Add support for an inputFP in SkTableColorFilter::asFragmentProcessor.

Additionally, fixed a performance bug which caused the fragment
processor to sample the alpha component from the color-table texture
twice. (This repeated evalation was not optimized away during
compilation.)

Change-Id: Ia2314d7918d849b894043321d6dca5da633cda4a
Bug: skia:10217
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/300699
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This commit is contained in:
John Stiles 2020-07-06 14:45:57 -04:00 committed by Skia Commit-Bot
parent 8d639e1a42
commit 3851be4555

View File

@ -20,48 +20,29 @@
#include "src/core/SkVM.h"
#include "src/core/SkWriteBuffer.h"
static const uint8_t gIdentityTable[] = {
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F,
0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27,
0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F,
0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37,
0x38, 0x39, 0x3A, 0x3B, 0x3C, 0x3D, 0x3E, 0x3F,
0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47,
0x48, 0x49, 0x4A, 0x4B, 0x4C, 0x4D, 0x4E, 0x4F,
0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57,
0x58, 0x59, 0x5A, 0x5B, 0x5C, 0x5D, 0x5E, 0x5F,
0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67,
0x68, 0x69, 0x6A, 0x6B, 0x6C, 0x6D, 0x6E, 0x6F,
0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77,
0x78, 0x79, 0x7A, 0x7B, 0x7C, 0x7D, 0x7E, 0x7F,
0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F,
0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97,
0x98, 0x99, 0x9A, 0x9B, 0x9C, 0x9D, 0x9E, 0x9F,
0xA0, 0xA1, 0xA2, 0xA3, 0xA4, 0xA5, 0xA6, 0xA7,
0xA8, 0xA9, 0xAA, 0xAB, 0xAC, 0xAD, 0xAE, 0xAF,
0xB0, 0xB1, 0xB2, 0xB3, 0xB4, 0xB5, 0xB6, 0xB7,
0xB8, 0xB9, 0xBA, 0xBB, 0xBC, 0xBD, 0xBE, 0xBF,
0xC0, 0xC1, 0xC2, 0xC3, 0xC4, 0xC5, 0xC6, 0xC7,
0xC8, 0xC9, 0xCA, 0xCB, 0xCC, 0xCD, 0xCE, 0xCF,
0xD0, 0xD1, 0xD2, 0xD3, 0xD4, 0xD5, 0xD6, 0xD7,
0xD8, 0xD9, 0xDA, 0xDB, 0xDC, 0xDD, 0xDE, 0xDF,
0xE0, 0xE1, 0xE2, 0xE3, 0xE4, 0xE5, 0xE6, 0xE7,
0xE8, 0xE9, 0xEA, 0xEB, 0xEC, 0xED, 0xEE, 0xEF,
0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7,
0xF8, 0xF9, 0xFA, 0xFB, 0xFC, 0xFD, 0xFE, 0xFF
static const uint8_t kIdentityTable[] = {
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F,
0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F,
0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B, 0x3C, 0x3D, 0x3E, 0x3F,
0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49, 0x4A, 0x4B, 0x4C, 0x4D, 0x4E, 0x4F,
0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5A, 0x5B, 0x5C, 0x5D, 0x5E, 0x5F,
0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6A, 0x6B, 0x6C, 0x6D, 0x6E, 0x6F,
0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7A, 0x7B, 0x7C, 0x7D, 0x7E, 0x7F,
0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F,
0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9A, 0x9B, 0x9C, 0x9D, 0x9E, 0x9F,
0xA0, 0xA1, 0xA2, 0xA3, 0xA4, 0xA5, 0xA6, 0xA7, 0xA8, 0xA9, 0xAA, 0xAB, 0xAC, 0xAD, 0xAE, 0xAF,
0xB0, 0xB1, 0xB2, 0xB3, 0xB4, 0xB5, 0xB6, 0xB7, 0xB8, 0xB9, 0xBA, 0xBB, 0xBC, 0xBD, 0xBE, 0xBF,
0xC0, 0xC1, 0xC2, 0xC3, 0xC4, 0xC5, 0xC6, 0xC7, 0xC8, 0xC9, 0xCA, 0xCB, 0xCC, 0xCD, 0xCE, 0xCF,
0xD0, 0xD1, 0xD2, 0xD3, 0xD4, 0xD5, 0xD6, 0xD7, 0xD8, 0xD9, 0xDA, 0xDB, 0xDC, 0xDD, 0xDE, 0xDF,
0xE0, 0xE1, 0xE2, 0xE3, 0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA, 0xEB, 0xEC, 0xED, 0xEE, 0xEF,
0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7, 0xF8, 0xF9, 0xFA, 0xFB, 0xFC, 0xFD, 0xFE, 0xFF
};
class SkTable_ColorFilter : public SkColorFilterBase {
public:
SkTable_ColorFilter(const uint8_t tableA[], const uint8_t tableR[],
const uint8_t tableG[], const uint8_t tableB[]) {
fBitmap = nullptr;
fFlags = 0;
uint8_t* dst = fStorage;
if (tableA) {
memcpy(dst, tableA, 256);
@ -87,8 +68,9 @@ public:
~SkTable_ColorFilter() override { delete fBitmap; }
#if SK_SUPPORT_GPU
std::unique_ptr<GrFragmentProcessor> asFragmentProcessor(GrRecordingContext*,
const GrColorInfo&) const override;
bool colorFilterAcceptsInputFP() const override { return true; }
GrFPResult asFragmentProcessor(std::unique_ptr<GrFragmentProcessor> inputFP,
GrRecordingContext*, const GrColorInfo&) const override;
#endif
enum {
@ -99,10 +81,10 @@ public:
};
bool onAppendStages(const SkStageRec& rec, bool shaderIsOpaque) const override {
const uint8_t *r = gIdentityTable,
*g = gIdentityTable,
*b = gIdentityTable,
*a = gIdentityTable;
const uint8_t *r = kIdentityTable,
*g = kIdentityTable,
*b = kIdentityTable,
*a = kIdentityTable;
const uint8_t* ptr = fStorage;
if (fFlags & kA_Flag) { a = ptr; ptr += 256; }
if (fFlags & kR_Flag) { r = ptr; ptr += 256; }
@ -163,17 +145,17 @@ private:
void getTableAsBitmap(SkBitmap* table) const;
mutable const SkBitmap* fBitmap; // lazily allocated
mutable const SkBitmap* fBitmap = nullptr; // lazily allocated
uint8_t fStorage[256 * 4];
unsigned fFlags;
unsigned int fFlags = 0;
friend class SkTableColorFilter;
typedef SkColorFilter INHERITED;
};
static const uint8_t gCountNibBits[] = {
static const uint8_t kCountNibBits[] = {
0, 1, 1, 2,
1, 2, 2, 3,
1, 2, 2, 3,
@ -184,9 +166,8 @@ static const uint8_t gCountNibBits[] = {
void SkTable_ColorFilter::flatten(SkWriteBuffer& buffer) const {
uint8_t storage[5*256];
int count = gCountNibBits[fFlags & 0xF];
size_t size = SkPackBits::Pack8(fStorage, count * 256, storage,
sizeof(storage));
int count = kCountNibBits[fFlags & 0xF];
size_t size = SkPackBits::Pack8(fStorage, count * 256, storage, sizeof(storage));
buffer.write32(fFlags);
buffer.writeByteArray(storage, size);
@ -194,7 +175,7 @@ void SkTable_ColorFilter::flatten(SkWriteBuffer& buffer) const {
sk_sp<SkFlattenable> SkTable_ColorFilter::CreateProc(SkReadBuffer& buffer) {
const int flags = buffer.read32();
const size_t count = gCountNibBits[flags & 0xF];
const size_t count = kCountNibBits[flags & 0xF];
SkASSERT(count <= 4);
uint8_t packedStorage[5*256];
@ -250,7 +231,7 @@ void SkTable_ColorFilter::getTableAsBitmap(SkBitmap* table) const {
for (int x = 0; x < 4; ++x) {
if (!(fFlags & kFlags[x])) {
memcpy(bitmapPixels, gIdentityTable, sizeof(gIdentityTable));
memcpy(bitmapPixels, kIdentityTable, sizeof(kIdentityTable));
} else {
memcpy(bitmapPixels, fStorage + offset, 256);
offset += 256;
@ -277,7 +258,8 @@ void SkTable_ColorFilter::getTableAsBitmap(SkBitmap* table) const {
class ColorTableEffect : public GrFragmentProcessor {
public:
static std::unique_ptr<GrFragmentProcessor> Make(GrRecordingContext* context,
static std::unique_ptr<GrFragmentProcessor> Make(std::unique_ptr<GrFragmentProcessor> inputFP,
GrRecordingContext* context,
const SkBitmap& bitmap);
~ColorTableEffect() override {}
@ -288,6 +270,15 @@ public:
return std::unique_ptr<GrFragmentProcessor>(new ColorTableEffect(*this));
}
bool hasInputFP() const {
// We always have a texture-effect child processor at index 0.
// If we have an input FP, it will be at child index 1.
return this->numChildProcessors() > 1;
}
static constexpr int kTexEffectFPIndex = 0;
static constexpr int kInputFPIndex = 1;
private:
GrGLSLFragmentProcessor* onCreateGLSLInstance() const override;
@ -295,7 +286,7 @@ private:
bool onIsEqual(const GrFragmentProcessor&) const override { return true; }
ColorTableEffect(GrSurfaceProxyView view);
ColorTableEffect(std::unique_ptr<GrFragmentProcessor> inputFP, GrSurfaceProxyView view);
explicit ColorTableEffect(const ColorTableEffect& that);
@ -304,11 +295,15 @@ private:
typedef GrFragmentProcessor INHERITED;
};
ColorTableEffect::ColorTableEffect(GrSurfaceProxyView view)
ColorTableEffect::ColorTableEffect(std::unique_ptr<GrFragmentProcessor> inputFP,
GrSurfaceProxyView view)
// Not bothering with table-specific optimizations.
: INHERITED(kColorTableEffect_ClassID, kNone_OptimizationFlags) {
auto te = GrTextureEffect::Make(std::move(view), kUnknown_SkAlphaType);
this->registerChild(std::move(te), SkSL::SampleUsage::Explicit());
this->registerChild(GrTextureEffect::Make(std::move(view), kUnknown_SkAlphaType),
SkSL::SampleUsage::Explicit());
if (inputFP) {
this->registerChild(std::move(inputFP));
}
}
ColorTableEffect::ColorTableEffect(const ColorTableEffect& that)
@ -320,23 +315,29 @@ GrGLSLFragmentProcessor* ColorTableEffect::onCreateGLSLInstance() const {
class Impl : public GrGLSLFragmentProcessor {
public:
void emitCode(EmitArgs& args) override {
const ColorTableEffect& proc = args.fFp.cast<ColorTableEffect>();
GrGLSLFPFragmentBuilder* fragBuilder = args.fFragBuilder;
fragBuilder->codeAppendf("\t\thalf4 coord = 255*unpremul(%s) + 0.5;\n",
args.fInputColor);
SkString a = this->invokeChild(0, args, "half2(coord.a, 0.5)");
SkString r = this->invokeChild(0, args, "half2(coord.r, 1.5)");
SkString g = this->invokeChild(0, args, "half2(coord.g, 2.5)");
SkString b = this->invokeChild(0, args, "half2(coord.b, 3.5)");
fragBuilder->codeAppendf("%s = half4(half3(%s.a, %s.a, %s.a) * %s.a, %s.a);",
args.fOutputColor, r.c_str(), g.c_str(), b.c_str(), a.c_str(),
a.c_str());
SkString inputColor = proc.hasInputFP()
? this->invokeChild(kInputFPIndex, args.fInputColor, args)
: SkString(args.fInputColor);
SkString a = this->invokeChild(kTexEffectFPIndex, args, "half2(coord.a, 0.5)");
SkString r = this->invokeChild(kTexEffectFPIndex, args, "half2(coord.r, 1.5)");
SkString g = this->invokeChild(kTexEffectFPIndex, args, "half2(coord.g, 2.5)");
SkString b = this->invokeChild(kTexEffectFPIndex, args, "half2(coord.b, 3.5)");
fragBuilder->codeAppendf("half4 coord = 255 * unpremul(%s) + 0.5;\n"
"%s = half4(%s.a, %s.a, %s.a, 1);\n"
"%s *= %s.a;\n",
inputColor.c_str(),
args.fOutputColor, r.c_str(), g.c_str(), b.c_str(),
args.fOutputColor, a.c_str());
}
};
return new Impl;
}
std::unique_ptr<GrFragmentProcessor> ColorTableEffect::Make(GrRecordingContext* context,
const SkBitmap& bitmap) {
std::unique_ptr<GrFragmentProcessor> ColorTableEffect::Make(
std::unique_ptr<GrFragmentProcessor> inputFP,
GrRecordingContext* context, const SkBitmap& bitmap) {
SkASSERT(kPremul_SkAlphaType == bitmap.alphaType());
SkASSERT(bitmap.isImmutable());
@ -345,7 +346,8 @@ std::unique_ptr<GrFragmentProcessor> ColorTableEffect::Make(GrRecordingContext*
return nullptr;
}
return std::unique_ptr<GrFragmentProcessor>(new ColorTableEffect(std::move(view)));
return std::unique_ptr<GrFragmentProcessor>(
new ColorTableEffect(std::move(inputFP), std::move(view)));
}
///////////////////////////////////////////////////////////////////////////////
@ -378,20 +380,21 @@ std::unique_ptr<GrFragmentProcessor> ColorTableEffect::TestCreate(GrProcessorTes
(flags & (1 << 3)) ? luts[3] : nullptr
));
sk_sp<SkColorSpace> colorSpace = GrTest::TestColorSpace(d->fRandom);
auto fp = as_CFB(filter)->asFragmentProcessor(
d->context(),
auto [success, fp] = as_CFB(filter)->asFragmentProcessor(
/*inputFP=*/nullptr, d->context(),
GrColorInfo(GrColorType::kRGBA_8888, kUnknown_SkAlphaType, std::move(colorSpace)));
SkASSERT(fp);
return fp;
SkASSERT(success);
return std::move(fp);
}
#endif
std::unique_ptr<GrFragmentProcessor> SkTable_ColorFilter::asFragmentProcessor(
GrRecordingContext* context, const GrColorInfo&) const {
GrFPResult SkTable_ColorFilter::asFragmentProcessor(std::unique_ptr<GrFragmentProcessor> inputFP,
GrRecordingContext* context,
const GrColorInfo&) const {
SkBitmap bitmap;
this->getTableAsBitmap(&bitmap);
return ColorTableEffect::Make(context, bitmap);
return GrFPSuccess(ColorTableEffect::Make(std::move(inputFP), context, bitmap));
}
#endif // SK_SUPPORT_GPU