Rebase gemmlowp to ecae4d1 am: 7d0d5a611e am: 9fa88931b4android-wear-8.0.0_r2android-o-mr1-iot-release-smart-display-r9android-o-mr1-iot-release-smart-display-r8android-o-mr1-iot-release-smart-display-r5android-o-mr1-iot-release-smart-display-r40.1Jandroid-o-mr1-iot-release-smart-display-r4android-o-mr1-iot-release-smart-display-r39android-o-mr1-iot-release-smart-display-r30android-o-mr1-iot-release-smart-display-r3android-o-mr1-iot-release-smart-display-r22android-o-mr1-iot-release-smart-display-r14android-o-mr1-iot-release-smart-clock-r6android-o-mr1-iot-release-smart-clock-r2android-o-mr1-iot-release-smart-clock-fsiandroid-o-mr1-iot-release-smart-clock-fcsandroid-o-mr1-iot-release-cube_r2android-o-mr1-iot-release-cube-fsiandroid-o-mr1-iot-release-cube-fcsandroid-o-mr1-iot-release-1.0.5android-o-mr1-iot-release-1.0.4android-o-mr1-iot-release-1.0.3android-n-iot-release-ihome-igv1android-9.0.0_r47android-9.0.0_r46android-9.0.0_r45android-9.0.0_r44android-9.0.0_r43android-9.0.0_r42android-9.0.0_r41android-9.0.0_r40android-9.0.0_r39android-9.0.0_r38android-9.0.0_r37android-9.0.0_r36android-9.0.0_r35android-9.0.0_r34android-9.0.0_r33android-9.0.0_r32android-9.0.0_r31android-9.0.0_r30android-9.0.0_r22android-9.0.0_r21android-9.0.0_r20android-9.0.0_r19android-9.0.0_r16android-9.0.0_r12android-9.0.0_r11pie-qpr3-s1-releasepie-qpr3-releasepie-qpr3-b-releasepie-qpr2-releasepie-qpr1-s3-releasepie-qpr1-s2-releasepie-qpr1-s1-releasepie-qpr1-releasepie-dr1-releasepie-dr1-devpie-devpie-b4s4-releasepie-b4s4-devoreo-mr1-1.2-iot-releasenougat-iot-releasemaster-cuttlefish-testing-release

am: 1f4ec3258f

Change-Id: Icb9df1558e7d87c03080597ffbb5a6212817cba6

1 files changed, 26 insertions, 0 deletions

diff --git a/internal/simd_wrappers_sse.h b/internal/simd_wrappers_sse.h
index 6480b66..3b78cb4 100644
--- a/internal/simd_wrappers_sse.h
+++ b/internal/simd_wrappers_sse.h
@@ -22,6 +22,7 @@
 namespace gemmlowp {
 
 using Int32x4 = __m128i;
+using Int16x8 = __m128i;
 using Uint8x16 = __m128i;
 
 template <int ScalarCount>
@@ -31,6 +32,12 @@ struct RegisterType<std::int32_t, ScalarCount> {
 };
 
 template <int ScalarCount>
+struct RegisterType<std::int16_t, ScalarCount> {
+  using Type =
+      typename std::conditional<ScalarCount >= 8, Int16x8, std::int16_t>::type;
+};
+
+template <int ScalarCount>
 struct RegisterType<std::uint8_t, ScalarCount> {
   using Type = typename std::conditional<
       ScalarCount >= 16, Uint8x16,
@@ -42,10 +49,18 @@ inline Int32x4 LoadInt32x4(const std::int32_t* src) {
   return _mm_loadu_si128(reinterpret_cast<const Int32x4*>(src));
 }
 
+inline Int32x4 LoadInt16x8(const std::int16_t* src) {
+  return _mm_loadu_si128(reinterpret_cast<const Int16x8*>(src));
+}
+
 inline void StoreInt32x4(std::int32_t* dst, Int32x4 value) {
   _mm_storeu_si128(reinterpret_cast<__m128i*>(dst), value);
 }
 
+inline void StoreInt16x8(std::int16_t* dst, Int16x8 value) {
+  _mm_storeu_si128(reinterpret_cast<__m128i*>(dst), value);
+}
+
 inline Uint8x16 LoadUint8x16(const std::uint8_t* src) {
   return _mm_loadu_si128(reinterpret_cast<const Uint8x16*>(src));
 }
@@ -116,6 +131,17 @@ struct LoadContiguousImpl<RegBlockInt32<8, 8>> {
   }
 };
 
+template <>
+struct LoadContiguousImpl<RegBlockInt16<8, 8>> {
+  static RegBlockInt16<8, 8> Run(const std::int16_t* src) {
+    RegBlockInt16<8, 8> result;
+    for (int i = 0; i < 8; i++) {
+      result.buf.reg[i] = LoadInt16x8(src + 8 * i);
+    }
+    return result;
+  }
+};
+
 }  // end namespace gemmlowp
 
 #include "simd_wrappers_common_neon_sse.h"