Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(498)

Side by Side Diff: source/libvpx/vp9/common/arm/neon/vp9_idct16x16_1_add_neon.c

Issue 812033011: libvpx: Pull from upstream (Closed) Base URL: svn://svn.chromium.org/chrome/trunk/deps/third_party/libvpx/
Patch Set: Created 5 years, 11 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch | Annotate | Revision Log
OLDNEW
(Empty)
1 /*
2 * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10
11 #include <arm_neon.h>
12 #include "vp9/common/vp9_idct.h"
13
14 void vp9_idct16x16_1_add_neon(
15 int16_t *input,
16 uint8_t *dest,
17 int dest_stride) {
18 uint8x8_t d2u8, d3u8, d30u8, d31u8;
19 uint64x1_t d2u64, d3u64, d4u64, d5u64;
20 uint16x8_t q0u16, q9u16, q10u16, q11u16, q12u16;
21 int16x8_t q0s16;
22 uint8_t *d1, *d2;
23 int16_t i, j, a1, cospi_16_64 = 11585;
24 int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
25 out = dct_const_round_shift(out * cospi_16_64);
26 a1 = ROUND_POWER_OF_TWO(out, 6);
27
28 q0s16 = vdupq_n_s16(a1);
29 q0u16 = vreinterpretq_u16_s16(q0s16);
30
31 for (d1 = d2 = dest, i = 0; i < 4; i++) {
32 for (j = 0; j < 2; j++) {
33 d2u64 = vld1_u64((const uint64_t *)d1);
34 d3u64 = vld1_u64((const uint64_t *)(d1 + 8));
35 d1 += dest_stride;
36 d4u64 = vld1_u64((const uint64_t *)d1);
37 d5u64 = vld1_u64((const uint64_t *)(d1 + 8));
38 d1 += dest_stride;
39
40 q9u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d2u64));
41 q10u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d3u64));
42 q11u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d4u64));
43 q12u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d5u64));
44
45 d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
46 d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
47 d30u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
48 d31u8 = vqmovun_s16(vreinterpretq_s16_u16(q12u16));
49
50 vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
51 vst1_u64((uint64_t *)(d2 + 8), vreinterpret_u64_u8(d3u8));
52 d2 += dest_stride;
53 vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d30u8));
54 vst1_u64((uint64_t *)(d2 + 8), vreinterpret_u64_u8(d31u8));
55 d2 += dest_stride;
56 }
57 }
58 return;
59 }
OLDNEW

Powered by Google App Engine
This is Rietveld 408576698