| OLD | NEW |
|---|
| (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 "vp8/encoder/block.h" |
| 13 |
| 14 void vp8_subtract_b_neon( |
| 15 BLOCK *be, |
| 16 BLOCKD *bd, |
| 17 int pitch) { |
| 18 unsigned char *src_ptr, *predictor; |
| 19 int src_stride; |
| 20 int16_t *src_diff; |
| 21 uint8x8_t d0u8, d1u8, d2u8, d3u8, d4u8, d5u8, d6u8, d7u8; |
| 22 uint16x8_t q10u16, q11u16, q12u16, q13u16; |
| 23 |
| 24 src_ptr = *be->base_src + be->src; |
| 25 src_stride = be->src_stride; |
| 26 predictor = bd->predictor; |
| 27 |
| 28 d0u8 = vld1_u8(src_ptr); |
| 29 src_ptr += src_stride; |
| 30 d2u8 = vld1_u8(src_ptr); |
| 31 src_ptr += src_stride; |
| 32 d4u8 = vld1_u8(src_ptr); |
| 33 src_ptr += src_stride; |
| 34 d6u8 = vld1_u8(src_ptr); |
| 35 |
| 36 d1u8 = vld1_u8(predictor); |
| 37 predictor += pitch; |
| 38 d3u8 = vld1_u8(predictor); |
| 39 predictor += pitch; |
| 40 d5u8 = vld1_u8(predictor); |
| 41 predictor += pitch; |
| 42 d7u8 = vld1_u8(predictor); |
| 43 |
| 44 q10u16 = vsubl_u8(d0u8, d1u8); |
| 45 q11u16 = vsubl_u8(d2u8, d3u8); |
| 46 q12u16 = vsubl_u8(d4u8, d5u8); |
| 47 q13u16 = vsubl_u8(d6u8, d7u8); |
| 48 |
| 49 src_diff = be->src_diff; |
| 50 vst1_u16((uint16_t *)src_diff, vget_low_u16(q10u16)); |
| 51 src_diff += pitch; |
| 52 vst1_u16((uint16_t *)src_diff, vget_low_u16(q11u16)); |
| 53 src_diff += pitch; |
| 54 vst1_u16((uint16_t *)src_diff, vget_low_u16(q12u16)); |
| 55 src_diff += pitch; |
| 56 vst1_u16((uint16_t *)src_diff, vget_low_u16(q13u16)); |
| 57 return; |
| 58 } |
| 59 |
| 60 void vp8_subtract_mby_neon( |
| 61 int16_t *diff, |
| 62 unsigned char *src, |
| 63 int src_stride, |
| 64 unsigned char *pred, |
| 65 int pred_stride) { |
| 66 int i; |
| 67 uint8x16_t q0u8, q1u8, q2u8, q3u8; |
| 68 uint16x8_t q8u16, q9u16, q10u16, q11u16; |
| 69 |
| 70 for (i = 0; i < 8; i++) { // subtract_mby_loop |
| 71 q0u8 = vld1q_u8(src); |
| 72 src += src_stride; |
| 73 q2u8 = vld1q_u8(src); |
| 74 src += src_stride; |
| 75 q1u8 = vld1q_u8(pred); |
| 76 pred += pred_stride; |
| 77 q3u8 = vld1q_u8(pred); |
| 78 pred += pred_stride; |
| 79 |
| 80 q8u16 = vsubl_u8(vget_low_u8(q0u8), vget_low_u8(q1u8)); |
| 81 q9u16 = vsubl_u8(vget_high_u8(q0u8), vget_high_u8(q1u8)); |
| 82 q10u16 = vsubl_u8(vget_low_u8(q2u8), vget_low_u8(q3u8)); |
| 83 q11u16 = vsubl_u8(vget_high_u8(q2u8), vget_high_u8(q3u8)); |
| 84 |
| 85 vst1q_u16((uint16_t *)diff, q8u16); |
| 86 diff += 8; |
| 87 vst1q_u16((uint16_t *)diff, q9u16); |
| 88 diff += 8; |
| 89 vst1q_u16((uint16_t *)diff, q10u16); |
| 90 diff += 8; |
| 91 vst1q_u16((uint16_t *)diff, q11u16); |
| 92 diff += 8; |
| 93 } |
| 94 return; |
| 95 } |
| 96 |
| 97 void vp8_subtract_mbuv_neon( |
| 98 int16_t *diff, |
| 99 unsigned char *usrc, |
| 100 unsigned char *vsrc, |
| 101 int src_stride, |
| 102 unsigned char *upred, |
| 103 unsigned char *vpred, |
| 104 int pred_stride) { |
| 105 int i, j; |
| 106 unsigned char *src_ptr, *pred_ptr; |
| 107 uint8x8_t d0u8, d1u8, d2u8, d3u8, d4u8, d5u8, d6u8, d7u8; |
| 108 uint16x8_t q8u16, q9u16, q10u16, q11u16; |
| 109 |
| 110 diff += 256; |
| 111 for (i = 0; i < 2; i++) { |
| 112 if (i == 0) { |
| 113 src_ptr = usrc; |
| 114 pred_ptr = upred; |
| 115 } else if (i == 1) { |
| 116 src_ptr = vsrc; |
| 117 pred_ptr = vpred; |
| 118 } |
| 119 |
| 120 for (j = 0; j < 2; j++) { |
| 121 d0u8 = vld1_u8(src_ptr); |
| 122 src_ptr += src_stride; |
| 123 d1u8 = vld1_u8(pred_ptr); |
| 124 pred_ptr += pred_stride; |
| 125 d2u8 = vld1_u8(src_ptr); |
| 126 src_ptr += src_stride; |
| 127 d3u8 = vld1_u8(pred_ptr); |
| 128 pred_ptr += pred_stride; |
| 129 d4u8 = vld1_u8(src_ptr); |
| 130 src_ptr += src_stride; |
| 131 d5u8 = vld1_u8(pred_ptr); |
| 132 pred_ptr += pred_stride; |
| 133 d6u8 = vld1_u8(src_ptr); |
| 134 src_ptr += src_stride; |
| 135 d7u8 = vld1_u8(pred_ptr); |
| 136 pred_ptr += pred_stride; |
| 137 |
| 138 q8u16 = vsubl_u8(d0u8, d1u8); |
| 139 q9u16 = vsubl_u8(d2u8, d3u8); |
| 140 q10u16 = vsubl_u8(d4u8, d5u8); |
| 141 q11u16 = vsubl_u8(d6u8, d7u8); |
| 142 |
| 143 vst1q_u16((uint16_t *)diff, q8u16); |
| 144 diff += 8; |
| 145 vst1q_u16((uint16_t *)diff, q9u16); |
| 146 diff += 8; |
| 147 vst1q_u16((uint16_t *)diff, q10u16); |
| 148 diff += 8; |
| 149 vst1q_u16((uint16_t *)diff, q11u16); |
| 150 diff += 8; |
| 151 } |
| 152 } |
| 153 return; |
| 154 } |
| OLD | NEW |
|---|
Chromium Code Reviews
Issue 484923003:
libvpx: Pull from upstream (Closed)
Base URL: svn://svn.chromium.org/chrome/trunk/deps/third_party/libvpx/