annotate arm/simple_idct_arm.S @ 8493:469f3e5bcf13 libavcodec

Bump minor number for many small improvements, tiff fax decoding, AVCodecContext refs is set during decoding.
author reimar
date Sat, 27 Dec 2008 09:22:21 +0000
parents 9281a8a9387a
children 04423b2f6e0b
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
2967
ef2149182f1c COSMETICS: Remove all trailing whitespace.
diego
parents: 1347
diff changeset
1 /*
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
2 * simple_idct_arm.S
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
3 * Copyright (C) 2002 Frederic 'dilb' Boulay.
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
4 *
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
5 * Author: Frederic Boulay <dilb@handhelds.org>
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
6 *
5214
470601203f44 Group all copyright and author notices together.
diego
parents: 3947
diff changeset
7 * The function defined in this file is derived from the simple_idct function
470601203f44 Group all copyright and author notices together.
diego
parents: 3947
diff changeset
8 * from the libavcodec library part of the FFmpeg project.
470601203f44 Group all copyright and author notices together.
diego
parents: 3947
diff changeset
9 *
3947
c8c591fe26f8 Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents: 3683
diff changeset
10 * This file is part of FFmpeg.
c8c591fe26f8 Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents: 3683
diff changeset
11 *
c8c591fe26f8 Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents: 3683
diff changeset
12 * FFmpeg is free software; you can redistribute it and/or
3683
dc1e28564bb2 Switch license from GPL to LGPL. The original author agreed to this as
diego
parents: 3036
diff changeset
13 * modify it under the terms of the GNU Lesser General Public
dc1e28564bb2 Switch license from GPL to LGPL. The original author agreed to this as
diego
parents: 3036
diff changeset
14 * License as published by the Free Software Foundation; either
3947
c8c591fe26f8 Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents: 3683
diff changeset
15 * version 2.1 of the License, or (at your option) any later version.
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
16 *
3947
c8c591fe26f8 Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents: 3683
diff changeset
17 * FFmpeg is distributed in the hope that it will be useful,
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
3683
dc1e28564bb2 Switch license from GPL to LGPL. The original author agreed to this as
diego
parents: 3036
diff changeset
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
dc1e28564bb2 Switch license from GPL to LGPL. The original author agreed to this as
diego
parents: 3036
diff changeset
20 * Lesser General Public License for more details.
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
21 *
3683
dc1e28564bb2 Switch license from GPL to LGPL. The original author agreed to this as
diego
parents: 3036
diff changeset
22 * You should have received a copy of the GNU Lesser General Public
3947
c8c591fe26f8 Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents: 3683
diff changeset
23 * License along with FFmpeg; if not, write to the Free Software
3036
0b546eab515d Update licensing information: The FSF changed postal address.
diego
parents: 2979
diff changeset
24 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
25 */
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
26
8069
316762ae96a7 ARM: use new macros for assembler function labels
mru
parents: 5967
diff changeset
27 #include "asm.S"
316762ae96a7 ARM: use new macros for assembler function labels
mru
parents: 5967
diff changeset
28
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
29 /* useful constants for the algorithm, they are save in __constant_ptr__ at */
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
30 /* the end of the source code.*/
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
31 #define W1 22725
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
32 #define W2 21407
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
33 #define W3 19266
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
34 #define W4 16383
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
35 #define W5 12873
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
36 #define W6 8867
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
37 #define W7 4520
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
38 #define MASK_MSHW 0xFFFF0000
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
39
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
40 /* offsets of the constants in the vector */
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
41 #define offW1 0
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
42 #define offW2 4
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
43 #define offW3 8
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
44 #define offW4 12
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
45 #define offW5 16
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
46 #define offW6 20
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
47 #define offW7 24
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
48 #define offMASK_MSHW 28
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
49
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
50 #define ROW_SHIFT 11
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
51 #define ROW_SHIFT2MSHW (16-11)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
52 #define COL_SHIFT 20
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
53 #define ROW_SHIFTED_1 1024 /* 1<< (ROW_SHIFT-1) */
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
54 #define COL_SHIFTED_1 524288 /* 1<< (COL_SHIFT-1) */
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
55
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
56
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
57 .text
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
58
8069
316762ae96a7 ARM: use new macros for assembler function labels
mru
parents: 5967
diff changeset
59 function simple_idct_ARM, export=1
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
60 @@ void simple_idct_ARM(int16_t *block)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
61 @@ save stack for reg needed (take all of them),
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
62 @@ R0-R3 are scratch regs, so no need to save them, but R0 contains the pointer to block
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
63 @@ so it must not be overwritten, if it is not saved!!
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
64 @@ R12 is another scratch register, so it should not be saved too
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
65 @@ save all registers
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
66 stmfd sp!, {r4-r11, r14} @ R14 is also called LR
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
67 @@ at this point, R0=block, other registers are free.
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
68 add r14, r0, #112 @ R14=&block[8*7], better start from the last row, and decrease the value until row=0, i.e. R12=block.
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
69 add r12, pc, #(__constant_ptr__-.-8) @ R12=__constant_ptr__, the vector containing the constants, probably not necessary to reserve a register for it
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
70 @@ add 2 temporary variables in the stack: R0 and R14
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
71 sub sp, sp, #8 @ allow 2 local variables
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
72 str r0, [sp, #0] @ save block in sp[0]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
73 @@ stack status
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
74 @@ sp+4 free
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
75 @@ sp+0 R0 (block)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
76
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
77
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
78 @@ at this point, R0=block, R14=&block[56], R12=__const_ptr_, R1-R11 free
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
79
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
80
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
81 __row_loop:
5967
15ed47af1838 Misc spelling fixes, prefer American over British English.
diego
parents: 5292
diff changeset
82 @@ read the row and check if it is null, almost null, or not, according to strongarm specs, it is not necessary to optimize ldr accesses (i.e. split 32bits in 2 16bits words), at least it gives more usable registers :)
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
83 ldr r1, [r14, #0] @ R1=(int32)(R12)[0]=ROWr32[0] (relative row cast to a 32b pointer)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
84 ldr r2, [r14, #4] @ R2=(int32)(R12)[1]=ROWr32[1]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
85 ldr r3, [r14, #8] @ R3=ROWr32[2]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
86 ldr r4, [r14, #12] @ R4=ROWr32[3]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
87 @@ check if the words are null, if all of them are null, then proceed with next row (branch __end_row_loop),
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
88 @@ if ROWr16[0] is the only one not null, then proceed with this special case (branch __almost_empty_row)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
89 @@ else follow the complete algorithm.
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
90 @@ at this point, R0=block, R14=&block[n], R12=__const_ptr_, R1=ROWr32[0], R2=ROWr32[1],
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
91 @@ R3=ROWr32[2], R4=ROWr32[3], R5-R11 free
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
92 orr r5, r4, r3 @ R5=R4 | R3
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
93 orr r5, r5, r2 @ R5=R4 | R3 | R2
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
94 orrs r6, r5, r1 @ Test R5 | R1 (the aim is to check if everything is null)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
95 beq __end_row_loop
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
96 mov r7, r1, asr #16 @ R7=R1>>16=ROWr16[1] (evaluate it now, as it could be useful later)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
97 ldrsh r6, [r14, #0] @ R6=ROWr16[0]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
98 orrs r5, r5, r7 @ R5=R4 | R3 | R2 | R7
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
99 beq __almost_empty_row
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
100
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
101 __b_evaluation:
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
102 @@ at this point, R0=block (temp), R1(free), R2=ROWr32[1], R3=ROWr32[2], R4=ROWr32[3],
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
103 @@ R5=(temp), R6=ROWr16[0], R7=ROWr16[1], R8-R11 free,
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
104 @@ R12=__const_ptr_, R14=&block[n]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
105 @@ to save some registers/calls, proceed with b0-b3 first, followed by a0-a3
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
106
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
107 @@ MUL16(b0, W1, row[1]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
108 @@ MUL16(b1, W3, row[1]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
109 @@ MUL16(b2, W5, row[1]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
110 @@ MUL16(b3, W7, row[1]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
111 @@ MAC16(b0, W3, row[3]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
112 @@ MAC16(b1, -W7, row[3]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
113 @@ MAC16(b2, -W1, row[3]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
114 @@ MAC16(b3, -W5, row[3]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
115 ldr r8, [r12, #offW1] @ R8=W1
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
116 mov r2, r2, asr #16 @ R2=ROWr16[3]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
117 mul r0, r8, r7 @ R0=W1*ROWr16[1]=b0 (ROWr16[1] must be the second arg, to have the possibility to save 1 cycle)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
118 ldr r9, [r12, #offW3] @ R9=W3
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
119 ldr r10, [r12, #offW5] @ R10=W5
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
120 mul r1, r9, r7 @ R1=W3*ROWr16[1]=b1 (ROWr16[1] must be the second arg, to have the possibility to save 1 cycle)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
121 ldr r11, [r12, #offW7] @ R11=W7
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
122 mul r5, r10, r7 @ R5=W5*ROWr16[1]=b2 (ROWr16[1] must be the second arg, to have the possibility to save 1 cycle)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
123 mul r7, r11, r7 @ R7=W7*ROWr16[1]=b3 (ROWr16[1] must be the second arg, to have the possibility to save 1 cycle)
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
124 teq r2, #0 @ if null avoid muls
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
125 mlane r0, r9, r2, r0 @ R0+=W3*ROWr16[3]=b0 (ROWr16[3] must be the second arg, to have the possibility to save 1 cycle)
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
126 rsbne r2, r2, #0 @ R2=-ROWr16[3]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
127 mlane r1, r11, r2, r1 @ R1-=W7*ROWr16[3]=b1 (ROWr16[3] must be the second arg, to have the possibility to save 1 cycle)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
128 mlane r5, r8, r2, r5 @ R5-=W1*ROWr16[3]=b2 (ROWr16[3] must be the second arg, to have the possibility to save 1 cycle)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
129 mlane r7, r10, r2, r7 @ R7-=W5*ROWr16[3]=b3 (ROWr16[3] must be the second arg, to have the possibility to save 1 cycle)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
130
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
131 @@ at this point, R0=b0, R1=b1, R2 (free), R3=ROWr32[2], R4=ROWr32[3],
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
132 @@ R5=b2, R6=ROWr16[0], R7=b3, R8=W1, R9=W3, R10=W5, R11=W7,
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
133 @@ R12=__const_ptr_, R14=&block[n]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
134 @@ temp = ((uint32_t*)row)[2] | ((uint32_t*)row)[3];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
135 @@ if (temp != 0) {}
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
136 orrs r2, r3, r4 @ R2=ROWr32[2] | ROWr32[3]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
137 beq __end_b_evaluation
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
138
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
139 @@ at this point, R0=b0, R1=b1, R2 (free), R3=ROWr32[2], R4=ROWr32[3],
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
140 @@ R5=b2, R6=ROWr16[0], R7=b3, R8=W1, R9=W3, R10=W5, R11=W7,
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
141 @@ R12=__const_ptr_, R14=&block[n]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
142 @@ MAC16(b0, W5, row[5]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
143 @@ MAC16(b2, W7, row[5]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
144 @@ MAC16(b3, W3, row[5]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
145 @@ MAC16(b1, -W1, row[5]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
146 @@ MAC16(b0, W7, row[7]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
147 @@ MAC16(b2, W3, row[7]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
148 @@ MAC16(b3, -W1, row[7]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
149 @@ MAC16(b1, -W5, row[7]);
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
150 mov r3, r3, asr #16 @ R3=ROWr16[5]
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
151 teq r3, #0 @ if null avoid muls
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
152 mlane r0, r10, r3, r0 @ R0+=W5*ROWr16[5]=b0
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
153 mov r4, r4, asr #16 @ R4=ROWr16[7]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
154 mlane r5, r11, r3, r5 @ R5+=W7*ROWr16[5]=b2
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
155 mlane r7, r9, r3, r7 @ R7+=W3*ROWr16[5]=b3
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
156 rsbne r3, r3, #0 @ R3=-ROWr16[5]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
157 mlane r1, r8, r3, r1 @ R7-=W1*ROWr16[5]=b1
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
158 @@ R3 is free now
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
159 teq r4, #0 @ if null avoid muls
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
160 mlane r0, r11, r4, r0 @ R0+=W7*ROWr16[7]=b0
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
161 mlane r5, r9, r4, r5 @ R5+=W3*ROWr16[7]=b2
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
162 rsbne r4, r4, #0 @ R4=-ROWr16[7]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
163 mlane r7, r8, r4, r7 @ R7-=W1*ROWr16[7]=b3
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
164 mlane r1, r10, r4, r1 @ R1-=W5*ROWr16[7]=b1
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
165 @@ R4 is free now
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
166 __end_b_evaluation:
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
167 @@ at this point, R0=b0, R1=b1, R2=ROWr32[2] | ROWr32[3] (tmp), R3 (free), R4 (free),
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
168 @@ R5=b2, R6=ROWr16[0], R7=b3, R8 (free), R9 (free), R10 (free), R11 (free),
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
169 @@ R12=__const_ptr_, R14=&block[n]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
170
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
171 __a_evaluation:
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
172 @@ a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
173 @@ a1 = a0 + W6 * row[2];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
174 @@ a2 = a0 - W6 * row[2];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
175 @@ a3 = a0 - W2 * row[2];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
176 @@ a0 = a0 + W2 * row[2];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
177 ldr r9, [r12, #offW4] @ R9=W4
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
178 mul r6, r9, r6 @ R6=W4*ROWr16[0]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
179 ldr r10, [r12, #offW6] @ R10=W6
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
180 ldrsh r4, [r14, #4] @ R4=ROWr16[2] (a3 not defined yet)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
181 add r6, r6, #ROW_SHIFTED_1 @ R6=W4*ROWr16[0] + 1<<(ROW_SHIFT-1) (a0)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
182
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
183 mul r11, r10, r4 @ R11=W6*ROWr16[2]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
184 ldr r8, [r12, #offW2] @ R8=W2
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
185 sub r3, r6, r11 @ R3=a0-W6*ROWr16[2] (a2)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
186 @@ temp = ((uint32_t*)row)[2] | ((uint32_t*)row)[3];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
187 @@ if (temp != 0) {}
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
188 teq r2, #0
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
189 beq __end_bef_a_evaluation
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
190
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
191 add r2, r6, r11 @ R2=a0+W6*ROWr16[2] (a1)
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
192 mul r11, r8, r4 @ R11=W2*ROWr16[2]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
193 sub r4, r6, r11 @ R4=a0-W2*ROWr16[2] (a3)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
194 add r6, r6, r11 @ R6=a0+W2*ROWr16[2] (a0)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
195
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
196
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
197 @@ at this point, R0=b0, R1=b1, R2=a1, R3=a2, R4=a3,
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
198 @@ R5=b2, R6=a0, R7=b3, R8=W2, R9=W4, R10=W6, R11 (free),
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
199 @@ R12=__const_ptr_, R14=&block[n]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
200
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
201
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
202 @@ a0 += W4*row[4]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
203 @@ a1 -= W4*row[4]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
204 @@ a2 -= W4*row[4]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
205 @@ a3 += W4*row[4]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
206 ldrsh r11, [r14, #8] @ R11=ROWr16[4]
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
207 teq r11, #0 @ if null avoid muls
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
208 mulne r11, r9, r11 @ R11=W4*ROWr16[4]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
209 @@ R9 is free now
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
210 ldrsh r9, [r14, #12] @ R9=ROWr16[6]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
211 addne r6, r6, r11 @ R6+=W4*ROWr16[4] (a0)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
212 subne r2, r2, r11 @ R2-=W4*ROWr16[4] (a1)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
213 subne r3, r3, r11 @ R3-=W4*ROWr16[4] (a2)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
214 addne r4, r4, r11 @ R4+=W4*ROWr16[4] (a3)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
215 @@ W6 alone is no more useful, save W2*ROWr16[6] in it instead
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
216 teq r9, #0 @ if null avoid muls
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
217 mulne r11, r10, r9 @ R11=W6*ROWr16[6]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
218 addne r6, r6, r11 @ R6+=W6*ROWr16[6] (a0)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
219 mulne r10, r8, r9 @ R10=W2*ROWr16[6]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
220 @@ a0 += W6*row[6];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
221 @@ a3 -= W6*row[6];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
222 @@ a1 -= W2*row[6];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
223 @@ a2 += W2*row[6];
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
224 subne r4, r4, r11 @ R4-=W6*ROWr16[6] (a3)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
225 subne r2, r2, r10 @ R2-=W2*ROWr16[6] (a1)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
226 addne r3, r3, r10 @ R3+=W2*ROWr16[6] (a2)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
227
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
228 __end_a_evaluation:
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
229 @@ at this point, R0=b0, R1=b1, R2=a1, R3=a2, R4=a3,
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
230 @@ R5=b2, R6=a0, R7=b3, R8 (free), R9 (free), R10 (free), R11 (free),
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
231 @@ R12=__const_ptr_, R14=&block[n]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
232 @@ row[0] = (a0 + b0) >> ROW_SHIFT;
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
233 @@ row[1] = (a1 + b1) >> ROW_SHIFT;
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
234 @@ row[2] = (a2 + b2) >> ROW_SHIFT;
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
235 @@ row[3] = (a3 + b3) >> ROW_SHIFT;
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
236 @@ row[4] = (a3 - b3) >> ROW_SHIFT;
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
237 @@ row[5] = (a2 - b2) >> ROW_SHIFT;
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
238 @@ row[6] = (a1 - b1) >> ROW_SHIFT;
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
239 @@ row[7] = (a0 - b0) >> ROW_SHIFT;
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
240 add r8, r6, r0 @ R8=a0+b0
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
241 add r9, r2, r1 @ R9=a1+b1
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
242 @@ put 2 16 bits half-words in a 32bits word
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
243 @@ ROWr32[0]=ROWr16[0] | (ROWr16[1]<<16) (only Little Endian compliant then!!!)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
244 ldr r10, [r12, #offMASK_MSHW] @ R10=0xFFFF0000
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
245 and r9, r10, r9, lsl #ROW_SHIFT2MSHW @ R9=0xFFFF0000 & ((a1+b1)<<5)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
246 mvn r11, r10 @ R11= NOT R10= 0x0000FFFF
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
247 and r8, r11, r8, asr #ROW_SHIFT @ R8=0x0000FFFF & ((a0+b0)>>11)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
248 orr r8, r8, r9
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
249 str r8, [r14, #0]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
250
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
251 add r8, r3, r5 @ R8=a2+b2
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
252 add r9, r4, r7 @ R9=a3+b3
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
253 and r9, r10, r9, lsl #ROW_SHIFT2MSHW @ R9=0xFFFF0000 & ((a3+b3)<<5)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
254 and r8, r11, r8, asr #ROW_SHIFT @ R8=0x0000FFFF & ((a2+b2)>>11)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
255 orr r8, r8, r9
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
256 str r8, [r14, #4]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
257
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
258 sub r8, r4, r7 @ R8=a3-b3
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
259 sub r9, r3, r5 @ R9=a2-b2
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
260 and r9, r10, r9, lsl #ROW_SHIFT2MSHW @ R9=0xFFFF0000 & ((a2-b2)<<5)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
261 and r8, r11, r8, asr #ROW_SHIFT @ R8=0x0000FFFF & ((a3-b3)>>11)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
262 orr r8, r8, r9
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
263 str r8, [r14, #8]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
264
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
265 sub r8, r2, r1 @ R8=a1-b1
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
266 sub r9, r6, r0 @ R9=a0-b0
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
267 and r9, r10, r9, lsl #ROW_SHIFT2MSHW @ R9=0xFFFF0000 & ((a0-b0)<<5)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
268 and r8, r11, r8, asr #ROW_SHIFT @ R8=0x0000FFFF & ((a1-b1)>>11)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
269 orr r8, r8, r9
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
270 str r8, [r14, #12]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
271
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
272 bal __end_row_loop
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
273
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
274 __almost_empty_row:
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
275 @@ the row was empty, except ROWr16[0], now, management of this special case
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
276 @@ at this point, R0=block, R14=&block[n], R12=__const_ptr_, R1=ROWr32[0], R2=ROWr32[1],
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
277 @@ R3=ROWr32[2], R4=ROWr32[3], R5=(temp), R6=ROWr16[0], R7=ROWr16[1],
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
278 @@ R8=0xFFFF (temp), R9-R11 free
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
279 mov r8, #0x10000 @ R8=0xFFFF (2 steps needed!) it saves a ldr call (because of delay run).
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
280 sub r8, r8, #1 @ R8 is now ready.
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
281 and r5, r8, r6, lsl #3 @ R5=R8 & (R6<<3)= (ROWr16[0]<<3) & 0xFFFF
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
282 orr r5, r5, r5, lsl #16 @ R5=R5 | (R5<<16)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
283 str r5, [r14, #0] @ R14[0]=ROWr32[0]=R5
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
284 str r5, [r14, #4] @ R14[4]=ROWr32[1]=R5
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
285 str r5, [r14, #8] @ R14[8]=ROWr32[2]=R5
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
286 str r5, [r14, #12] @ R14[12]=ROWr32[3]=R5
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
287
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
288 __end_row_loop:
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
289 @@ at this point, R0-R11 (free)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
290 @@ R12=__const_ptr_, R14=&block[n]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
291 ldr r0, [sp, #0] @ R0=block
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
292 teq r0, r14 @ compare current &block[8*n] to block, when block is reached, the loop is finished.
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
293 sub r14, r14, #16
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
294 bne __row_loop
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
295
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
296
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
297
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
298 @@ at this point, R0=block, R1-R11 (free)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
299 @@ R12=__const_ptr_, R14=&block[n]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
300 add r14, r0, #14 @ R14=&block[7], better start from the last col, and decrease the value until col=0, i.e. R14=block.
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
301 __col_loop:
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
302
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
303 __b_evaluation2:
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
304 @@ at this point, R0=block (temp), R1-R11 (free)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
305 @@ R12=__const_ptr_, R14=&block[n]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
306 @@ proceed with b0-b3 first, followed by a0-a3
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
307 @@ MUL16(b0, W1, col[8x1]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
308 @@ MUL16(b1, W3, col[8x1]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
309 @@ MUL16(b2, W5, col[8x1]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
310 @@ MUL16(b3, W7, col[8x1]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
311 @@ MAC16(b0, W3, col[8x3]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
312 @@ MAC16(b1, -W7, col[8x3]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
313 @@ MAC16(b2, -W1, col[8x3]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
314 @@ MAC16(b3, -W5, col[8x3]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
315 ldr r8, [r12, #offW1] @ R8=W1
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
316 ldrsh r7, [r14, #16]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
317 mul r0, r8, r7 @ R0=W1*ROWr16[1]=b0 (ROWr16[1] must be the second arg, to have the possibility to save 1 cycle)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
318 ldr r9, [r12, #offW3] @ R9=W3
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
319 ldr r10, [r12, #offW5] @ R10=W5
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
320 mul r1, r9, r7 @ R1=W3*ROWr16[1]=b1 (ROWr16[1] must be the second arg, to have the possibility to save 1 cycle)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
321 ldr r11, [r12, #offW7] @ R11=W7
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
322 mul r5, r10, r7 @ R5=W5*ROWr16[1]=b2 (ROWr16[1] must be the second arg, to have the possibility to save 1 cycle)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
323 ldrsh r2, [r14, #48]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
324 mul r7, r11, r7 @ R7=W7*ROWr16[1]=b3 (ROWr16[1] must be the second arg, to have the possibility to save 1 cycle)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
325 teq r2, #0 @ if 0, then avoid muls
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
326 mlane r0, r9, r2, r0 @ R0+=W3*ROWr16[3]=b0 (ROWr16[3] must be the second arg, to have the possibility to save 1 cycle)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
327 rsbne r2, r2, #0 @ R2=-ROWr16[3]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
328 mlane r1, r11, r2, r1 @ R1-=W7*ROWr16[3]=b1 (ROWr16[3] must be the second arg, to have the possibility to save 1 cycle)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
329 mlane r5, r8, r2, r5 @ R5-=W1*ROWr16[3]=b2 (ROWr16[3] must be the second arg, to have the possibility to save 1 cycle)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
330 mlane r7, r10, r2, r7 @ R7-=W5*ROWr16[3]=b3 (ROWr16[3] must be the second arg, to have the possibility to save 1 cycle)
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
331
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
332 @@ at this point, R0=b0, R1=b1, R2 (free), R3 (free), R4 (free),
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
333 @@ R5=b2, R6 (free), R7=b3, R8=W1, R9=W3, R10=W5, R11=W7,
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
334 @@ R12=__const_ptr_, R14=&block[n]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
335 @@ MAC16(b0, W5, col[5x8]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
336 @@ MAC16(b2, W7, col[5x8]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
337 @@ MAC16(b3, W3, col[5x8]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
338 @@ MAC16(b1, -W1, col[5x8]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
339 @@ MAC16(b0, W7, col[7x8]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
340 @@ MAC16(b2, W3, col[7x8]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
341 @@ MAC16(b3, -W1, col[7x8]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
342 @@ MAC16(b1, -W5, col[7x8]);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
343 ldrsh r3, [r14, #80] @ R3=COLr16[5x8]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
344 teq r3, #0 @ if 0 then avoid muls
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
345 mlane r0, r10, r3, r0 @ R0+=W5*ROWr16[5x8]=b0
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
346 mlane r5, r11, r3, r5 @ R5+=W7*ROWr16[5x8]=b2
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
347 mlane r7, r9, r3, r7 @ R7+=W3*ROWr16[5x8]=b3
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
348 rsbne r3, r3, #0 @ R3=-ROWr16[5x8]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
349 ldrsh r4, [r14, #112] @ R4=COLr16[7x8]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
350 mlane r1, r8, r3, r1 @ R7-=W1*ROWr16[5x8]=b1
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
351 @@ R3 is free now
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
352 teq r4, #0 @ if 0 then avoid muls
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
353 mlane r0, r11, r4, r0 @ R0+=W7*ROWr16[7x8]=b0
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
354 mlane r5, r9, r4, r5 @ R5+=W3*ROWr16[7x8]=b2
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
355 rsbne r4, r4, #0 @ R4=-ROWr16[7x8]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
356 mlane r7, r8, r4, r7 @ R7-=W1*ROWr16[7x8]=b3
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
357 mlane r1, r10, r4, r1 @ R1-=W5*ROWr16[7x8]=b1
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
358 @@ R4 is free now
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
359 __end_b_evaluation2:
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
360 @@ at this point, R0=b0, R1=b1, R2 (free), R3 (free), R4 (free),
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
361 @@ R5=b2, R6 (free), R7=b3, R8 (free), R9 (free), R10 (free), R11 (free),
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
362 @@ R12=__const_ptr_, R14=&block[n]
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
363
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
364 __a_evaluation2:
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
365 @@ a0 = (W4 * col[8x0]) + (1 << (COL_SHIFT - 1));
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
366 @@ a1 = a0 + W6 * row[2];
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
367 @@ a2 = a0 - W6 * row[2];
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
368 @@ a3 = a0 - W2 * row[2];
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
369 @@ a0 = a0 + W2 * row[2];
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
370 ldrsh r6, [r14, #0]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
371 ldr r9, [r12, #offW4] @ R9=W4
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
372 mul r6, r9, r6 @ R6=W4*ROWr16[0]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
373 ldr r10, [r12, #offW6] @ R10=W6
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
374 ldrsh r4, [r14, #32] @ R4=ROWr16[2] (a3 not defined yet)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
375 add r6, r6, #COL_SHIFTED_1 @ R6=W4*ROWr16[0] + 1<<(COL_SHIFT-1) (a0)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
376 mul r11, r10, r4 @ R11=W6*ROWr16[2]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
377 ldr r8, [r12, #offW2] @ R8=W2
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
378 add r2, r6, r11 @ R2=a0+W6*ROWr16[2] (a1)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
379 sub r3, r6, r11 @ R3=a0-W6*ROWr16[2] (a2)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
380 mul r11, r8, r4 @ R11=W2*ROWr16[2]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
381 sub r4, r6, r11 @ R4=a0-W2*ROWr16[2] (a3)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
382 add r6, r6, r11 @ R6=a0+W2*ROWr16[2] (a0)
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
383
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
384 @@ at this point, R0=b0, R1=b1, R2=a1, R3=a2, R4=a3,
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
385 @@ R5=b2, R6=a0, R7=b3, R8=W2, R9=W4, R10=W6, R11 (free),
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
386 @@ R12=__const_ptr_, R14=&block[n]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
387 @@ a0 += W4*row[4]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
388 @@ a1 -= W4*row[4]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
389 @@ a2 -= W4*row[4]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
390 @@ a3 += W4*row[4]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
391 ldrsh r11, [r14, #64] @ R11=ROWr16[4]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
392 teq r11, #0 @ if null avoid muls
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
393 mulne r11, r9, r11 @ R11=W4*ROWr16[4]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
394 @@ R9 is free now
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
395 addne r6, r6, r11 @ R6+=W4*ROWr16[4] (a0)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
396 subne r2, r2, r11 @ R2-=W4*ROWr16[4] (a1)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
397 subne r3, r3, r11 @ R3-=W4*ROWr16[4] (a2)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
398 ldrsh r9, [r14, #96] @ R9=ROWr16[6]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
399 addne r4, r4, r11 @ R4+=W4*ROWr16[4] (a3)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
400 @@ W6 alone is no more useful, save W2*ROWr16[6] in it instead
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
401 teq r9, #0 @ if null avoid muls
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
402 mulne r11, r10, r9 @ R11=W6*ROWr16[6]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
403 addne r6, r6, r11 @ R6+=W6*ROWr16[6] (a0)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
404 mulne r10, r8, r9 @ R10=W2*ROWr16[6]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
405 @@ a0 += W6*row[6];
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
406 @@ a3 -= W6*row[6];
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
407 @@ a1 -= W2*row[6];
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
408 @@ a2 += W2*row[6];
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
409 subne r4, r4, r11 @ R4-=W6*ROWr16[6] (a3)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
410 subne r2, r2, r10 @ R2-=W2*ROWr16[6] (a1)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
411 addne r3, r3, r10 @ R3+=W2*ROWr16[6] (a2)
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
412 __end_a_evaluation2:
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
413 @@ at this point, R0=b0, R1=b1, R2=a1, R3=a2, R4=a3,
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
414 @@ R5=b2, R6=a0, R7=b3, R8 (free), R9 (free), R10 (free), R11 (free),
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
415 @@ R12=__const_ptr_, R14=&block[n]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
416 @@ col[0 ] = ((a0 + b0) >> COL_SHIFT);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
417 @@ col[8 ] = ((a1 + b1) >> COL_SHIFT);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
418 @@ col[16] = ((a2 + b2) >> COL_SHIFT);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
419 @@ col[24] = ((a3 + b3) >> COL_SHIFT);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
420 @@ col[32] = ((a3 - b3) >> COL_SHIFT);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
421 @@ col[40] = ((a2 - b2) >> COL_SHIFT);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
422 @@ col[48] = ((a1 - b1) >> COL_SHIFT);
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
423 @@ col[56] = ((a0 - b0) >> COL_SHIFT);
5967
15ed47af1838 Misc spelling fixes, prefer American over British English.
diego
parents: 5292
diff changeset
424 @@@@@ no optimization here @@@@@
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
425 add r8, r6, r0 @ R8=a0+b0
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
426 add r9, r2, r1 @ R9=a1+b1
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
427 mov r8, r8, asr #COL_SHIFT
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
428 mov r9, r9, asr #COL_SHIFT
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
429 strh r8, [r14, #0]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
430 strh r9, [r14, #16]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
431 add r8, r3, r5 @ R8=a2+b2
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
432 add r9, r4, r7 @ R9=a3+b3
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
433 mov r8, r8, asr #COL_SHIFT
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
434 mov r9, r9, asr #COL_SHIFT
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
435 strh r8, [r14, #32]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
436 strh r9, [r14, #48]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
437 sub r8, r4, r7 @ R8=a3-b3
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
438 sub r9, r3, r5 @ R9=a2-b2
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
439 mov r8, r8, asr #COL_SHIFT
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
440 mov r9, r9, asr #COL_SHIFT
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
441 strh r8, [r14, #64]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
442 strh r9, [r14, #80]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
443 sub r8, r2, r1 @ R8=a1-b1
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
444 sub r9, r6, r0 @ R9=a0-b0
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
445 mov r8, r8, asr #COL_SHIFT
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
446 mov r9, r9, asr #COL_SHIFT
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
447 strh r8, [r14, #96]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
448 strh r9, [r14, #112]
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
449
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
450 __end_col_loop:
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
451 @@ at this point, R0-R11 (free)
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
452 @@ R12=__const_ptr_, R14=&block[n]
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
453 ldr r0, [sp, #0] @ R0=block
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
454 teq r0, r14 @ compare current &block[n] to block, when block is reached, the loop is finished.
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
455 sub r14, r14, #2
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
456 bne __col_loop
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
457
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
458
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
459
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
460
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
461 __end_simple_idct_ARM:
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
462 @@ restore registers to previous status!
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
463 add sp, sp, #8 @@ the local variables!
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
464 ldmfd sp!, {r4-r11, r15} @@ update PC with LR content.
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
465
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
466
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
467
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
468 @@ kind of sub-function, here not to overload the common case.
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
469 __end_bef_a_evaluation:
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
470 add r2, r6, r11 @ R2=a0+W6*ROWr16[2] (a1)
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
471 mul r11, r8, r4 @ R11=W2*ROWr16[2]
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
472 sub r4, r6, r11 @ R4=a0-W2*ROWr16[2] (a3)
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
473 add r6, r6, r11 @ R6=a0+W2*ROWr16[2] (a0)
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
474 bal __end_a_evaluation
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
475
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
476
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
477 __constant_ptr__: @@ see #defines at the beginning of the source code for values.
2979
bfabfdf9ce55 COSMETICS: tabs --> spaces, some prettyprinting
diego
parents: 2967
diff changeset
478 .align
1347
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
479 .word W1
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
480 .word W2
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
481 .word W3
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
482 .word W4
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
483 .word W5
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
484 .word W6
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
485 .word W7
cca26199ab17 Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb@handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
al3x
parents:
diff changeset
486 .word MASK_MSHW