Mercurial > libavcodec.hg
annotate ps2/mpegvideo_mmi.c @ 11352:6e0af2cfdcfe libavcodec
Do MC and IDCT in coding (hilbert) order
This increases the slice size to 64 pixels, due to having to decode an
entire chroma superblock row per slice.
This can be up to 6% slower depending on clip and CPU, but is necessary
for future optimizations that gain significantly more than was lost.
author | conrad |
---|---|
date | Wed, 03 Mar 2010 23:27:40 +0000 |
parents | 04423b2f6e0b |
children |
rev | line source |
---|---|
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
1 /* |
8629
04423b2f6e0b
cosmetics: Remove pointless period after copyright statement non-sentences.
diego
parents:
8031
diff
changeset
|
2 * Copyright (c) 2000,2001 Fabrice Bellard |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
3 * |
5214 | 4 * MMI optimization by Leon van Stuivenberg |
5 * | |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3036
diff
changeset
|
6 * This file is part of FFmpeg. |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3036
diff
changeset
|
7 * |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3036
diff
changeset
|
8 * FFmpeg is free software; you can redistribute it and/or |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
9 * modify it under the terms of the GNU Lesser General Public |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
10 * 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:
3036
diff
changeset
|
11 * version 2.1 of the License, or (at your option) any later version. |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
12 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3036
diff
changeset
|
13 * FFmpeg is distributed in the hope that it will be useful, |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
16 * Lesser General Public License for more details. |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
17 * |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
18 * 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:
3036
diff
changeset
|
19 * 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
|
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
21 */ |
2967 | 22 |
6763 | 23 #include "libavcodec/avcodec.h" |
24 #include "libavcodec/dsputil.h" | |
25 #include "libavcodec/mpegvideo.h" | |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
26 |
2967 | 27 static void dct_unquantize_h263_mmi(MpegEncContext *s, |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
28 DCTELEM *block, int n, int qscale) |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
29 { |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
30 int level=0, qmul, qadd; |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
31 int nCoeffs; |
2967 | 32 |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
33 assert(s->block_last_index[n]>=0); |
2967 | 34 |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
35 qadd = (qscale - 1) | 1; |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
36 qmul = qscale << 1; |
2967 | 37 |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
38 if (s->mb_intra) { |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
39 if (!s->h263_aic) { |
2967 | 40 if (n < 4) |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
41 level = block[0] * s->y_dc_scale; |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
42 else |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
43 level = block[0] * s->c_dc_scale; |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
44 }else { |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
45 qadd = 0; |
2979 | 46 level = block[0]; |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
47 } |
4795 | 48 nCoeffs= 63; //does not always use zigzag table |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
49 } else { |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
50 nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ]; |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
51 } |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
52 |
8031 | 53 __asm__ volatile( |
2979 | 54 "add $14, $0, %3 \n\t" |
55 "pcpyld $8, %0, %0 \n\t" | |
56 "pcpyh $8, $8 \n\t" //r8 = qmul | |
57 "pcpyld $9, %1, %1 \n\t" | |
58 "pcpyh $9, $9 \n\t" //r9 = qadd | |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
59 ".p2align 2 \n\t" |
2979 | 60 "1: \n\t" |
61 "lq $10, 0($14) \n\t" //r10 = level | |
62 "addi $14, $14, 16 \n\t" //block+=8 | |
63 "addi %2, %2, -8 \n\t" | |
64 "pcgth $11, $0, $10 \n\t" //r11 = level < 0 ? -1 : 0 | |
65 "pcgth $12, $10, $0 \n\t" //r12 = level > 0 ? -1 : 0 | |
66 "por $12, $11, $12 \n\t" | |
67 "pmulth $10, $10, $8 \n\t" | |
68 "paddh $13, $9, $11 \n\t" | |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
69 "pxor $13, $13, $11 \n\t" //r13 = level < 0 ? -qadd : qadd |
2979 | 70 "pmfhl.uw $11 \n\t" |
71 "pinteh $10, $11, $10 \n\t" //r10 = level * qmul | |
72 "paddh $10, $10, $13 \n\t" | |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
73 "pand $10, $10, $12 \n\t" |
2979 | 74 "sq $10, -16($14) \n\t" |
75 "bgez %2, 1b \n\t" | |
76 :: "r"(qmul), "r" (qadd), "r" (nCoeffs), "r" (block) : "$8", "$9", "$10", "$11", "$12", "$13", "$14", "memory" ); | |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
77 |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
78 if(s->mb_intra) |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
79 block[0]= level; |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
80 } |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
81 |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
82 |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
83 void MPV_common_init_mmi(MpegEncContext *s) |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
84 { |
2967 | 85 s->dct_unquantize_h263_intra = |
1835 | 86 s->dct_unquantize_h263_inter = dct_unquantize_h263_mmi; |
721
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
87 } |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
88 |
71f669e9f633
ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
michaelni
parents:
diff
changeset
|
89 |