Add mulHi to SkNx

Add mulHi to base SkNx, and specialize implementations for Sk4u for
neon and sse.

Add casts for converting from uint8_t by 4 to uint32_t by 4.

Cq-Include-Trybots: skia.primary:Test-Debian9-Clang-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD
Change-Id: I29a32e2ad9812a47fff841ceca334e562362836f
Reviewed-on: https://skia-review.googlesource.com/57960
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Herb Derby <herb@google.com>

1 files changed, 10 insertions, 1 deletions

diff --git a/src/core/SkNx.h b/src/core/SkNx.h
index 65e3fcb8c8..6957cb0d38 100644
--- a/src/core/SkNx.h
+++ b/src/core/SkNx.h
@@ -119,10 +119,13 @@ struct SkNx {
     AI SkNx saturatedAdd(const SkNx& y) const {
         return { fLo.saturatedAdd(y.fLo), fHi.saturatedAdd(y.fHi) };
     }
+
+    AI SkNx mulHi(const SkNx& m) const {
+        return { fLo.mulHi(m.fLo), fHi.mulHi(m.fHi) };
+    }
     AI SkNx thenElse(const SkNx& t, const SkNx& e) const {
         return { fLo.thenElse(t.fLo, e.fLo), fHi.thenElse(t.fHi, e.fHi) };
     }
-
     AI static SkNx Min(const SkNx& x, const SkNx& y) {
         return { Half::Min(x.fLo, y.fLo), Half::Min(x.fHi, y.fHi) };
     }
@@ -214,6 +217,12 @@ struct SkNx<1,T> {
         return sum < fVal ? std::numeric_limits<T>::max() : sum;
     }
 
+    AI SkNx mulHi(const SkNx& m) const {
+        static_assert(std::is_unsigned<T>::value, "");
+        static_assert(sizeof(T) <= 4, "");
+        return static_cast<T>((static_cast<uint64_t>(fVal) * m.fVal) >> (sizeof(T)*8));
+    }
+
     AI SkNx thenElse(const SkNx& t, const SkNx& e) const { return fVal != 0 ? t : e; }
 
 private: