libwebp: update to 0.5.0

third_party/libwebp/dsp/dec_sse41.c

Index: third_party/libwebp/dsp/dec_sse41.c
diff --git a/third_party/libwebp/dsp/dec_sse41.c b/third_party/libwebp/dsp/dec_sse41.c
new file mode 100644
index 0000000000000000000000000000000000000000..224c6f895416030b21006e157d01fdfae913da9d
--- /dev/null
+++ b/third_party/libwebp/dsp/dec_sse41.c
@@ -0,0 +1,45 @@
+// Copyright 2015 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// SSE4 version of some decoding functions.
+//
+// Author: Skal (pascal.massimino@gmail.com)
+
+#include "./dsp.h"
+
+#if defined(WEBP_USE_SSE41)
+
+#include <smmintrin.h>
+#include "../dec/vp8i.h"
+
+static void HE16(uint8_t* dst) {     // horizontal
+  int j;
+  const __m128i kShuffle3 = _mm_set1_epi8(3);
+  for (j = 16; j > 0; --j) {
+    const __m128i in = _mm_cvtsi32_si128(WebPMemToUint32(dst - 4));
+    const __m128i values = _mm_shuffle_epi8(in, kShuffle3);
+    _mm_storeu_si128((__m128i*)dst, values);
+    dst += BPS;
+  }
+}
+
+//------------------------------------------------------------------------------
+// Entry point
+
+extern void VP8DspInitSSE41(void);
+
+WEBP_TSAN_IGNORE_FUNCTION void VP8DspInitSSE41(void) {
+  VP8PredLuma16[3] = HE16;
+}
+
+#else  // !WEBP_USE_SSE41
+
+WEBP_DSP_INIT_STUB(VP8DspInitSSE41)
+
+#endif  // WEBP_USE_SSE41