1 ;*****************************************************************************
2 ;* x86-optimized functions for transpose filter
4 ;* Copyright (C) 2019 Paul B Mahol
6 ;* This file is part of FFmpeg.
8 ;* FFmpeg is free software; you can redistribute it and/or
9 ;* modify it under the terms of the GNU Lesser General Public
10 ;* License as published by the Free Software Foundation; either
11 ;* version 2.1 of the License, or (at your option) any later version.
13 ;* FFmpeg 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 GNU
16 ;* Lesser General Public License for more details.
18 ;* You should have received a copy of the GNU Lesser General Public
19 ;* License along with FFmpeg; if not, write to the Free Software
20 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 ;******************************************************************************
23 %include "libavutil/x86/x86util.asm"
27 ;------------------------------------------------------------------------------
28 ; void ff_transpose_8x8(uint8_t *src, ptrdiff_t src_linesize,
29 ; uint8_t *dst, ptrdiff_t dst_linesize)
30 ;------------------------------------------------------------------------------
33 cglobal transpose_8x8_8, 4,5,8, src, src_linesize, dst, dst_linesize, linesize3
34 lea linesize3q, [src_linesizeq * 3]
35 movq m0, [srcq + src_linesizeq * 0]
36 movq m1, [srcq + src_linesizeq * 1]
37 movq m2, [srcq + src_linesizeq * 2]
38 movq m3, [srcq + linesize3q]
39 lea srcq, [srcq + src_linesizeq * 4]
40 movq m4, [srcq + src_linesizeq * 0]
41 movq m5, [srcq + src_linesizeq * 1]
42 movq m6, [srcq + src_linesizeq * 2]
43 movq m7, [srcq + linesize3q]
45 TRANSPOSE_8X8B 0, 1, 2, 3, 4, 5, 6, 7
47 lea linesize3q, [dst_linesizeq * 3]
48 movq [dstq + dst_linesizeq * 0], m0
49 movq [dstq + dst_linesizeq * 1], m1
50 movq [dstq + dst_linesizeq * 2], m2
51 movq [dstq + linesize3q], m3
52 lea dstq, [dstq + dst_linesizeq * 4]
53 movq [dstq + dst_linesizeq * 0], m4
54 movq [dstq + dst_linesizeq * 1], m5
55 movq [dstq + dst_linesizeq * 2], m6
56 movq [dstq + linesize3q], m7
59 cglobal transpose_8x8_16, 4,5,9, ARCH_X86_32 * 32, src, src_linesize, dst, dst_linesize, linesize3
60 lea linesize3q, [src_linesizeq * 3]
61 movu m0, [srcq + src_linesizeq * 0]
62 movu m1, [srcq + src_linesizeq * 1]
63 movu m2, [srcq + src_linesizeq * 2]
64 movu m3, [srcq + linesize3q]
65 lea srcq, [srcq + src_linesizeq * 4]
66 movu m4, [srcq + src_linesizeq * 0]
67 movu m5, [srcq + src_linesizeq * 1]
68 movu m6, [srcq + src_linesizeq * 2]
69 movu m7, [srcq + linesize3q]
72 TRANSPOSE8x8W 0, 1, 2, 3, 4, 5, 6, 7, 8
74 TRANSPOSE8x8W 0, 1, 2, 3, 4, 5, 6, 7, [rsp], [rsp + 16]
77 lea linesize3q, [dst_linesizeq * 3]
78 movu [dstq + dst_linesizeq * 0], m0
79 movu [dstq + dst_linesizeq * 1], m1
80 movu [dstq + dst_linesizeq * 2], m2
81 movu [dstq + linesize3q], m3
82 lea dstq, [dstq + dst_linesizeq * 4]
83 movu [dstq + dst_linesizeq * 0], m4
84 movu [dstq + dst_linesizeq * 1], m5
85 movu [dstq + dst_linesizeq * 2], m6
86 movu [dstq + linesize3q], m7