blob: ce4af0bfca8a13471b3c2218f42c2d7c31c8f750 [file] [log] [blame]
/*
* Copyright 2010 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include "GrTextureMaker.h"
#include "SkGr.h"
#include "GrCaps.h"
#include "GrContext.h"
#include "GrDrawContext.h"
#include "GrXferProcessor.h"
#include "GrYUVProvider.h"
#include "SkColorFilter.h"
#include "SkConfig8888.h"
#include "SkCanvas.h"
#include "SkData.h"
#include "SkErrorInternals.h"
#include "SkGrPixelRef.h"
#include "SkMessageBus.h"
#include "SkPixelRef.h"
#include "SkResourceCache.h"
#include "SkTextureCompressor.h"
#include "SkYUVPlanesCache.h"
#include "effects/GrBicubicEffect.h"
#include "effects/GrConstColorProcessor.h"
#include "effects/GrDitherEffect.h"
#include "effects/GrPorterDuffXferProcessor.h"
#include "effects/GrXfermodeFragmentProcessor.h"
#include "effects/GrYUVtoRGBEffect.h"
#ifndef SK_IGNORE_ETC1_SUPPORT
# include "ktx.h"
# include "etc1.h"
#endif
/* Fill out buffer with the compressed format Ganesh expects from a colortable
based bitmap. [palette (colortable) + indices].
At the moment Ganesh only supports 8bit version. If Ganesh allowed we others
we could detect that the colortable.count is <= 16, and then repack the
indices as nibbles to save RAM, but it would take more time (i.e. a lot
slower than memcpy), so skipping that for now.
Ganesh wants a full 256 palette entry, even though Skia's ctable is only as big
as the colortable.count says it is.
*/
static void build_index8_data(void* buffer, const SkBitmap& bitmap) {
SkASSERT(kIndex_8_SkColorType == bitmap.colorType());
SkAutoLockPixels alp(bitmap);
if (!bitmap.readyToDraw()) {
SkDEBUGFAIL("bitmap not ready to draw!");
return;
}
SkColorTable* ctable = bitmap.getColorTable();
char* dst = (char*)buffer;
const int count = ctable->count();
SkDstPixelInfo dstPI;
dstPI.fColorType = kRGBA_8888_SkColorType;
dstPI.fAlphaType = kPremul_SkAlphaType;
dstPI.fPixels = buffer;
dstPI.fRowBytes = count * sizeof(SkPMColor);
SkSrcPixelInfo srcPI;
srcPI.fColorType = kN32_SkColorType;
srcPI.fAlphaType = kPremul_SkAlphaType;
srcPI.fPixels = ctable->readColors();
srcPI.fRowBytes = count * sizeof(SkPMColor);
srcPI.convertPixelsTo(&dstPI, count, 1);
// always skip a full 256 number of entries, even if we memcpy'd fewer
dst += 256 * sizeof(GrColor);
if ((unsigned)bitmap.width() == bitmap.rowBytes()) {
memcpy(dst, bitmap.getPixels(), bitmap.getSize());
} else {
// need to trim off the extra bytes per row
size_t width = bitmap.width();
size_t rowBytes = bitmap.rowBytes();
const char* src = (const char*)bitmap.getPixels();
for (int y = 0; y < bitmap.height(); y++) {
memcpy(dst, src, width);
src += rowBytes;
dst += width;
}
}
}
////////////////////////////////////////////////////////////////////////////////
static void get_stretch(const GrCaps& caps, int width, int height,
const GrTextureParams& params, SkGrStretch* stretch) {
stretch->fType = SkGrStretch::kNone_Type;
bool doStretch = false;
if (params.isTiled() && !caps.npotTextureTileSupport() &&
(!SkIsPow2(width) || !SkIsPow2(height))) {
doStretch = true;
stretch->fWidth = GrNextPow2(SkTMax(width, caps.minTextureSize()));
stretch->fHeight = GrNextPow2(SkTMax(height, caps.minTextureSize()));
} else if (width < caps.minTextureSize() || height < caps.minTextureSize()) {
// The small texture issues appear to be with tiling. Hence it seems ok to scale them
// up using the GPU. If issues persist we may need to CPU-stretch.
doStretch = true;
stretch->fWidth = SkTMax(width, caps.minTextureSize());
stretch->fHeight = SkTMax(height, caps.minTextureSize());
}
if (doStretch) {
switch(params.filterMode()) {
case GrTextureParams::kNone_FilterMode:
stretch->fType = SkGrStretch::kNearest_Type;
break;
case GrTextureParams::kBilerp_FilterMode:
case GrTextureParams::kMipMap_FilterMode:
stretch->fType = SkGrStretch::kBilerp_Type;
break;
}
} else {
stretch->fWidth = -1;
stretch->fHeight = -1;
stretch->fType = SkGrStretch::kNone_Type;
}
}
bool GrMakeStretchedKey(const GrUniqueKey& origKey, const SkGrStretch& stretch,
GrUniqueKey* stretchedKey) {
if (origKey.isValid() && SkGrStretch::kNone_Type != stretch.fType) {
uint32_t width = SkToU16(stretch.fWidth);
uint32_t height = SkToU16(stretch.fHeight);
static const GrUniqueKey::Domain kDomain = GrUniqueKey::GenerateDomain();
GrUniqueKey::Builder builder(stretchedKey, origKey, kDomain, 2);
builder[0] = stretch.fType;
builder[1] = width | (height << 16);
builder.finish();
return true;
}
SkASSERT(!stretchedKey->isValid());
return false;
}
static void make_unstretched_key(GrUniqueKey* key, uint32_t imageID, const SkIRect& subset) {
SkASSERT(SkIsU16(subset.width()));
SkASSERT(SkIsU16(subset.height()));
static const GrUniqueKey::Domain kDomain = GrUniqueKey::GenerateDomain();
GrUniqueKey::Builder builder(key, kDomain, 4);
builder[0] = imageID;
builder[1] = subset.x();
builder[2] = subset.y();
builder[3] = subset.width() | (subset.height() << 16);
}
void GrMakeKeyFromImageID(GrUniqueKey* key, uint32_t imageID, const SkIRect& subset,
const GrCaps& caps, const GrTextureParams& params) {
SkGrStretch stretch;
get_stretch(caps, subset.width(), subset.height(), params, &stretch);
if (SkGrStretch::kNone_Type != stretch.fType) {
GrUniqueKey tmpKey;
make_unstretched_key(&tmpKey, imageID, subset);
if (!GrMakeStretchedKey(tmpKey, stretch, key)) {
*key = tmpKey;
}
} else {
make_unstretched_key(key, imageID, subset);
}
}
static void make_image_keys(uint32_t imageID, const SkIRect& subset, const SkGrStretch& stretch,
GrUniqueKey* key, GrUniqueKey* stretchedKey) {
make_unstretched_key(key, imageID, subset);
if (SkGrStretch::kNone_Type != stretch.fType) {
GrMakeStretchedKey(*key, stretch, stretchedKey);
}
}
GrSurfaceDesc GrImageInfoToSurfaceDesc(const SkImageInfo& info) {
GrSurfaceDesc desc;
desc.fFlags = kNone_GrSurfaceFlags;
desc.fWidth = info.width();
desc.fHeight = info.height();
desc.fConfig = SkImageInfo2GrPixelConfig(info);
desc.fSampleCnt = 0;
return desc;
}
namespace {
// When the SkPixelRef genID changes, invalidate a corresponding GrResource described by key.
class BitmapInvalidator : public SkPixelRef::GenIDChangeListener {
public:
explicit BitmapInvalidator(const GrUniqueKey& key) : fMsg(key) {}
private:
GrUniqueKeyInvalidatedMessage fMsg;
void onChange() override {
SkMessageBus<GrUniqueKeyInvalidatedMessage>::Post(fMsg);
}
};
} // namespace
GrTexture* GrCreateTextureForPixels(GrContext* ctx,
const GrUniqueKey& optionalKey,
GrSurfaceDesc desc,
SkPixelRef* pixelRefForInvalidationNotification,
const void* pixels,
size_t rowBytes) {
GrTexture* result = ctx->textureProvider()->createTexture(desc, true, pixels, rowBytes);
if (result && optionalKey.isValid()) {
if (pixelRefForInvalidationNotification) {
BitmapInvalidator* listener = new BitmapInvalidator(optionalKey);
pixelRefForInvalidationNotification->addGenIDChangeListener(listener);
}
ctx->textureProvider()->assignUniqueKeyToTexture(optionalKey, result);
}
return result;
}
// creates a new texture that is the input texture scaled up. If optionalKey is valid it will be
// set on the new texture. stretch controls whether the scaling is done using nearest or bilerp
// filtering and the size to stretch the texture to.
GrTexture* stretch_texture(GrTexture* inputTexture, const SkGrStretch& stretch,
SkPixelRef* pixelRef,
const GrUniqueKey& optionalKey) {
SkASSERT(SkGrStretch::kNone_Type != stretch.fType);
GrContext* context = inputTexture->getContext();
SkASSERT(context);
const GrCaps* caps = context->caps();
// Either it's a cache miss or the original wasn't cached to begin with.
GrSurfaceDesc rtDesc = inputTexture->desc();
rtDesc.fFlags = rtDesc.fFlags | kRenderTarget_GrSurfaceFlag;
rtDesc.fWidth = stretch.fWidth;
rtDesc.fHeight = stretch.fHeight;
rtDesc.fConfig = GrMakePixelConfigUncompressed(rtDesc.fConfig);
// If the config isn't renderable try converting to either A8 or an 32 bit config. Otherwise,
// fail.
if (!caps->isConfigRenderable(rtDesc.fConfig, false)) {
if (GrPixelConfigIsAlphaOnly(rtDesc.fConfig)) {
if (caps->isConfigRenderable(kAlpha_8_GrPixelConfig, false)) {
rtDesc.fConfig = kAlpha_8_GrPixelConfig;
} else if (caps->isConfigRenderable(kSkia8888_GrPixelConfig, false)) {
rtDesc.fConfig = kSkia8888_GrPixelConfig;
} else {
return nullptr;
}
} else if (kRGB_GrColorComponentFlags ==
(kRGB_GrColorComponentFlags & GrPixelConfigComponentMask(rtDesc.fConfig))) {
if (caps->isConfigRenderable(kSkia8888_GrPixelConfig, false)) {
rtDesc.fConfig = kSkia8888_GrPixelConfig;
} else {
return nullptr;
}
} else {
return nullptr;
}
}
SkAutoTUnref<GrTexture> stretched(GrCreateTextureForPixels(context, optionalKey, rtDesc,
pixelRef, nullptr,0));
if (!stretched) {
return nullptr;
}
GrPaint paint;
// If filtering is not desired then we want to ensure all texels in the resampled image are
// copies of texels from the original.
GrTextureParams params(SkShader::kClamp_TileMode,
SkGrStretch::kBilerp_Type == stretch.fType ?
GrTextureParams::kBilerp_FilterMode :
GrTextureParams::kNone_FilterMode);
paint.addColorTextureProcessor(inputTexture, SkMatrix::I(), params);
SkRect rect = SkRect::MakeWH(SkIntToScalar(rtDesc.fWidth), SkIntToScalar(rtDesc.fHeight));
SkRect localRect = SkRect::MakeWH(1.f, 1.f);
SkAutoTUnref<GrDrawContext> drawContext(context->drawContext());
if (!drawContext) {
return nullptr;
}
drawContext->drawNonAARectToRect(stretched->asRenderTarget(), GrClip::WideOpen(), paint,
SkMatrix::I(), rect, localRect);
return stretched.detach();
}
GrPixelConfig GrIsCompressedTextureDataSupported(GrContext* ctx, SkData* data,
int expectedW, int expectedH,
const void** outStartOfDataToUpload) {
*outStartOfDataToUpload = nullptr;
#ifndef SK_IGNORE_ETC1_SUPPORT
if (!ctx->caps()->isConfigTexturable(kETC1_GrPixelConfig)) {
return kUnknown_GrPixelConfig;
}
const uint8_t* bytes = data->bytes();
if (data->size() > ETC_PKM_HEADER_SIZE && etc1_pkm_is_valid(bytes)) {
// Does the data match the dimensions of the bitmap? If not,
// then we don't know how to scale the image to match it...
if (etc1_pkm_get_width(bytes) != (unsigned)expectedW ||
etc1_pkm_get_height(bytes) != (unsigned)expectedH)
{
return kUnknown_GrPixelConfig;
}
*outStartOfDataToUpload = bytes + ETC_PKM_HEADER_SIZE;
return kETC1_GrPixelConfig;
} else if (SkKTXFile::is_ktx(bytes)) {
SkKTXFile ktx(data);
// Is it actually an ETC1 texture?
if (!ktx.isCompressedFormat(SkTextureCompressor::kETC1_Format)) {
return kUnknown_GrPixelConfig;
}
// Does the data match the dimensions of the bitmap? If not,
// then we don't know how to scale the image to match it...
if (ktx.width() != expectedW || ktx.height() != expectedH) {
return kUnknown_GrPixelConfig;
}
*outStartOfDataToUpload = ktx.pixelData();
return kETC1_GrPixelConfig;
}
#endif
return kUnknown_GrPixelConfig;
}
static GrTexture* load_etc1_texture(GrContext* ctx, const GrUniqueKey& optionalKey,
const SkBitmap &bm, GrSurfaceDesc desc) {
SkAutoTUnref<SkData> data(bm.pixelRef()->refEncodedData());
if (!data) {
return nullptr;
}
const void* startOfTexData;
desc.fConfig = GrIsCompressedTextureDataSupported(ctx, data, bm.width(), bm.height(),
&startOfTexData);
if (kUnknown_GrPixelConfig == desc.fConfig) {
return nullptr;
}
return GrCreateTextureForPixels(ctx, optionalKey, desc, bm.pixelRef(), startOfTexData, 0);
}
/*
* Once we have made SkImages handle all lazy/deferred/generated content, the YUV apis will
* be gone from SkPixelRef, and we can remove this subclass entirely.
*/
class PixelRef_GrYUVProvider : public GrYUVProvider {
SkPixelRef* fPR;
public:
PixelRef_GrYUVProvider(SkPixelRef* pr) : fPR(pr) {}
uint32_t onGetID() override { return fPR->getGenerationID(); }
bool onGetYUVSizes(SkISize sizes[3]) override {
return fPR->getYUV8Planes(sizes, nullptr, nullptr, nullptr);
}
bool onGetYUVPlanes(SkISize sizes[3], void* planes[3], size_t rowBytes[3],
SkYUVColorSpace* space) override {
return fPR->getYUV8Planes(sizes, planes, rowBytes, space);
}
};
static GrTexture* load_yuv_texture(GrContext* ctx, const GrUniqueKey& optionalKey,
const SkBitmap& bm, const GrSurfaceDesc& desc) {
// Subsets are not supported, the whole pixelRef is loaded when using YUV decoding
SkPixelRef* pixelRef = bm.pixelRef();
if ((nullptr == pixelRef) ||
(pixelRef->info().width() != bm.info().width()) ||
(pixelRef->info().height() != bm.info().height())) {
return nullptr;
}
const bool useCache = optionalKey.isValid();
PixelRef_GrYUVProvider provider(pixelRef);
GrTexture* texture = provider.refAsTexture(ctx, desc, useCache);
if (!texture) {
return nullptr;
}
if (useCache) {
BitmapInvalidator* listener = new BitmapInvalidator(optionalKey);
pixelRef->addGenIDChangeListener(listener);
ctx->textureProvider()->assignUniqueKeyToTexture(optionalKey, texture);
}
return texture;
}
static GrTexture* create_unstretched_bitmap_texture(GrContext* ctx,
const SkBitmap& origBitmap,
const GrUniqueKey& optionalKey) {
if (origBitmap.width() < ctx->caps()->minTextureSize() ||
origBitmap.height() < ctx->caps()->minTextureSize()) {
return nullptr;
}
SkBitmap tmpBitmap;
const SkBitmap* bitmap = &origBitmap;
GrSurfaceDesc desc = GrImageInfoToSurfaceDesc(bitmap->info());
const GrCaps* caps = ctx->caps();
if (kIndex_8_SkColorType == bitmap->colorType()) {
if (caps->isConfigTexturable(kIndex_8_GrPixelConfig)) {
size_t imageSize = GrCompressedFormatDataSize(kIndex_8_GrPixelConfig,
bitmap->width(), bitmap->height());
SkAutoMalloc storage(imageSize);
build_index8_data(storage.get(), origBitmap);
// our compressed data will be trimmed, so pass width() for its
// "rowBytes", since they are the same now.
return GrCreateTextureForPixels(ctx, optionalKey, desc, origBitmap.pixelRef(),
storage.get(), bitmap->width());
} else {
origBitmap.copyTo(&tmpBitmap, kN32_SkColorType);
// now bitmap points to our temp, which has been promoted to 32bits
bitmap = &tmpBitmap;
desc.fConfig = SkImageInfo2GrPixelConfig(bitmap->info());
}
} else if (!bitmap->readyToDraw()) {
// If the bitmap had compressed data and was then uncompressed, it'll still return
// compressed data on 'refEncodedData' and upload it. Probably not good, since if
// the bitmap has available pixels, then they might not be what the decompressed
// data is.
GrTexture *texture = load_etc1_texture(ctx, optionalKey, *bitmap, desc);
if (texture) {
return texture;
}
}
GrTexture *texture = load_yuv_texture(ctx, optionalKey, *bitmap, desc);
if (texture) {
return texture;
}
SkAutoLockPixels alp(*bitmap);
if (!bitmap->readyToDraw()) {
return nullptr;
}
return GrCreateTextureForPixels(ctx, optionalKey, desc, origBitmap.pixelRef(),
bitmap->getPixels(), bitmap->rowBytes());
}
static SkBitmap stretch_on_cpu(const SkBitmap& bmp, const SkGrStretch& stretch) {
SkBitmap stretched;
stretched.allocN32Pixels(stretch.fWidth, stretch.fHeight);
SkCanvas canvas(stretched);
SkPaint paint;
switch (stretch.fType) {
case SkGrStretch::kNearest_Type:
paint.setFilterQuality(kNone_SkFilterQuality);
break;
case SkGrStretch::kBilerp_Type:
paint.setFilterQuality(kLow_SkFilterQuality);
break;
case SkGrStretch::kNone_Type:
SkDEBUGFAIL("Shouldn't get here.");
break;
}
SkRect dstRect = SkRect::MakeWH(SkIntToScalar(stretch.fWidth), SkIntToScalar(stretch.fHeight));
canvas.drawBitmapRect(bmp, dstRect, &paint);
return stretched;
}
bool GrIsImageInCache(const GrContext* ctx, uint32_t imageID, const SkIRect& subset,
GrTexture* nativeTexture, const GrTextureParams& params) {
SkGrStretch stretch;
get_stretch(*ctx->caps(), subset.width(), subset.height(), params, &stretch);
// Handle the case where the bitmap/image is explicitly texture backed.
if (nativeTexture) {
if (SkGrStretch::kNone_Type == stretch.fType) {
return true;
}
const GrUniqueKey& key = nativeTexture->getUniqueKey();
if (!key.isValid()) {
return false;
}
GrUniqueKey stretchedKey;
GrMakeStretchedKey(key, stretch, &stretchedKey);
return ctx->textureProvider()->existsTextureWithUniqueKey(stretchedKey);
}
GrUniqueKey key, stretchedKey;
make_image_keys(imageID, subset, stretch, &key, &stretchedKey);
return ctx->textureProvider()->existsTextureWithUniqueKey(
(SkGrStretch::kNone_Type == stretch.fType) ? key : stretchedKey);
}
class Bitmap_GrTextureMaker : public GrTextureMaker {
public:
Bitmap_GrTextureMaker(const SkBitmap& bitmap)
: INHERITED(bitmap.width(), bitmap.height())
, fBitmap(bitmap)
{}
protected:
GrTexture* onRefUnstretchedTexture(GrContext* ctx) override {
GrTexture* tex = fBitmap.getTexture();
if (tex) {
return SkRef(tex);
}
GrUniqueKey unstretchedKey;
make_unstretched_key(&unstretchedKey, fBitmap.getGenerationID(), fBitmap.getSubset());
GrTexture* result = ctx->textureProvider()->findAndRefTextureByUniqueKey(unstretchedKey);
if (result) {
return result;
}
return create_unstretched_bitmap_texture(ctx, fBitmap, unstretchedKey);
}
bool onMakeStretchedKey(const SkGrStretch& stretch, GrUniqueKey* stretchedKey) override {
if (fBitmap.isVolatile()) {
return false;
}
GrUniqueKey unstretchedKey;
make_unstretched_key(&unstretchedKey, fBitmap.getGenerationID(), fBitmap.getSubset());
return GrMakeStretchedKey(unstretchedKey, stretch, stretchedKey);
}
void onNotifyStretchCached(const GrUniqueKey& stretchedKey) override {
fBitmap.pixelRef()->addGenIDChangeListener(new BitmapInvalidator(stretchedKey));
}
bool onGetROBitmap(SkBitmap* bitmap) override {
*bitmap = fBitmap;
return true;
}
private:
const SkBitmap fBitmap;
typedef GrTextureMaker INHERITED;
};
GrTexture* GrRefCachedBitmapTexture(GrContext* ctx, const SkBitmap& bitmap,
const GrTextureParams& params) {
return Bitmap_GrTextureMaker(bitmap).refCachedTexture(ctx, params);
}
///////////////////////////////////////////////////////////////////////////////
// alphatype is ignore for now, but if GrPixelConfig is expanded to encompass
// alpha info, that will be considered.
GrPixelConfig SkImageInfo2GrPixelConfig(SkColorType ct, SkAlphaType, SkColorProfileType pt) {
switch (ct) {
case kUnknown_SkColorType:
return kUnknown_GrPixelConfig;
case kAlpha_8_SkColorType:
return kAlpha_8_GrPixelConfig;
case kRGB_565_SkColorType:
return kRGB_565_GrPixelConfig;
case kARGB_4444_SkColorType:
return kRGBA_4444_GrPixelConfig;
case kRGBA_8888_SkColorType:
//if (kSRGB_SkColorProfileType == pt) {
// return kSRGBA_8888_GrPixelConfig;
//}
return kRGBA_8888_GrPixelConfig;
case kBGRA_8888_SkColorType:
return kBGRA_8888_GrPixelConfig;
case kIndex_8_SkColorType:
return kIndex_8_GrPixelConfig;
case kGray_8_SkColorType:
return kAlpha_8_GrPixelConfig; // TODO: gray8 support on gpu
}
SkASSERT(0); // shouldn't get here
return kUnknown_GrPixelConfig;
}
bool GrPixelConfig2ColorAndProfileType(GrPixelConfig config, SkColorType* ctOut,
SkColorProfileType* ptOut) {
SkColorType ct;
SkColorProfileType pt = kLinear_SkColorProfileType;
switch (config) {
case kAlpha_8_GrPixelConfig:
ct = kAlpha_8_SkColorType;
break;
case kIndex_8_GrPixelConfig:
ct = kIndex_8_SkColorType;
break;
case kRGB_565_GrPixelConfig:
ct = kRGB_565_SkColorType;
break;
case kRGBA_4444_GrPixelConfig:
ct = kARGB_4444_SkColorType;
break;
case kRGBA_8888_GrPixelConfig:
ct = kRGBA_8888_SkColorType;
break;
case kBGRA_8888_GrPixelConfig:
ct = kBGRA_8888_SkColorType;
break;
case kSRGBA_8888_GrPixelConfig:
ct = kRGBA_8888_SkColorType;
pt = kSRGB_SkColorProfileType;
break;
default:
return false;
}
if (ctOut) {
*ctOut = ct;
}
if (ptOut) {
*ptOut = pt;
}
return true;
}
////////////////////////////////////////////////////////////////////////////////////////////////
static inline bool blend_requires_shader(const SkXfermode::Mode mode, bool primitiveIsSrc) {
if (primitiveIsSrc) {
return SkXfermode::kSrc_Mode != mode;
} else {
return SkXfermode::kDst_Mode != mode;
}
}
static inline bool skpaint_to_grpaint_impl(GrContext* context,
const SkPaint& skPaint,
const SkMatrix& viewM,
const GrFragmentProcessor** shaderProcessor,
SkXfermode::Mode* primColorMode,
bool primitiveIsSrc,
GrPaint* grPaint) {
grPaint->setAntiAlias(skPaint.isAntiAlias());
// Setup the initial color considering the shader, the SkPaint color, and the presence or not
// of per-vertex colors.
SkAutoTUnref<const GrFragmentProcessor> aufp;
const GrFragmentProcessor* shaderFP = nullptr;
if (!primColorMode || blend_requires_shader(*primColorMode, primitiveIsSrc)) {
if (shaderProcessor) {
shaderFP = *shaderProcessor;
} else if (const SkShader* shader = skPaint.getShader()) {
aufp.reset(shader->asFragmentProcessor(context, viewM, nullptr,
skPaint.getFilterQuality()));
shaderFP = aufp;
if (!shaderFP) {
return false;
}
}
}
// Set this in below cases if the output of the shader/paint-color/paint-alpha/primXfermode is
// a known constant value. In that case we can simply apply a color filter during this
// conversion without converting the color filter to a GrFragmentProcessor.
bool applyColorFilterToPaintColor = false;
if (shaderFP) {
if (primColorMode) {
// There is a blend between the primitive color and the shader color. The shader sees
// the opaque paint color. The shader's output is blended using the provided mode by
// the primitive color. The blended color is then modulated by the paint's alpha.
// The geometry processor will insert the primitive color to start the color chain, so
// the GrPaint color will be ignored.
GrColor shaderInput = SkColorToOpaqueGrColor(skPaint.getColor());
shaderFP = GrFragmentProcessor::OverrideInput(shaderFP, shaderInput);
aufp.reset(shaderFP);
if (primitiveIsSrc) {
shaderFP = GrXfermodeFragmentProcessor::CreateFromDstProcessor(shaderFP,
*primColorMode);
} else {
shaderFP = GrXfermodeFragmentProcessor::CreateFromSrcProcessor(shaderFP,
*primColorMode);
}
aufp.reset(shaderFP);
// The above may return null if compose results in a pass through of the prim color.
if (shaderFP) {
grPaint->addColorFragmentProcessor(shaderFP);
}
GrColor paintAlpha = SkColorAlphaToGrColor(skPaint.getColor());
if (GrColor_WHITE != paintAlpha) {
grPaint->addColorFragmentProcessor(GrConstColorProcessor::Create(
paintAlpha, GrConstColorProcessor::kModulateRGBA_InputMode))->unref();
}
} else {
// The shader's FP sees the paint unpremul color
grPaint->setColor(SkColorToUnpremulGrColor(skPaint.getColor()));
grPaint->addColorFragmentProcessor(shaderFP);
}
} else {
if (primColorMode) {
// There is a blend between the primitive color and the paint color. The blend considers
// the opaque paint color. The paint's alpha is applied to the post-blended color.
SkAutoTUnref<const GrFragmentProcessor> processor(
GrConstColorProcessor::Create(SkColorToOpaqueGrColor(skPaint.getColor()),
GrConstColorProcessor::kIgnore_InputMode));
if (primitiveIsSrc) {
processor.reset(GrXfermodeFragmentProcessor::CreateFromDstProcessor(processor,
*primColorMode));
} else {
processor.reset(GrXfermodeFragmentProcessor::CreateFromSrcProcessor(processor,
*primColorMode));
}
if (processor) {
grPaint->addColorFragmentProcessor(processor);
}
grPaint->setColor(SkColorToOpaqueGrColor(skPaint.getColor()));
GrColor paintAlpha = SkColorAlphaToGrColor(skPaint.getColor());
if (GrColor_WHITE != paintAlpha) {
grPaint->addColorFragmentProcessor(GrConstColorProcessor::Create(
paintAlpha, GrConstColorProcessor::kModulateRGBA_InputMode))->unref();
}
} else {
// No shader, no primitive color.
grPaint->setColor(SkColorToPremulGrColor(skPaint.getColor()));
applyColorFilterToPaintColor = true;
}
}
SkColorFilter* colorFilter = skPaint.getColorFilter();
if (colorFilter) {
if (applyColorFilterToPaintColor) {
grPaint->setColor(SkColorToPremulGrColor(colorFilter->filterColor(skPaint.getColor())));
} else {
SkAutoTUnref<const GrFragmentProcessor> cfFP(
colorFilter->asFragmentProcessor(context));
if (cfFP) {
grPaint->addColorFragmentProcessor(cfFP);
} else {
return false;
}
}
}
SkXfermode* mode = skPaint.getXfermode();
GrXPFactory* xpFactory = nullptr;
if (!SkXfermode::AsXPFactory(mode, &xpFactory)) {
// Fall back to src-over
// return false here?
xpFactory = GrPorterDuffXPFactory::Create(SkXfermode::kSrcOver_Mode);
}
SkASSERT(xpFactory);
grPaint->setXPFactory(xpFactory)->unref();
#ifndef SK_IGNORE_GPU_DITHER
if (skPaint.isDither() && grPaint->numColorFragmentProcessors() > 0) {
grPaint->addColorFragmentProcessor(GrDitherEffect::Create())->unref();
}
#endif
return true;
}
bool SkPaintToGrPaint(GrContext* context, const SkPaint& skPaint, const SkMatrix& viewM,
GrPaint* grPaint) {
return skpaint_to_grpaint_impl(context, skPaint, viewM, nullptr, nullptr, false, grPaint);
}
/** Replaces the SkShader (if any) on skPaint with the passed in GrFragmentProcessor. */
bool SkPaintToGrPaintReplaceShader(GrContext* context,
const SkPaint& skPaint,
const GrFragmentProcessor* shaderFP,
GrPaint* grPaint) {
if (!shaderFP) {
return false;
}
return skpaint_to_grpaint_impl(context, skPaint, SkMatrix::I(), &shaderFP, nullptr, false,
grPaint);
}
/** Ignores the SkShader (if any) on skPaint. */
bool SkPaintToGrPaintNoShader(GrContext* context,
const SkPaint& skPaint,
GrPaint* grPaint) {
// Use a ptr to a nullptr to to indicate that the SkShader is ignored and not replaced.
static const GrFragmentProcessor* kNullShaderFP = nullptr;
static const GrFragmentProcessor** kIgnoreShader = &kNullShaderFP;
return skpaint_to_grpaint_impl(context, skPaint, SkMatrix::I(), kIgnoreShader, nullptr, false,
grPaint);
}
/** Blends the SkPaint's shader (or color if no shader) with a per-primitive color which must
be setup as a vertex attribute using the specified SkXfermode::Mode. */
bool SkPaintToGrPaintWithXfermode(GrContext* context,
const SkPaint& skPaint,
const SkMatrix& viewM,
SkXfermode::Mode primColorMode,
bool primitiveIsSrc,
GrPaint* grPaint) {
return skpaint_to_grpaint_impl(context, skPaint, viewM, nullptr, &primColorMode, primitiveIsSrc,
grPaint);
}
////////////////////////////////////////////////////////////////////////////////////////////////
SkImageInfo GrMakeInfoFromTexture(GrTexture* tex, int w, int h, bool isOpaque) {
#ifdef SK_DEBUG
const GrSurfaceDesc& desc = tex->desc();
SkASSERT(w <= desc.fWidth);
SkASSERT(h <= desc.fHeight);
#endif
const GrPixelConfig config = tex->config();
SkColorType ct;
SkAlphaType at = isOpaque ? kOpaque_SkAlphaType : kPremul_SkAlphaType;
if (!GrPixelConfig2ColorAndProfileType(config, &ct, nullptr)) {
ct = kUnknown_SkColorType;
}
return SkImageInfo::Make(w, h, ct, at);
}
void GrWrapTextureInBitmap(GrTexture* src, int w, int h, bool isOpaque, SkBitmap* dst) {
const SkImageInfo info = GrMakeInfoFromTexture(src, w, h, isOpaque);
dst->setInfo(info);
dst->setPixelRef(new SkGrPixelRef(info, src))->unref();
}
GrTextureParams::FilterMode GrSkFilterQualityToGrFilterMode(SkFilterQuality paintFilterQuality,
const SkMatrix& viewM,
const SkMatrix& localM,
bool* doBicubic) {
*doBicubic = false;
GrTextureParams::FilterMode textureFilterMode;
switch (paintFilterQuality) {
case kNone_SkFilterQuality:
textureFilterMode = GrTextureParams::kNone_FilterMode;
break;
case kLow_SkFilterQuality:
textureFilterMode = GrTextureParams::kBilerp_FilterMode;
break;
case kMedium_SkFilterQuality: {
SkMatrix matrix;
matrix.setConcat(viewM, localM);
if (matrix.getMinScale() < SK_Scalar1) {
textureFilterMode = GrTextureParams::kMipMap_FilterMode;
} else {
// Don't trigger MIP level generation unnecessarily.
textureFilterMode = GrTextureParams::kBilerp_FilterMode;
}
break;
}
case kHigh_SkFilterQuality: {
SkMatrix matrix;
matrix.setConcat(viewM, localM);
*doBicubic = GrBicubicEffect::ShouldUseBicubic(matrix, &textureFilterMode);
break;
}
default:
SkErrorInternals::SetError( kInvalidPaint_SkError,
"Sorry, I don't understand the filtering "
"mode you asked for. Falling back to "
"MIPMaps.");
textureFilterMode = GrTextureParams::kMipMap_FilterMode;
break;
}
return textureFilterMode;
}
////////////////////////////////////////////////////////////////////////////////////////////////
GrTexture* GrTextureMaker::refCachedTexture(GrContext* ctx, const GrTextureParams& params) {
SkGrStretch stretch;
get_stretch(*ctx->caps(), this->width(), this->height(), params, &stretch);
if (SkGrStretch::kNone_Type == stretch.fType) {
return this->onRefUnstretchedTexture(ctx);
}
GrUniqueKey stretchedKey;
if (this->onMakeStretchedKey(stretch, &stretchedKey)) {
GrTexture* result = ctx->textureProvider()->findAndRefTextureByUniqueKey(stretchedKey);
if (result) {
return result;
}
}
GrTexture* result = this->onGenerateStretchedTexture(ctx, stretch);
if (!result) {
return nullptr;
}
if (stretchedKey.isValid()) {
ctx->textureProvider()->assignUniqueKeyToTexture(stretchedKey, result);
this->onNotifyStretchCached(stretchedKey);
}
return result;
}
GrTexture* GrTextureMaker::onGenerateStretchedTexture(GrContext* ctx, const SkGrStretch& stretch) {
if (this->width() < ctx->caps()->minTextureSize() ||
this->height() < ctx->caps()->minTextureSize())
{
// we can't trust our ability to use HW to perform the stretch, so we request
// a raster instead, and perform the stretch on the CPU.
SkBitmap bitmap;
if (!this->onGetROBitmap(&bitmap)) {
return nullptr;
}
SkBitmap stretchedBmp = stretch_on_cpu(bitmap, stretch);
return create_unstretched_bitmap_texture(ctx, stretchedBmp, GrUniqueKey());
} else {
SkAutoTUnref<GrTexture> unstretched(this->onRefUnstretchedTexture(ctx));
if (!unstretched) {
return nullptr;
}
return stretch_texture(unstretched, stretch, nullptr, GrUniqueKey());
}
}