkColor.h"
#include "SkColorPriv.h"
#include "SkMathPriv.h"
#include "SkRandom.h"
#include "SkUnPreMultiply.h"
#define GetPackedR16As32(packed) (SkGetPackedR16(dc) << (8 - SK_R16_BITS))
#define GetPackedG16As32(packed) (SkGetPackedG16(dc) << (8 - SK_G16_BITS))
#define GetPackedB16As32(packed) (SkGetPackedB16(dc) << (8 - SK_B16_BITS))
static inline bool S32A_D565_Blend_0(SkPMColor sc, uint16_t dc, U8CPU alpha) {
unsigned dst_scale = 255 - SkMulDiv255Round(SkGetPackedA32(sc), alpha);
unsigned dr = SkMulS16(SkPacked32ToR16(sc), alpha) + SkMulS16(SkGetPackedR16(dc), dst_scale);
unsigned dg = SkMulS16(SkPacked32ToG16(sc), alpha) + SkMulS16(SkGetPackedG16(dc), dst_scale);
unsigned rr = SkDiv255Round(dr);
unsigned rg = SkDiv255Round(dg);
if (rr <= 31 && rg <= 63) {
return true;
}
return false;
}
static inline bool S32A_D565_Blend_01(SkPMColor sc, uint16_t dc, U8CPU alpha) {
unsigned dst_scale = 255 - SkMulDiv255Round(SkGetPackedA32(sc), alpha);
unsigned dr = SkMulS16(SkGetPackedR32(sc), alpha) + SkMulS16(SkGetPackedR16(dc) << 3, dst_scale);
unsigned dg = SkMulS16(SkGetPackedG32(sc), alpha) + SkMulS16(SkGetPackedG16(dc) << 2, dst_scale);
unsigned rr = SkDiv255Round(dr) >> 3;
unsigned rg = SkDiv255Round(dg) >> 2;
if (rr <= 31 && rg <= 63) {
return true;
}
return false;
}
static inline bool S32A_D565_Blend_02(SkPMColor sc, uint16_t dc, U8CPU alpha) {
unsigned dst_scale = 255 - SkMulDiv255Round(SkGetPackedA32(sc), alpha);
unsigned dr = SkMulS16(SkGetPackedR32(sc), alpha) + SkMulS16(GetPackedR16As32(dc), dst_scale);
unsigned dg = SkMulS16(SkGetPackedG32(sc), alpha) + SkMulS16(GetPackedG16As32(dc), dst_scale);
unsigned db = SkMulS16(SkGetPackedB32(sc), alpha) + SkMulS16(GetPackedB16As32(dc), dst_scale);
int rc = SkPack888ToRGB16(SkDiv255Round(dr),
SkDiv255Round(dg),
SkDiv255Round(db));
unsigned rr = SkGetPackedR16(rc);
unsigned rg = SkGetPackedG16(rc);
if (rr <= 31 && rg <= 63) {
return true;
}
return false;
}
static inline bool S32A_D565_Blend_1(SkPMColor sc, uint16_t dc, U8CPU alpha) {
unsigned dst_scale = 255 - SkMulDiv255Round(SkGetPackedA32(sc), alpha);
unsigned dr = (SkMulS16(SkGetPackedR32(sc), alpha) >> 3) + SkMulS16(SkGetPackedR16(dc), dst_scale);
unsigned dg = (SkMulS16(SkGetPackedG32(sc), alpha) >> 2) + SkMulS16(SkGetPackedG16(dc), dst_scale);
unsigned rr = SkDiv255Round(dr);
unsigned rg = SkDiv255Round(dg);
if (rr <= 31 && rg <= 63) {
return true;
}
return false;
}
static inline int SkDiv65025Round(int x) {
return (x + 65025/2) / 65025;
// return x / 65025;
}
static inline bool S32A_D565_Blend_2(SkPMColor sc, uint16_t dc, U8CPU alpha) {
unsigned dst_scale = 255*255 - SkGetPackedA32(sc) * alpha;
alpha *= 255;
unsigned dr = (SkGetPackedR32(sc) >> 3) * alpha + SkGetPackedR16(dc) * dst_scale;
unsigned dg = (SkGetPackedG32(sc) >> 2) * alpha + SkGetPackedG16(dc) * dst_scale;
unsigned rr = SkDiv65025Round(dr);
unsigned rg = SkDiv65025Round(dg);
if (rr <= 31 && rg <= 63) {
return true;
}
return false;
}
static inline void test_565blend(skiatest::Reporter* reporter) {
int total_failures = 0;
for (int global_alpha = 0; global_alpha <= 255; ++global_alpha) {
int failures = 0;
int total = 0;
for (int src_a = 0; src_a <= 255; ++src_a) {
for (int src_c = 0; src_c <= src_a; ++src_c) {
SkPMColor sc = SkPackARGB32(src_a, src_c, src_c, src_c);
for (int dst_r = 0; dst_r <= 31; ++dst_r) {
for (int dst_g = 0; dst_g <= 63; ++dst_g) {
uint16_t dc = SkPackRGB16(dst_r, dst_g, dst_r);
failures += !S32A_D565_Blend_0(sc, dc, global_alpha);
total += 1;
}
}
}
}
SkDebugf("global_alpha=%d failures=%d total=%d %g\n", global_alpha, failures, total, failures * 100.0 / total);
total_failures += failures;
}
SkDebugf("total failures %d\n", total_failures);
}
static inline void test_premul(skiatest::Reporter* reporter) {
for (int a = 0; a <= 255; a++) {
for (int x = 0; x <= 255; x++) {
SkColor c0 = SkColorSetARGB(a, x, x, x);
SkPMColor p0 = SkPreMultiplyColor(c0);
SkColor c1 = SkUnPreMultiply::PMColorToColor(p0);
SkPMColor p1 = SkPreMultiplyColor(c1);
// we can't promise that c0 == c1, since c0 -> p0 is a many to one
// function, however, we can promise that p0 -> c1 -> p1 : p0 == p1
REPORTER_ASSERT(reporter, p0 == p1);
{
int ax = SkMulDiv255Ceiling(x, a);
REPORTER_ASSERT(reporter, ax <= a);
}
}
}
}
/**
This test fails: SkFourByteInterp does *not* preserve opaque destinations.
SkAlpha255To256 implemented as (alpha + 1) is faster than
(alpha + (alpha >> 7)), but inaccurate, and Skia intends to phase it out.
*/
/*
static void test_interp(skiatest::Reporter* reporter) {
SkRandom r;
U8CPU a0 = 0;
U8CPU a255 = 255;
for (int i = 0; i < 200; i++) {
SkColor colorSrc = r.nextU();
SkColor colorDst = r.nextU();
SkPMColor src = SkPreMultiplyColor(colorSrc);
SkPMColor dst = SkPreMultiplyColor(colorDst);
REPORTER_ASSERT(reporter, SkFourByteInterp(src, dst, a0) == dst);
REPORTER_ASSERT(reporter, SkFourByteInterp(src, dst, a255) == src);
}
}
*/
static inline void test_fast_interp(skiatest::Reporter* reporter) {
SkRandom r;
U8CPU a0 = 0;
U8CPU a255 = 255;
for (int i = 0; i < 200; i++) {
SkColor colorSrc = r.nextU();
SkColor colorDst = r.nextU();
SkPMColor src = SkPreMultiplyColor(colorSrc);
SkPMColor dst = SkPreMultiplyColor(colorDst);
REPORTER_ASSERT(reporter, SkFastFourByteInterp(src, dst, a0) == dst);
REPORTER_ASSERT(reporter, SkFastFourByteInterp(src, dst, a255) == src);
}
}
static void TestColor(skiatest::Reporter* reporter) {
test_premul(reporter);
//test_interp(reporter);
test_fast_interp(reporter);
// test_565blend(reporter);
}
#include "TestClassDef.h"
DEFINE_TESTCLASS("Color", ColorTestClass, TestColor)