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
13 #include "./vpx_dsp_rtcd.h"
14 #include "vpx_dsp/arm/idct_neon.h"
15 #include "vpx_dsp/inv_txfm.h"
16
idct32x32_1_add_pos_kernel(uint8_t ** dest,const int stride,const uint8x16_t res)17 static INLINE void idct32x32_1_add_pos_kernel(uint8_t **dest, const int stride,
18 const uint8x16_t res) {
19 const uint8x16_t a0 = vld1q_u8(*dest);
20 const uint8x16_t a1 = vld1q_u8(*dest + 16);
21 const uint8x16_t b0 = vqaddq_u8(a0, res);
22 const uint8x16_t b1 = vqaddq_u8(a1, res);
23 vst1q_u8(*dest, b0);
24 vst1q_u8(*dest + 16, b1);
25 *dest += stride;
26 }
27
idct32x32_1_add_neg_kernel(uint8_t ** dest,const int stride,const uint8x16_t res)28 static INLINE void idct32x32_1_add_neg_kernel(uint8_t **dest, const int stride,
29 const uint8x16_t res) {
30 const uint8x16_t a0 = vld1q_u8(*dest);
31 const uint8x16_t a1 = vld1q_u8(*dest + 16);
32 const uint8x16_t b0 = vqsubq_u8(a0, res);
33 const uint8x16_t b1 = vqsubq_u8(a1, res);
34 vst1q_u8(*dest, b0);
35 vst1q_u8(*dest + 16, b1);
36 *dest += stride;
37 }
38
vpx_idct32x32_1_add_neon(const tran_low_t * input,uint8_t * dest,int stride)39 void vpx_idct32x32_1_add_neon(const tran_low_t *input, uint8_t *dest,
40 int stride) {
41 int i;
42 const int16_t out0 =
43 WRAPLOW(dct_const_round_shift((int16_t)input[0] * cospi_16_64));
44 const int16_t out1 = WRAPLOW(dct_const_round_shift(out0 * cospi_16_64));
45 const int16_t a1 = ROUND_POWER_OF_TWO(out1, 6);
46
47 if (a1 >= 0) {
48 const uint8x16_t dc = create_dcq(a1);
49 for (i = 0; i < 32; i++) {
50 idct32x32_1_add_pos_kernel(&dest, stride, dc);
51 }
52 } else {
53 const uint8x16_t dc = create_dcq(-a1);
54 for (i = 0; i < 32; i++) {
55 idct32x32_1_add_neg_kernel(&dest, stride, dc);
56 }
57 }
58 }
59