try to speed-up maprect + round2i + contains

We call roundOut in a few places. If we can get SkNx::Ceil we could efficiently implement that as well.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2133413002
CQ_INCLUDE_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review-Url: https://codereview.chromium.org/2133413002

1 files changed, 90 insertions, 0 deletions

diff --git a/bench/RectBench.cpp b/bench/RectBench.cpp
index 46a515de29..0f757a2b5f 100644
--- a/bench/RectBench.cpp
+++ b/bench/RectBench.cpp
@@ -302,3 +302,93 @@ DEF_BENCH(return new BlitMaskBench(SkCanvas::kPoints_PointMode,
 DEF_BENCH(return new BlitMaskBench(SkCanvas::kPoints_PointMode,
                                    BlitMaskBench::KMaskShader,
                                    "maskshader");)
+
+///////////////////////////////////////////////////////////////////////////////////////////////////
+
+typedef int (*RectMathProc)(const SkMatrix&, const SkRect[], const SkIRect[], int count);
+
+class RectMathBench : public Benchmark {
+    SkString fName;
+    SkRandom fRand;
+    SkString fSuffix;
+    RectMathProc fProc;
+
+public:
+    enum {
+        N = 300,
+        OUTER = 10000,
+    };
+    SkRect  fRects[N];
+    SkIRect fIRects[N];
+    volatile int fCounter;
+    
+    RectMathBench(RectMathProc proc, const char* suffix) {
+        fProc = proc;
+        fSuffix.set(suffix);
+        SkRandom rand;
+        for (int i = 0; i < N; ++i) {
+            fRects[i].setXYWH(rand.nextUScalar1() * 100, rand.nextUScalar1() * 100,
+                              rand.nextUScalar1() * 100, rand.nextUScalar1() * 100);
+            fIRects[i].setXYWH(i, i, 10, 10);
+        }
+    }
+    
+    bool isVisual() override { return false; }
+    
+protected:
+    const char* onGetName() override {
+        fName.printf("rect_math_%s", fSuffix.c_str());
+        return fName.c_str();
+    }
+    
+    void onDraw(int loops, SkCanvas* canvas) override {
+        SkMatrix mat;
+        for (int j = 0; j < OUTER; ++j) {
+            mat.setScaleTranslate(fRand.nextUScalar1(), fRand.nextUScalar1(),
+                                  fRand.nextUScalar1(), fRand.nextUScalar1());
+            fCounter += fProc(mat, fRects, fIRects, N);
+        }
+    }
+
+private:
+    typedef Benchmark INHERITED;
+};
+
+static int rectmath0(const SkMatrix& mat, const SkRect rr[], const SkIRect ir[], int count) {
+    int counter = 0;
+    for (int i = 0; i < count; ++i) {
+        SkRect dst;
+        mat.mapRectScaleTranslate(&dst, rr[i]);
+        counter += dst.round().contains(ir[i]);
+    }
+    return counter;
+}
+
+static int rectmath1(const SkMatrix& mat, const SkRect rr[], const SkIRect ir[], int count) {
+    int counter = 0;
+    for (int i = 0; i < count; ++i) {
+        SkRect dst;
+        mat.mapRectScaleTranslate(&dst, rr[i]);
+        counter += dst.round2i().contains(ir[i]);
+    }
+    return counter;
+}
+
+#if 0
+static bool contains(SkIRect outer, SkIRect inner) {
+    Sk4i le(outer.fLeft, outer.fTop, inner.fRight, inner.fBottom);
+    Sk4i ge(inner.fLeft, inner.fTop, outer.fRight, outer.fBottom);
+    return (le <= ge).allTrue();
+}
+
+static int rectmath3(const SkMatrix& mat, const SkRect rr[], const SkIRect ir[], int count) {
+    int counter = 0;
+    for (int i = 0; i < count; ++i) {
+        counter += contains(mat.mapRectScaleTranslate(rr[i]).round2i(), ir[i]);
+    }
+    return counter;
+}
+#endif
+
+DEF_BENCH(return new RectMathBench(rectmath0, "0");)
+DEF_BENCH(return new RectMathBench(rectmath1, "1");)