Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
Download
52868 views
Tweet
Share
Share
1
/*****************************************************************************

2
 * bitstream-a.S: aarch64 bitstream functions

3
 *****************************************************************************

4
 * Copyright (C) 2014-2016 x264 project

5
 *

6
 * Authors: Janne Grunau <[email protected]>

7
 *

8
 * This program is free software; you can redistribute it and/or modify

9
 * it under the terms of the GNU General Public License as published by

10
 * the Free Software Foundation; either version 2 of the License, or

11
 * (at your option) any later version.

12
 *

13
 * This program is distributed in the hope that it will be useful,

14
 * but WITHOUT ANY WARRANTY; without even the implied warranty of

15
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

16
 * GNU General Public License for more details.

17
 *

18
 * You should have received a copy of the GNU General Public License

19
 * along with this program; if not, write to the Free Software

20
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.

21
 *

22
 * This program is also available under a commercial proprietary license.

23
 * For more information, contact us at [email protected].

24
 *****************************************************************************/

25


26
#include "asm.S"

27


28
function x264_nal_escape_neon, export=1

29
    movi        v0.16b,  #0xff

30
    movi        v4.16b,  #4

31
    mov         w3,  #3

32
    subs        x6,  x1,  x2

33
    cbz         x6,  99f

34
0:

35
    cmn         x6,  #15

36
    b.lt        16f

37
    mov         x1,  x2

38
    b           100f

39
16:

40
    ld1         {v1.16b}, [x1], #16

41
    ext         v2.16b, v0.16b, v1.16b, #14

42
    ext         v3.16b, v0.16b, v1.16b, #15

43
    cmhi        v7.16b, v4.16b, v1.16b

44
    cmeq        v5.16b, v2.16b, #0

45
    cmeq        v6.16b, v3.16b, #0

46
    and         v5.16b, v5.16b, v7.16b

47
    and         v5.16b, v5.16b, v6.16b

48
    shrn        v7.8b,  v5.8h,  #4

49
    mov         x7,  v7.d[0]

50
    cbz         x7,  16f

51
    mov         x6,  #-16

52
100:

53
    umov        w5,  v0.b[14]

54
    umov        w4,  v0.b[15]

55
    orr         w5,  w4,  w5, lsl #8

56
101:

57
    ldrb        w4,  [x1, x6]

58
    orr         w9,  w4,  w5, lsl #16

59
    cmp         w9,  #3

60
    b.hi        102f

61
    strb        w3,  [x0], #1

62
    orr         w5,  w3,  w5, lsl #8

63
102:

64
    adds        x6,  x6,  #1

65
    strb        w4,  [x0], #1

66
    orr         w5,  w4,  w5, lsl #8

67
    b.lt        101b

68
    subs        x6,  x1,  x2

69
    lsr         w9,  w5,  #8

70
    mov         v0.b[14],  w9

71
    mov         v0.b[15],  w5

72
    b.lt        0b

73


74
    ret

75
16:

76
    subs        x6,  x1,  x2

77
    st1         {v1.16b}, [x0], #16

78
    mov         v0.16b, v1.16b

79
    b.lt        0b

80
99:

81
    ret

82
endfunc

83


84