Add tests and benches to support the sRGB blitter for SkOpts

 1,370.85 LinearSrcOvericonstrip.pngVSkOptsSSE41
 2,359.69 LinearSrcOvericonstrip.pngVSkOptsDefault
 1,828.72 LinearSrcOvericonstrip.pngVSkOptsNonSimdCore
 3,277.40 LinearSrcOvericonstrip.pngVSkOptsTrivial
 9,862.34 LinearSrcOvericonstrip.pngVSkOptsBruteForce

   633.55 LinearSrcOvermandrill_512.pngVSkOptsSSE41
   684.29 LinearSrcOvermandrill_512.pngVSkOptsDefault
 1,201.88 LinearSrcOvermandrill_512.pngVSkOptsNonSimdCore
 2,382.63 LinearSrcOvermandrill_512.pngVSkOptsTrivial
10,888.74 LinearSrcOvermandrill_512.pngVSkOptsBruteForce

   209.14 LinearSrcOverplane.pngVSkOptsSSE41
   562.24 LinearSrcOverplane.pngVSkOptsDefault
   272.64 LinearSrcOverplane.pngVSkOptsNonSimdCore
   436.46 LinearSrcOverplane.pngVSkOptsTrivial
 1,327.23 LinearSrcOverplane.pngVSkOptsBruteForce

   318.01 LinearSrcOverbaby_tux.pngVSkOptsSSE41
   529.05 LinearSrcOverbaby_tux.pngVSkOptsDefault
   441.33 LinearSrcOverbaby_tux.pngVSkOptsNonSimdCore
   720.50 LinearSrcOverbaby_tux.pngVSkOptsTrivial
 2,191.10 LinearSrcOverbaby_tux.pngVSkOptsBruteForce

   479.68 LinearSrcOveryellow_rose.pngVSkOptsSSE41
 1,095.03 LinearSrcOveryellow_rose.pngVSkOptsDefault
   668.60 LinearSrcOveryellow_rose.pngVSkOptsNonSimdCore
 1,257.19 LinearSrcOveryellow_rose.pngVSkOptsTrivial
 4,970.25 LinearSrcOveryellow_rose.pngVSkOptsBruteForce

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1939513002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed: https://skia.googlesource.com/skia/+/554784cd85029c05d9ed04b1aeb71520d196153a

Review-Url: https://codereview.chromium.org/1939513002

1 files changed, 167 insertions, 0 deletions

diff --git a/bench/SkBlend_optsBench.cpp b/bench/SkBlend_optsBench.cpp
new file mode 100644
index 0000000000..ec276893ce
--- /dev/null
+++ b/bench/SkBlend_optsBench.cpp
@@ -0,0 +1,167 @@
+/*
+ * Copyright 2016 Google Inc.
+ *
+ * Use of this source code is governed by a BSD-style license that can be
+ * found in the LICENSE file.
+ */
+
+#include <tuple>
+
+#include "Benchmark.h"
+#include "Resources.h"
+#include "SkCpu.h"
+#include "SkImage.h"
+#include "SkImage_Base.h"
+#include "SkNx.h"
+#include "SkOpts.h"
+#include "SkString.h"
+
+#define INNER_LOOPS 10
+
+namespace sk_default {
+extern void brute_force_srcover_srgb_srgb(
+    uint32_t* dst, const uint32_t* const srcStart, int ndst, const int nsrc);
+}
+
+class SrcOverVSkOptsBruteForce {
+public:
+    static SkString Name() { return SkString{"VSkOptsBruteForce"}; }
+    static bool WorksOnCpu() { return true; }
+    static void BlendN(uint32_t* dst, int count, const uint32_t* src) {
+        sk_default::brute_force_srcover_srgb_srgb(dst, src, count, count);
+    }
+};
+
+namespace sk_default {
+extern void trivial_srcover_srgb_srgb(
+    uint32_t* dst, const uint32_t* const srcStart, int ndst, const int nsrc);
+}
+
+class SrcOverVSkOptsTrivial {
+public:
+    static SkString Name() { return SkString{"VSkOptsTrivial"}; }
+    static bool WorksOnCpu() { return true; }
+    static void BlendN(uint32_t* dst, int count, const uint32_t* src) {
+        sk_default::trivial_srcover_srgb_srgb(dst, src, count, count);
+    }
+};
+
+namespace sk_default {
+extern void best_non_simd_srcover_srgb_srgb(
+    uint32_t* dst, const uint32_t* const srcStart, int ndst, const int nsrc);
+}
+
+class SrcOverVSkOptsNonSimdCore {
+public:
+    static SkString Name() { return SkString{"VSkOptsNonSimdCore"}; }
+    static bool WorksOnCpu() { return true; }
+    static void BlendN(uint32_t* dst, int count, const uint32_t* src) {
+        sk_default::best_non_simd_srcover_srgb_srgb(dst, src, count, count);
+    }
+};
+
+namespace sk_default {
+extern void srcover_srgb_srgb(
+    uint32_t* dst, const uint32_t* const srcStart, int ndst, const int nsrc);
+}
+
+class SrcOverVSkOptsDefault {
+public:
+    static SkString Name() { return SkString{"VSkOptsDefault"}; }
+    static bool WorksOnCpu() { return true; }
+    static void BlendN(uint32_t* dst, int count, const uint32_t* src) {
+        sk_default::srcover_srgb_srgb(dst, src, count, count);
+    }
+};
+
+namespace sk_sse41 {
+   extern void srcover_srgb_srgb(
+       uint32_t* dst, const uint32_t* const srcStart, int ndst, const int nsrc);
+}
+
+class SrcOverVSkOptsSSE41 {
+public:
+    static SkString Name() { return SkString{"VSkOptsSSE41"}; }
+    static bool WorksOnCpu() { return SkCpu::Supports(SkCpu::SSE41); }
+    static void BlendN(uint32_t* dst, int count, const uint32_t* src) {
+        sk_sse41::srcover_srgb_srgb(dst, src, count, count);
+    }
+};
+
+///////////////////////////////////////////////////////////////////////////////////////////////////
+
+template <typename Blender>
+class LinearSrcOverBench : public Benchmark {
+public:
+    LinearSrcOverBench(const char* fileName) {
+        fName = "LinearSrcOver";
+        fName.append(fileName);
+        fName.append(Blender::Name());
+
+        sk_sp<SkImage> image = GetResourceAsImage(fileName);
+        SkBitmap bm;
+        if (!as_IB(image)->getROPixels(&bm)) {
+            SkFAIL("Could not read resource");
+        }
+        bm.peekPixels(&fPixmap);
+        fCount = fPixmap.rowBytesAsPixels();
+        fDst.reset(fCount);
+        memset(fDst.get(), 0, fPixmap.rowBytes());
+    }
+
+protected:
+    bool isSuitableFor(Backend backend) override {
+        return backend == kNonRendering_Backend && Blender::WorksOnCpu();
+    }
+    const char* onGetName() override { return fName.c_str(); }
+    void onDraw(int loops, SkCanvas*) override {
+        SkASSERT(fPixmap.colorType() == kRGBA_8888_SkColorType);
+
+        const int width = fPixmap.rowBytesAsPixels();
+
+        for (int i = 0; i < loops * INNER_LOOPS; ++i) {
+            const uint32_t* src = fPixmap.addr32();
+            for (int y = 0; y < fPixmap.height(); y++) {
+                Blender::BlendN(fDst.get(), width, src);
+                src += width;
+            }
+        }
+    }
+
+    void onPostDraw(SkCanvas*) override {
+        // Make sure the compiler does not optimize away the operation.
+        volatile uint32_t v = 0;
+        for (int i = 0; i < fCount; i++) {
+            v ^= fDst[i];
+        }
+    }
+
+private:
+    int fCount;
+    SkAutoTArray<uint32_t> fDst;
+    SkString fName;
+    SkPixmap fPixmap;
+
+    typedef Benchmark INHERITED;
+};
+
+#if defined(SK_CPU_X86) && !defined(SK_BUILD_FOR_IOS)
+#define BENCHES(fileName)                                                        \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsBruteForce>(fileName); )  \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsTrivial>(fileName); )     \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsNonSimdCore>(fileName); ) \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsDefault>(fileName); )     \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsSSE41>(fileName); )
+#else
+#define BENCHES(fileName)                                                        \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsBruteForce>(fileName); )  \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsTrivial>(fileName); )     \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsNonSimdCore>(fileName); ) \
+DEF_BENCH( return new LinearSrcOverBench<SrcOverVSkOptsDefault>(fileName); )
+#endif
+
+BENCHES("yellow_rose.png")
+BENCHES("baby_tux.png")
+BENCHES("plane.png")
+BENCHES("mandrill_512.png")
+BENCHES("iconstrip.png")