1 | // Copyright 2015 Google Inc. All Rights Reserved. |
2 | // |
3 | // Use of this source code is governed by a BSD-style license |
4 | // that can be found in the COPYING file in the root of the source |
5 | // tree. An additional intellectual property rights grant can be found |
6 | // in the file PATENTS. All contributing project authors may |
7 | // be found in the AUTHORS file in the root of the source tree. |
8 | // ----------------------------------------------------------------------------- |
9 | // |
10 | // SSE4 version of some decoding functions. |
11 | // |
12 | // Author: Skal (pascal.massimino@gmail.com) |
13 | |
14 | #include "src/dsp/dsp.h" |
15 | |
16 | #if defined(WEBP_USE_SSE41) |
17 | |
18 | #include <smmintrin.h> |
19 | #include "src/dec/vp8i_dec.h" |
20 | #include "src/utils/utils.h" |
21 | |
22 | static void HE16_SSE41(uint8_t* dst) { // horizontal |
23 | int j; |
24 | const __m128i kShuffle3 = _mm_set1_epi8(3); |
25 | for (j = 16; j > 0; --j) { |
26 | const __m128i in = _mm_cvtsi32_si128(WebPMemToUint32(dst - 4)); |
27 | const __m128i values = _mm_shuffle_epi8(in, kShuffle3); |
28 | _mm_storeu_si128((__m128i*)dst, values); |
29 | dst += BPS; |
30 | } |
31 | } |
32 | |
33 | //------------------------------------------------------------------------------ |
34 | // Entry point |
35 | |
36 | extern void VP8DspInitSSE41(void); |
37 | |
38 | WEBP_TSAN_IGNORE_FUNCTION void VP8DspInitSSE41(void) { |
39 | VP8PredLuma16[3] = HE16_SSE41; |
40 | } |
41 | |
42 | #else // !WEBP_USE_SSE41 |
43 | |
44 | WEBP_DSP_INIT_STUB(VP8DspInitSSE41) |
45 | |
46 | #endif // WEBP_USE_SSE41 |
47 | |