1 ;*****************************************************************************
\r
2 ;* dct-a.asm: h264 encoder library
\r
3 ;*****************************************************************************
\r
4 ;* Copyright (C) 2003-2008 x264 project
\r
6 ;* Authors: Laurent Aimar <fenrir@via.ecp.fr>
\r
7 ;* Loren Merritt <lorenm@u.washington.edu>
\r
8 ;* Holger Lubitz <hal@duncan.ol.sub.de>
\r
9 ;* Min Chen <chenm001.163.com>
\r
11 ;* This program is free software; you can redistribute it and/or modify
\r
12 ;* it under the terms of the GNU General Public License as published by
\r
13 ;* the Free Software Foundation; either version 2 of the License, or
\r
14 ;* (at your option) any later version.
\r
16 ;* This program is distributed in the hope that it will be useful,
\r
17 ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
\r
18 ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
\r
19 ;* GNU General Public License for more details.
\r
21 ;* You should have received a copy of the GNU General Public License
\r
22 ;* along with this program; if not, write to the Free Software
\r
23 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
\r
24 ;*****************************************************************************
\r
26 %include "x86inc.asm"
\r
27 %include "x86util.asm"
\r
30 pw_32: times 8 dw 32
\r
36 SUMSUBD2_AB m%2, m%4, m%6, m%5
\r
37 SUMSUB_BADC m%2, m%3, m%5, m%1
\r
38 SWAP %1, %2, %5, %4, %3
\r
42 cglobal x264_add8x4_idct_sse2, 3,3
\r
51 IDCT4_1D 0,1,2,3,4,5
\r
52 TRANSPOSE2x4x4W 0,1,2,3,4
\r
53 paddw m0, [pw_32 GLOBAL]
\r
54 IDCT4_1D 0,1,2,3,4,5
\r
56 STORE_DIFF m0, m4, m7, [r0]
\r
57 STORE_DIFF m1, m4, m7, [r0+r2]
\r
59 STORE_DIFF m2, m4, m7, [r0]
\r
60 STORE_DIFF m3, m4, m7, [r0+r2]
\r