Add a hook for CPU-optimized sRGB-sRGB srcover.

Herb's really starting to get serious about tweaking this, which becomes
a lot easier when you've got SkOpts' runtime CPU detection. We should be
able to optimize this usefully for SSSE3, SSE4.1, AVX, AVX2, or NEON.
(We can of course implement a subset.)

This function takes two counts to give us flexibility to write src patterns:
   nsrc >= ndst -> the usual srcover function
   nsrc <  ndst -> repeat src until it fills dst
   nsrc << ndst -> possibly preprocess src into registers
   nsrc == 1    -> equivalent of blitrow_color32, srcover_1, etc.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1939783003

Review-Url: https://codereview.chromium.org/1939783003

src/core/SkOpts.cpp

@@ -11,6 +11,7 @@
 #include "SkOpts.h"
 
 #define SK_OPTS_NS sk_default
+#include "SkBlend_opts.h"
 #include "SkBlitMask_opts.h"
 #include "SkBlitRow_opts.h"
 #include "SkBlurImageFilter_opts.h"
@@ -73,6 +74,8 @@ namespace SkOpts {
     decltype(half_to_float) half_to_float = sk_default::half_to_float;
     decltype(float_to_half) float_to_half = sk_default::float_to_half;
 
+    decltype(srcover_srgb_srgb) srcover_srgb_srgb = sk_default::srcover_srgb_srgb;
+
     // Each Init_foo() is defined in src/opts/SkOpts_foo.cpp.
     void Init_ssse3();
     void Init_sse41();

src/core/SkOpts.h

@@ -64,6 +64,10 @@ namespace SkOpts {
 
     extern void (*half_to_float)(float[], const uint16_t[], int);
     extern void (*float_to_half)(uint16_t[], const float[], int);
+
+    // Blend ndst src pixels over dst, where both src and dst point to sRGB pixels (RGBA or BGRA).
+    // If nsrc < ndst, we loop over src to create a pattern.
+    extern void (*srcover_srgb_srgb)(uint32_t* dst, const uint32_t* src, int ndst, int nsrc);
 }
 
 #endif//SkOpts_DEFINED

src/opts/SkBlend_opts.h

@@ -0,0 +1,58 @@
+/*
+ * Copyright 2016 Google Inc.
+ *
+ * Use of this source code is governed by a BSD-style license that can be
+ * found in the LICENSE file.
+ */
+
+#ifndef SkBlend_opts_DEFINED
+#define SkBlend_opts_DEFINED
+
+namespace SK_OPTS_NS {
+
+#if 0
+
+#else
+
+    static inline void srcover_srgb_srgb_1(uint32_t* dst, uint32_t src) {
+        switch (src >> 24) {
+            case 0x00:             return; 
+            case 0xff: *dst = src; return;
+        }
+
+        Sk4f d = SkNx_cast<float>(Sk4b::Load( dst)),
+             s = SkNx_cast<float>(Sk4b::Load(&src));
+
+        // Approximate sRGB gamma as 2.0.
+        Sk4f d_sq = d*d,
+             s_sq = s*s;
+        d = Sk4f{d_sq[0], d_sq[1], d_sq[2], d[3]};
+        s = Sk4f{s_sq[0], s_sq[1], s_sq[2], s[3]};
+
+        // SrcOver.
+        Sk4f invA = 1.0f - s[3]*(1/255.0f);
+        d = s + d * invA;
+
+        // Re-apply approximate sRGB gamma.
+        Sk4f d_sqrt = d.sqrt();
+        d = Sk4f{d_sqrt[0], d_sqrt[1], d_sqrt[2], d[3]};
+
+        SkNx_cast<uint8_t>(d).store(dst);
+    }
+
+    static inline void srcover_srgb_srgb(uint32_t* dst, const uint32_t* const src, int ndst, const int nsrc) {
+        while (ndst > 0) {
+            int n = SkTMin(ndst, nsrc);
+
+            for (int i = 0; i < n; i++) {
+                srcover_srgb_srgb_1(dst++, src[i]);
+            }
+            ndst -= n;
+        }
+    }
+    
+#endif
+
+}  // namespace SK_OPTS_NS
+
+#endif//SkBlend_opts_DEFINED