From b94cee98fbad26c421ca79082b3b5db9b4580fea Mon Sep 17 00:00:00 2001
From: James Zern <jzern@google.com>
Date: Mon, 20 Nov 2017 20:34:05 -0800
Subject: [PATCH] dec_sse2: remove HE8uv_SSE2

with gcc-4.8, clang-4.0.1/5 this is no faster (actually up to 2x slower)
than the code generated for memset (0x01010... * dst[-1]). shuffles in
sse4 recover a bit, but performance is still down.

Change-Id: Ie85e8353f8ede559d0b05a1d388787fd18ecc80f
---
 src/dsp/dec_sse2.c | 10 ----------
 1 file changed, 10 deletions(-)

diff --git a/src/dsp/dec_sse2.c b/src/dsp/dec_sse2.c
index c96e8e96..b3840faf 100644
--- a/src/dsp/dec_sse2.c
+++ b/src/dsp/dec_sse2.c
@@ -1127,15 +1127,6 @@ static void VE8uv_SSE2(uint8_t* dst) {    // vertical
   }
 }
 
-static void HE8uv_SSE2(uint8_t* dst) {    // horizontal
-  int j;
-  for (j = 0; j < 8; ++j) {
-    const __m128i values = _mm_set1_epi8(dst[-1]);
-    _mm_storel_epi64((__m128i*)dst, values);
-    dst += BPS;
-  }
-}
-
 // helper for chroma-DC predictions
 static WEBP_INLINE void Put8x8uv_SSE2(uint8_t v, uint8_t* dst) {
   int j;
@@ -1224,7 +1215,6 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8DspInitSSE2(void) {
   VP8PredChroma8[0] = DC8uv_SSE2;
   VP8PredChroma8[1] = TM8uv_SSE2;
   VP8PredChroma8[2] = VE8uv_SSE2;
-  VP8PredChroma8[3] = HE8uv_SSE2;
   VP8PredChroma8[4] = DC8uvNoTop_SSE2;
   VP8PredChroma8[5] = DC8uvNoLeft_SSE2;
   VP8PredChroma8[6] = DC8uvNoTopLeft_SSE2;