Remove bloat from SkBlend_opts.

src/opts/SkBlend_opts.h

Index: src/opts/SkBlend_opts.h
diff --git a/src/opts/SkBlend_opts.h b/src/opts/SkBlend_opts.h
index 2dcdcc7aaf1cd65f84b67f7742c1a5d1830347b1..39cbbe8d7efcc7516d73e444e11e3860075f76c6 100644
--- a/src/opts/SkBlend_opts.h
+++ b/src/opts/SkBlend_opts.h
@@ -21,95 +21,19 @@ ninja -C out/Release dm nanobench ; and ./out/Release/dm --match Blend_opts ; an
 
 namespace SK_OPTS_NS {
 
-// An implementation of SrcOver from bytes to bytes in linear space that takes advantage of the
-// observation that the 255's cancel.
-//    invA = 1 - (As / 255);
-//
-//    R = 255 * sqrt((Rs/255)^2 + (Rd/255)^2 * invA)
-// => R = 255 * sqrt((Rs^2 + Rd^2 * invA)/255^2)
-// => R = sqrt(Rs^2 + Rd^2 * invA)
-static inline void blend_srgb_srgb_1(uint32_t* dst, const uint32_t pixel) {
-    Sk4f s = srgb_to_linear(to_4f(pixel));
-    Sk4f d = srgb_to_linear(to_4f(*dst));
-    Sk4f invAlpha = 1.0f - Sk4f{s[SkPM4f::A]} * (1.0f / 255.0f);
-    Sk4f r = linear_to_srgb(s + d * invAlpha) + 0.5f;
-    *dst = to_4b(r);
-}
-
-static inline void srcover_srgb_srgb_1(uint32_t* dst, const uint32_t pixel) {
+static inline void srcover_srgb8888_srgb_1(uint32_t* dst, const uint32_t pixel) {
     if ((~pixel & 0xFF000000) == 0) {
         *dst = pixel;
     } else if ((pixel & 0xFF000000) != 0) {
-        blend_srgb_srgb_1(dst, pixel);
+        srcover_blend_srgb8888_srgb_1(dst, srgb_to_linear(to_4f(pixel)));
     }
 }
 
-static inline void srcover_srgb_srgb_2(uint32_t* dst, const uint32_t* src) {
-    srcover_srgb_srgb_1(dst++, *src++);
-    srcover_srgb_srgb_1(dst, *src);
-}
-
 static inline void srcover_srgb_srgb_4(uint32_t* dst, const uint32_t* src) {
-    srcover_srgb_srgb_1(dst++, *src++);
-    srcover_srgb_srgb_1(dst++, *src++);
-    srcover_srgb_srgb_1(dst++, *src++);
-    srcover_srgb_srgb_1(dst, *src);
-}
-
-void best_non_simd_srcover_srgb_srgb(
-    uint32_t* dst, const uint32_t* const src, int ndst, const int nsrc) {
-    uint64_t* ddst = reinterpret_cast<uint64_t*>(dst);
-
-    while (ndst >0) {
-        int count = SkTMin(ndst, nsrc);
-        ndst -= count;
-        const uint64_t* dsrc = reinterpret_cast<const uint64_t*>(src);
-        const uint64_t* end = dsrc + (count >> 1);
-        do {
-            if ((~*dsrc & 0xFF000000FF000000) == 0) {
-                do {
-                    *ddst++ = *dsrc++;
-                } while (dsrc < end && (~*dsrc & 0xFF000000FF000000) == 0);
-            } else if ((*dsrc & 0xFF000000FF000000) == 0) {
-                do {
-                    dsrc++;
-                    ddst++;
-                } while (dsrc < end && (*dsrc & 0xFF000000FF000000) == 0);
-            } else {
-                srcover_srgb_srgb_2(reinterpret_cast<uint32_t*>(ddst++),
-                                    reinterpret_cast<const uint32_t*>(dsrc++));
-            }
-        } while (dsrc < end);
-
-        if ((count & 1) != 0) {
-            srcover_srgb_srgb_1(reinterpret_cast<uint32_t*>(ddst),
-                                *reinterpret_cast<const uint32_t*>(dsrc));
-        }
-    }
-}
-
-void brute_force_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++) {
-            blend_srgb_srgb_1(dst++, src[i]);
-        }
-        ndst -= n;
-    }
-}
-
-void trivial_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;
-    }
+    srcover_srgb8888_srgb_1(dst++, *src++);
+    srcover_srgb8888_srgb_1(dst++, *src++);
+    srcover_srgb8888_srgb_1(dst++, *src++);
+    srcover_srgb8888_srgb_1(dst, *src);
 }
 
 #if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
@@ -163,7 +87,7 @@ void trivial_srcover_srgb_srgb(
 
                 count = count & 3;
                 while (count-- > 0) {
-                    srcover_srgb_srgb_1(dst++, *src++);
+                    srcover_srgb8888_srgb_1(dst++, *src++);
                 }
             }
         }
@@ -235,7 +159,7 @@ void trivial_srcover_srgb_srgb(
 
                 count = count & 3;
                 while (count-- > 0) {
-                    srcover_srgb_srgb_1(dst++, *src++);
+                    srcover_srgb8888_srgb_1(dst++, *src++);
                 }
             }
         }
@@ -244,7 +168,14 @@ void trivial_srcover_srgb_srgb(
 
     void srcover_srgb_srgb(
         uint32_t* dst, const uint32_t* const src, int ndst, const int nsrc) {
-        trivial_srcover_srgb_srgb(dst, src, ndst, nsrc);
+        while (ndst > 0) {
+            int n = SkTMin(ndst, nsrc);
+
+            for (int i = 0; i < n; i++) {
+                srcover_srgb8888_srgb_1(dst++, src[i]);
+            }
+            ndst -= n;
+        }
     }
 
 #endif