Planar ramp() for S32 linear gradients

src/effects/gradients/Sk4fGradientPriv.h

Index: src/effects/gradients/Sk4fGradientPriv.h
diff --git a/src/effects/gradients/Sk4fGradientPriv.h b/src/effects/gradients/Sk4fGradientPriv.h
index 4f71d3f14df6929d403872c1c22f7a014090ca42..ae6fe7ce466acd7197aabddfc5cf2fc50273a02e 100644
--- a/src/effects/gradients/Sk4fGradientPriv.h
+++ b/src/effects/gradients/Sk4fGradientPriv.h
@@ -108,19 +108,19 @@ struct DstTraits<DstType::S32, premul> {
     using PM   = PremulTraits<premul>;
     using Type = SkPMColor;
 
-    // TODO: prescale by something like (255^2, 255^2, 255^2, 255) and use
-    //       linear_to_srgb to save a mult in store?
     static Sk4f load(const SkPM4f& c) {
-        return c.to4f_pmorder();
+        // Prescaling by (255^2, 255^2, 255^2, 255) on load, to avoid a 255 multiply on
+        // each store (S32 conversion yields a uniform 255 factor).
+        return c.to4f_pmorder() * Sk4f(255 * 255, 255 * 255, 255 * 255, 255);
     }
 
     static void store(const Sk4f& c, Type* dst) {
         // FIXME: this assumes opaque colors.  Handle unpremultiplication.
-        *dst = Sk4f_toS32(PM::apply(c));
+        *dst = to_4b(linear_to_srgb(PM::apply(c)));
     }
 
     static void store(const Sk4f& c, Type* dst, int n) {
-        sk_memset32(dst, Sk4f_toS32(PM::apply(c)), n);
+        sk_memset32(dst, to_4b(linear_to_srgb(PM::apply(c))), n);
     }
 
     static void store4x(const Sk4f& c0, const Sk4f& c1,