1;
2;  Copyright (c) 2013 The WebM project authors. All Rights Reserved.
3;
4;  Use of this source code is governed by a BSD-style license and patent
5;  grant that can be found in the LICENSE file in the root of the source
6;  tree. All contributing project authors may be found in the AUTHORS
7;  file in the root of the source tree.
8;
9
10
11    EXPORT  |vpx_idct4x4_1_add_neon|
12    ARM
13    REQUIRE8
14    PRESERVE8
15
16    AREA ||.text||, CODE, READONLY, ALIGN=2
17
18;void vpx_idct4x4_1_add_neon(int16_t *input, uint8_t *dest,
19;                                  int dest_stride)
20;
21; r0  int16_t input
22; r1  uint8_t *dest
23; r2  int dest_stride)
24
25|vpx_idct4x4_1_add_neon| PROC
26    ldrsh            r0, [r0]
27
28    ; generate cospi_16_64 = 11585
29    mov              r12, #0x2d00
30    add              r12, #0x41
31
32    ; out = dct_const_round_shift(input[0] * cospi_16_64)
33    mul              r0, r0, r12               ; input[0] * cospi_16_64
34    add              r0, r0, #0x2000           ; +(1 << ((DCT_CONST_BITS) - 1))
35    asr              r0, r0, #14               ; >> DCT_CONST_BITS
36
37    ; out = dct_const_round_shift(out * cospi_16_64)
38    mul              r0, r0, r12               ; out * cospi_16_64
39    mov              r12, r1                   ; save dest
40    add              r0, r0, #0x2000           ; +(1 << ((DCT_CONST_BITS) - 1))
41    asr              r0, r0, #14               ; >> DCT_CONST_BITS
42
43    ; a1 = ROUND_POWER_OF_TWO(out, 4)
44    add              r0, r0, #8                ; + (1 <<((4) - 1))
45    asr              r0, r0, #4                ; >> 4
46
47    vdup.s16         q0, r0                    ; duplicate a1
48
49    vld1.32          {d2[0]}, [r1], r2
50    vld1.32          {d2[1]}, [r1], r2
51    vld1.32          {d4[0]}, [r1], r2
52    vld1.32          {d4[1]}, [r1]
53
54    vaddw.u8         q8, q0, d2                ; dest[x] + a1
55    vaddw.u8         q9, q0, d4
56
57    vqmovun.s16      d6, q8                    ; clip_pixel
58    vqmovun.s16      d7, q9
59
60    vst1.32          {d6[0]}, [r12], r2
61    vst1.32          {d6[1]}, [r12], r2
62    vst1.32          {d7[0]}, [r12], r2
63    vst1.32          {d7[1]}, [r12]
64
65    bx               lr
66    ENDP             ; |vpx_idct4x4_1_add_neon|
67
68    END
69