skia2/tests/PrimitiveProcessorTest.cpp
Brian Salomon bf6b9795f6 GrSurfaceContext not ref counted.
Also don't specify redundant width/height to SkGpuDevice.

Change-Id: I389df5c4b073c2c05632ba6b7c95b02a22dfaf98
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/235824
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
2019-08-21 15:31:23 +00:00

171 lines
6.6 KiB
C++

/*
* Copyright 2016 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
// This is a GPU-backend specific test. It relies on static intializers to work
#include "include/core/SkTypes.h"
#include "tests/Test.h"
#include "include/core/SkString.h"
#include "include/gpu/GrContext.h"
#include "src/core/SkPointPriv.h"
#include "src/gpu/GrContextPriv.h"
#include "src/gpu/GrGeometryProcessor.h"
#include "src/gpu/GrGpu.h"
#include "src/gpu/GrMemoryPool.h"
#include "src/gpu/GrOpFlushState.h"
#include "src/gpu/GrRenderTargetContext.h"
#include "src/gpu/GrRenderTargetContextPriv.h"
#include "src/gpu/glsl/GrGLSLFragmentShaderBuilder.h"
#include "src/gpu/glsl/GrGLSLGeometryProcessor.h"
#include "src/gpu/glsl/GrGLSLVarying.h"
#include "src/gpu/ops/GrMeshDrawOp.h"
namespace {
class Op : public GrMeshDrawOp {
public:
DEFINE_OP_CLASS_ID
const char* name() const override { return "Dummy Op"; }
static std::unique_ptr<GrDrawOp> Make(GrContext* context, int numAttribs) {
GrOpMemoryPool* pool = context->priv().opMemoryPool();
return pool->allocate<Op>(numAttribs);
}
FixedFunctionFlags fixedFunctionFlags() const override {
return FixedFunctionFlags::kNone;
}
GrProcessorSet::Analysis finalize(const GrCaps&, const GrAppliedClip*,
bool hasMixedSampledCoverage, GrClampType) override {
return GrProcessorSet::EmptySetAnalysis();
}
private:
friend class ::GrOpMemoryPool;
Op(int numAttribs) : INHERITED(ClassID()), fNumAttribs(numAttribs) {
this->setBounds(SkRect::MakeWH(1.f, 1.f), HasAABloat::kNo, IsZeroArea::kNo);
}
void onPrepareDraws(Target* target) override {
class GP : public GrGeometryProcessor {
public:
GP(int numAttribs) : INHERITED(kGP_ClassID), fNumAttribs(numAttribs) {
SkASSERT(numAttribs > 1);
fAttribNames.reset(new SkString[numAttribs]);
fAttributes.reset(new Attribute[numAttribs]);
for (auto i = 0; i < numAttribs; ++i) {
fAttribNames[i].printf("attr%d", i);
// This gives us more of a mix of attribute types, and allows the
// component count to fit within the limits for iOS Metal.
if (i & 0x1) {
fAttributes[i] = {fAttribNames[i].c_str(), kFloat_GrVertexAttribType,
kFloat_GrSLType};
} else {
fAttributes[i] = {fAttribNames[i].c_str(), kFloat2_GrVertexAttribType,
kFloat2_GrSLType};
}
}
this->setVertexAttributes(fAttributes.get(), numAttribs);
}
const char* name() const override { return "Dummy GP"; }
GrGLSLPrimitiveProcessor* createGLSLInstance(const GrShaderCaps&) const override {
class GLSLGP : public GrGLSLGeometryProcessor {
public:
void onEmitCode(EmitArgs& args, GrGPArgs* gpArgs) override {
const GP& gp = args.fGP.cast<GP>();
args.fVaryingHandler->emitAttributes(gp);
this->writeOutputPosition(args.fVertBuilder, gpArgs,
gp.fAttributes[0].name());
GrGLSLFPFragmentBuilder* fragBuilder = args.fFragBuilder;
fragBuilder->codeAppendf("%s = half4(1);", args.fOutputColor);
fragBuilder->codeAppendf("%s = half4(1);", args.fOutputCoverage);
}
void setData(const GrGLSLProgramDataManager& pdman,
const GrPrimitiveProcessor& primProc,
FPCoordTransformIter&&) override {}
};
return new GLSLGP();
}
void getGLSLProcessorKey(const GrShaderCaps&,
GrProcessorKeyBuilder* builder) const override {
builder->add32(fNumAttribs);
}
private:
int fNumAttribs;
std::unique_ptr<SkString[]> fAttribNames;
std::unique_ptr<Attribute[]> fAttributes;
typedef GrGeometryProcessor INHERITED;
};
sk_sp<GrGeometryProcessor> gp(new GP(fNumAttribs));
size_t vertexStride = gp->vertexStride();
QuadHelper helper(target, vertexStride, 1);
SkPoint* vertices = reinterpret_cast<SkPoint*>(helper.vertices());
SkPointPriv::SetRectTriStrip(vertices, 0.f, 0.f, 1.f, 1.f, vertexStride);
helper.recordDraw(target, std::move(gp));
}
void onExecute(GrOpFlushState* flushState, const SkRect& chainBounds) override {
flushState->executeDrawsAndUploadsForMeshDrawOp(
this, chainBounds, GrProcessorSet::MakeEmptySet());
}
int fNumAttribs;
typedef GrMeshDrawOp INHERITED;
};
}
DEF_GPUTEST_FOR_ALL_CONTEXTS(VertexAttributeCount, reporter, ctxInfo) {
GrContext* context = ctxInfo.grContext();
#if GR_GPU_STATS
GrGpu* gpu = context->priv().getGpu();
#endif
auto renderTargetContext = context->priv().makeDeferredRenderTargetContext(
SkBackingFit::kApprox, 1, 1, GrColorType::kRGBA_8888, nullptr);
if (!renderTargetContext) {
ERRORF(reporter, "Could not create render target context.");
return;
}
int attribCnt = context->priv().caps()->maxVertexAttributes();
if (!attribCnt) {
ERRORF(reporter, "No attributes allowed?!");
return;
}
context->flush();
context->priv().resetGpuStats();
#if GR_GPU_STATS
REPORTER_ASSERT(reporter, gpu->stats()->numDraws() == 0);
REPORTER_ASSERT(reporter, gpu->stats()->numFailedDraws() == 0);
#endif
// Adding discard to appease vulkan validation warning about loading uninitialized data on draw
renderTargetContext->discard();
GrPaint grPaint;
// This one should succeed.
renderTargetContext->priv().testingOnly_addDrawOp(Op::Make(context, attribCnt));
context->flush();
#if GR_GPU_STATS
REPORTER_ASSERT(reporter, gpu->stats()->numDraws() == 1);
REPORTER_ASSERT(reporter, gpu->stats()->numFailedDraws() == 0);
#endif
context->priv().resetGpuStats();
renderTargetContext->priv().testingOnly_addDrawOp(Op::Make(context, attribCnt + 1));
context->flush();
#if GR_GPU_STATS
REPORTER_ASSERT(reporter, gpu->stats()->numDraws() == 0);
REPORTER_ASSERT(reporter, gpu->stats()->numFailedDraws() == 1);
#endif
}