Mercurial > libavcodec.hg
annotate simple_idct.c @ 932:176fd8c8e8a8 libavcodec
mb qp limits
author | michaelni |
---|---|
date | Wed, 18 Dec 2002 10:09:43 +0000 |
parents | 2f7da29ede37 |
children | caa77cd960c0 |
rev | line source |
---|---|
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
1 /* |
429 | 2 * Simple IDCT |
3 * | |
4 * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at> | |
5 * | |
6 * This library is free software; you can redistribute it and/or | |
7 * modify it under the terms of the GNU Lesser General Public | |
8 * License as published by the Free Software Foundation; either | |
9 * version 2 of the License, or (at your option) any later version. | |
10 * | |
11 * This library is distributed in the hope that it will be useful, | |
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 * Lesser General Public License for more details. | |
15 * | |
16 * You should have received a copy of the GNU Lesser General Public | |
17 * License along with this library; if not, write to the Free Software | |
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
19 */ | |
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
20 /* |
429 | 21 based upon some outcommented c code from mpeg2dec (idct_mmx.c |
22 written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>) | |
23 */ | |
396
fce0a2520551
removed useless header includes - use av memory functions
glantau
parents:
352
diff
changeset
|
24 #include "avcodec.h" |
479 | 25 #include "dsputil.h" |
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
26 #include "simple_idct.h" |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
27 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
28 #if 0 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
29 #define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */ |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
30 #define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */ |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
31 #define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */ |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
32 #define W4 2048 /* 2048*sqrt (2)*cos (4*pi/16) */ |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
33 #define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */ |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
34 #define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */ |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
35 #define W7 565 /* 2048*sqrt (2)*cos (7*pi/16) */ |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
36 #define ROW_SHIFT 8 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
37 #define COL_SHIFT 17 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
38 #else |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
39 #define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
40 #define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
41 #define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
352
5a8eb5cf9f92
C4=16383 for the c version too and even for some outcommented code
michaelni
parents:
215
diff
changeset
|
42 #define W4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
43 #define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
44 #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
45 #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
46 #define ROW_SHIFT 11 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
47 #define COL_SHIFT 20 // 6 |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
48 #endif |
205 | 49 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
50 #if defined(ARCH_POWERPC_405) |
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
51 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
52 /* signed 16x16 -> 32 multiply add accumulate */ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
53 #define MAC16(rt, ra, rb) \ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
54 asm ("maclhw %0, %2, %3" : "=r" (rt) : "0" (rt), "r" (ra), "r" (rb)); |
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
55 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
56 /* signed 16x16 -> 32 multiply */ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
57 #define MUL16(rt, ra, rb) \ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
58 asm ("mullhw %0, %1, %2" : "=r" (rt) : "r" (ra), "r" (rb)); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
59 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
60 #else |
205 | 61 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
62 /* signed 16x16 -> 32 multiply add accumulate */ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
63 #define MAC16(rt, ra, rb) rt += (ra) * (rb) |
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
64 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
65 /* signed 16x16 -> 32 multiply */ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
66 #define MUL16(rt, ra, rb) rt = (ra) * (rb) |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
67 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
68 #endif |
205 | 69 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
70 static inline void idctRowCondDC (int16_t * row) |
205 | 71 { |
72 int a0, a1, a2, a3, b0, b1, b2, b3; | |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
73 #ifdef FAST_64BIT |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
74 uint64_t temp; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
75 #else |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
76 uint32_t temp; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
77 #endif |
205 | 78 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
79 #ifdef FAST_64BIT |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
80 #ifdef WORDS_BIGENDIAN |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
81 #define ROW0_MASK 0xffff000000000000LL |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
82 #else |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
83 #define ROW0_MASK 0xffffLL |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
84 #endif |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
85 if ( ((((uint64_t *)row)[0] & ~ROW0_MASK) | |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
86 ((uint64_t *)row)[1]) == 0) { |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
87 temp = (row[0] << 3) & 0xffff; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
88 temp += temp << 16; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
89 temp += temp << 32; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
90 ((uint64_t *)row)[0] = temp; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
91 ((uint64_t *)row)[1] = temp; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
92 return; |
205 | 93 } |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
94 #else |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
95 if (!(((uint32_t*)row)[1] | |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
96 ((uint32_t*)row)[2] | |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
97 ((uint32_t*)row)[3] | |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
98 row[1])) { |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
99 temp = (row[0] << 3) & 0xffff; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
100 temp += temp << 16; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
101 ((uint32_t*)row)[0]=((uint32_t*)row)[1] = |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
102 ((uint32_t*)row)[2]=((uint32_t*)row)[3] = temp; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
103 return; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
104 } |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
105 #endif |
205 | 106 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
107 a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1)); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
108 a1 = a0; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
109 a2 = a0; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
110 a3 = a0; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
111 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
112 /* no need to optimize : gcc does it */ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
113 a0 += W2 * row[2]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
114 a1 += W6 * row[2]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
115 a2 -= W6 * row[2]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
116 a3 -= W2 * row[2]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
117 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
118 MUL16(b0, W1, row[1]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
119 MAC16(b0, W3, row[3]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
120 MUL16(b1, W3, row[1]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
121 MAC16(b1, -W7, row[3]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
122 MUL16(b2, W5, row[1]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
123 MAC16(b2, -W1, row[3]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
124 MUL16(b3, W7, row[1]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
125 MAC16(b3, -W5, row[3]); |
205 | 126 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
127 #ifdef FAST_64BIT |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
128 temp = ((uint64_t*)row)[1]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
129 #else |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
130 temp = ((uint32_t*)row)[2] | ((uint32_t*)row)[3]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
131 #endif |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
132 if (temp != 0) { |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
133 a0 += W4*row[4] + W6*row[6]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
134 a1 += - W4*row[4] - W2*row[6]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
135 a2 += - W4*row[4] + W2*row[6]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
136 a3 += W4*row[4] - W6*row[6]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
137 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
138 MAC16(b0, W5, row[5]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
139 MAC16(b0, W7, row[7]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
140 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
141 MAC16(b1, -W1, row[5]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
142 MAC16(b1, -W5, row[7]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
143 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
144 MAC16(b2, W7, row[5]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
145 MAC16(b2, W3, row[7]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
146 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
147 MAC16(b3, W3, row[5]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
148 MAC16(b3, -W1, row[7]); |
205 | 149 } |
150 | |
151 row[0] = (a0 + b0) >> ROW_SHIFT; | |
152 row[7] = (a0 - b0) >> ROW_SHIFT; | |
153 row[1] = (a1 + b1) >> ROW_SHIFT; | |
154 row[6] = (a1 - b1) >> ROW_SHIFT; | |
155 row[2] = (a2 + b2) >> ROW_SHIFT; | |
156 row[5] = (a2 - b2) >> ROW_SHIFT; | |
157 row[3] = (a3 + b3) >> ROW_SHIFT; | |
158 row[4] = (a3 - b3) >> ROW_SHIFT; | |
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
159 } |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
160 |
479 | 161 static inline void idctSparseColPut (UINT8 *dest, int line_size, |
162 int16_t * col) | |
205 | 163 { |
164 int a0, a1, a2, a3, b0, b1, b2, b3; | |
479 | 165 UINT8 *cm = cropTbl + MAX_NEG_CROP; |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
166 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
167 /* XXX: I did that only to give same values as previous code */ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
168 a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
169 a1 = a0; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
170 a2 = a0; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
171 a3 = a0; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
172 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
173 a0 += + W2*col[8*2]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
174 a1 += + W6*col[8*2]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
175 a2 += - W6*col[8*2]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
176 a3 += - W2*col[8*2]; |
205 | 177 |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
178 MUL16(b0, W1, col[8*1]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
179 MUL16(b1, W3, col[8*1]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
180 MUL16(b2, W5, col[8*1]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
181 MUL16(b3, W7, col[8*1]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
182 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
183 MAC16(b0, + W3, col[8*3]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
184 MAC16(b1, - W7, col[8*3]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
185 MAC16(b2, - W1, col[8*3]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
186 MAC16(b3, - W5, col[8*3]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
187 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
188 if(col[8*4]){ |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
189 a0 += + W4*col[8*4]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
190 a1 += - W4*col[8*4]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
191 a2 += - W4*col[8*4]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
192 a3 += + W4*col[8*4]; |
205 | 193 } |
194 | |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
195 if (col[8*5]) { |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
196 MAC16(b0, + W5, col[8*5]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
197 MAC16(b1, - W1, col[8*5]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
198 MAC16(b2, + W7, col[8*5]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
199 MAC16(b3, + W3, col[8*5]); |
205 | 200 } |
201 | |
202 if(col[8*6]){ | |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
203 a0 += + W6*col[8*6]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
204 a1 += - W2*col[8*6]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
205 a2 += + W2*col[8*6]; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
206 a3 += - W6*col[8*6]; |
205 | 207 } |
208 | |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
209 if (col[8*7]) { |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
210 MAC16(b0, + W7, col[8*7]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
211 MAC16(b1, - W5, col[8*7]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
212 MAC16(b2, + W3, col[8*7]); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
213 MAC16(b3, - W1, col[8*7]); |
205 | 214 } |
215 | |
479 | 216 dest[0] = cm[(a0 + b0) >> COL_SHIFT]; |
217 dest += line_size; | |
218 dest[0] = cm[(a1 + b1) >> COL_SHIFT]; | |
219 dest += line_size; | |
220 dest[0] = cm[(a2 + b2) >> COL_SHIFT]; | |
221 dest += line_size; | |
222 dest[0] = cm[(a3 + b3) >> COL_SHIFT]; | |
223 dest += line_size; | |
224 dest[0] = cm[(a3 - b3) >> COL_SHIFT]; | |
225 dest += line_size; | |
226 dest[0] = cm[(a2 - b2) >> COL_SHIFT]; | |
227 dest += line_size; | |
228 dest[0] = cm[(a1 - b1) >> COL_SHIFT]; | |
229 dest += line_size; | |
230 dest[0] = cm[(a0 - b0) >> COL_SHIFT]; | |
231 } | |
232 | |
233 static inline void idctSparseColAdd (UINT8 *dest, int line_size, | |
234 int16_t * col) | |
235 { | |
236 int a0, a1, a2, a3, b0, b1, b2, b3; | |
237 UINT8 *cm = cropTbl + MAX_NEG_CROP; | |
238 | |
239 /* XXX: I did that only to give same values as previous code */ | |
240 a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); | |
241 a1 = a0; | |
242 a2 = a0; | |
243 a3 = a0; | |
244 | |
245 a0 += + W2*col[8*2]; | |
246 a1 += + W6*col[8*2]; | |
247 a2 += - W6*col[8*2]; | |
248 a3 += - W2*col[8*2]; | |
249 | |
250 MUL16(b0, W1, col[8*1]); | |
251 MUL16(b1, W3, col[8*1]); | |
252 MUL16(b2, W5, col[8*1]); | |
253 MUL16(b3, W7, col[8*1]); | |
254 | |
255 MAC16(b0, + W3, col[8*3]); | |
256 MAC16(b1, - W7, col[8*3]); | |
257 MAC16(b2, - W1, col[8*3]); | |
258 MAC16(b3, - W5, col[8*3]); | |
259 | |
260 if(col[8*4]){ | |
261 a0 += + W4*col[8*4]; | |
262 a1 += - W4*col[8*4]; | |
263 a2 += - W4*col[8*4]; | |
264 a3 += + W4*col[8*4]; | |
265 } | |
266 | |
267 if (col[8*5]) { | |
268 MAC16(b0, + W5, col[8*5]); | |
269 MAC16(b1, - W1, col[8*5]); | |
270 MAC16(b2, + W7, col[8*5]); | |
271 MAC16(b3, + W3, col[8*5]); | |
272 } | |
273 | |
274 if(col[8*6]){ | |
275 a0 += + W6*col[8*6]; | |
276 a1 += - W2*col[8*6]; | |
277 a2 += + W2*col[8*6]; | |
278 a3 += - W6*col[8*6]; | |
279 } | |
280 | |
281 if (col[8*7]) { | |
282 MAC16(b0, + W7, col[8*7]); | |
283 MAC16(b1, - W5, col[8*7]); | |
284 MAC16(b2, + W3, col[8*7]); | |
285 MAC16(b3, - W1, col[8*7]); | |
286 } | |
287 | |
288 dest[0] = cm[dest[0] + ((a0 + b0) >> COL_SHIFT)]; | |
289 dest += line_size; | |
290 dest[0] = cm[dest[0] + ((a1 + b1) >> COL_SHIFT)]; | |
291 dest += line_size; | |
292 dest[0] = cm[dest[0] + ((a2 + b2) >> COL_SHIFT)]; | |
293 dest += line_size; | |
294 dest[0] = cm[dest[0] + ((a3 + b3) >> COL_SHIFT)]; | |
295 dest += line_size; | |
296 dest[0] = cm[dest[0] + ((a3 - b3) >> COL_SHIFT)]; | |
297 dest += line_size; | |
298 dest[0] = cm[dest[0] + ((a2 - b2) >> COL_SHIFT)]; | |
299 dest += line_size; | |
300 dest[0] = cm[dest[0] + ((a1 - b1) >> COL_SHIFT)]; | |
301 dest += line_size; | |
302 dest[0] = cm[dest[0] + ((a0 - b0) >> COL_SHIFT)]; | |
175
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
303 } |
bd77d3cbb233
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
diff
changeset
|
304 |
633 | 305 static inline void idctSparseCol (int16_t * col) |
306 { | |
307 int a0, a1, a2, a3, b0, b1, b2, b3; | |
308 | |
309 /* XXX: I did that only to give same values as previous code */ | |
310 a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); | |
311 a1 = a0; | |
312 a2 = a0; | |
313 a3 = a0; | |
314 | |
315 a0 += + W2*col[8*2]; | |
316 a1 += + W6*col[8*2]; | |
317 a2 += - W6*col[8*2]; | |
318 a3 += - W2*col[8*2]; | |
319 | |
320 MUL16(b0, W1, col[8*1]); | |
321 MUL16(b1, W3, col[8*1]); | |
322 MUL16(b2, W5, col[8*1]); | |
323 MUL16(b3, W7, col[8*1]); | |
324 | |
325 MAC16(b0, + W3, col[8*3]); | |
326 MAC16(b1, - W7, col[8*3]); | |
327 MAC16(b2, - W1, col[8*3]); | |
328 MAC16(b3, - W5, col[8*3]); | |
329 | |
330 if(col[8*4]){ | |
331 a0 += + W4*col[8*4]; | |
332 a1 += - W4*col[8*4]; | |
333 a2 += - W4*col[8*4]; | |
334 a3 += + W4*col[8*4]; | |
335 } | |
336 | |
337 if (col[8*5]) { | |
338 MAC16(b0, + W5, col[8*5]); | |
339 MAC16(b1, - W1, col[8*5]); | |
340 MAC16(b2, + W7, col[8*5]); | |
341 MAC16(b3, + W3, col[8*5]); | |
342 } | |
343 | |
344 if(col[8*6]){ | |
345 a0 += + W6*col[8*6]; | |
346 a1 += - W2*col[8*6]; | |
347 a2 += + W2*col[8*6]; | |
348 a3 += - W6*col[8*6]; | |
349 } | |
350 | |
351 if (col[8*7]) { | |
352 MAC16(b0, + W7, col[8*7]); | |
353 MAC16(b1, - W5, col[8*7]); | |
354 MAC16(b2, + W3, col[8*7]); | |
355 MAC16(b3, - W1, col[8*7]); | |
356 } | |
357 | |
358 col[0 ] = ((a0 + b0) >> COL_SHIFT); | |
359 col[8 ] = ((a1 + b1) >> COL_SHIFT); | |
360 col[16] = ((a2 + b2) >> COL_SHIFT); | |
361 col[24] = ((a3 + b3) >> COL_SHIFT); | |
362 col[32] = ((a3 - b3) >> COL_SHIFT); | |
363 col[40] = ((a2 - b2) >> COL_SHIFT); | |
364 col[48] = ((a1 - b1) >> COL_SHIFT); | |
365 col[56] = ((a0 - b0) >> COL_SHIFT); | |
366 } | |
367 | |
479 | 368 void simple_idct_put(UINT8 *dest, int line_size, INT16 *block) |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
369 { |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
370 int i; |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
371 for(i=0; i<8; i++) |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
372 idctRowCondDC(block + i*8); |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
373 |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
374 for(i=0; i<8; i++) |
479 | 375 idctSparseColPut(dest + i, line_size, block + i); |
376 } | |
377 | |
378 void simple_idct_add(UINT8 *dest, int line_size, INT16 *block) | |
379 { | |
380 int i; | |
381 for(i=0; i<8; i++) | |
382 idctRowCondDC(block + i*8); | |
383 | |
384 for(i=0; i<8; i++) | |
385 idctSparseColAdd(dest + i, line_size, block + i); | |
476
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
386 } |
ec13b0a726c3
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
bellard
parents:
464
diff
changeset
|
387 |
633 | 388 void simple_idct(INT16 *block) |
389 { | |
390 int i; | |
391 for(i=0; i<8; i++) | |
392 idctRowCondDC(block + i*8); | |
393 | |
394 for(i=0; i<8; i++) | |
395 idctSparseCol(block + i); | |
396 } | |
397 | |
719 | 398 /* 2x4x8 idct */ |
399 | |
400 #define CN_SHIFT 12 | |
401 #define C_FIX(x) ((int)((x) * (1 << CN_SHIFT) + 0.5)) | |
722
ff90043f4a2d
in fact IDCT248 needs to be normalized as I suspected
bellard
parents:
719
diff
changeset
|
402 #define C1 C_FIX(0.6532814824) |
ff90043f4a2d
in fact IDCT248 needs to be normalized as I suspected
bellard
parents:
719
diff
changeset
|
403 #define C2 C_FIX(0.2705980501) |
719 | 404 |
722
ff90043f4a2d
in fact IDCT248 needs to be normalized as I suspected
bellard
parents:
719
diff
changeset
|
405 /* row idct is multiple by 16 * sqrt(2.0), col idct4 is normalized, |
ff90043f4a2d
in fact IDCT248 needs to be normalized as I suspected
bellard
parents:
719
diff
changeset
|
406 and the butterfly must be multiplied by 0.5 * sqrt(2.0) */ |
ff90043f4a2d
in fact IDCT248 needs to be normalized as I suspected
bellard
parents:
719
diff
changeset
|
407 #define C_SHIFT (4+1+12) |
719 | 408 |
409 static inline void idct4col(UINT8 *dest, int line_size, const INT16 *col) | |
410 { | |
411 int c0, c1, c2, c3, a0, a1, a2, a3; | |
412 const UINT8 *cm = cropTbl + MAX_NEG_CROP; | |
413 | |
414 a0 = col[8*0]; | |
415 a1 = col[8*2]; | |
416 a2 = col[8*4]; | |
417 a3 = col[8*6]; | |
722
ff90043f4a2d
in fact IDCT248 needs to be normalized as I suspected
bellard
parents:
719
diff
changeset
|
418 c0 = ((a0 + a2) << (CN_SHIFT - 1)) + (1 << (C_SHIFT - 1)); |
ff90043f4a2d
in fact IDCT248 needs to be normalized as I suspected
bellard
parents:
719
diff
changeset
|
419 c2 = ((a0 - a2) << (CN_SHIFT - 1)) + (1 << (C_SHIFT - 1)); |
719 | 420 c1 = a1 * C1 + a3 * C2; |
421 c3 = a1 * C2 - a3 * C1; | |
422 dest[0] = cm[(c0 + c1) >> C_SHIFT]; | |
423 dest += line_size; | |
424 dest[0] = cm[(c2 + c3) >> C_SHIFT]; | |
425 dest += line_size; | |
426 dest[0] = cm[(c2 - c3) >> C_SHIFT]; | |
427 dest += line_size; | |
428 dest[0] = cm[(c0 - c1) >> C_SHIFT]; | |
429 } | |
430 | |
431 #define BF(k) \ | |
432 {\ | |
433 int a0, a1;\ | |
434 a0 = ptr[k];\ | |
435 a1 = ptr[8 + k];\ | |
436 ptr[k] = a0 + a1;\ | |
437 ptr[8 + k] = a0 - a1;\ | |
438 } | |
439 | |
440 /* only used by DV codec. The input must be interlaced. 128 is added | |
441 to the pixels before clamping to avoid systematic error | |
442 (1024*sqrt(2)) offset would be needed otherwise. */ | |
443 /* XXX: I think a 1.0/sqrt(2) normalization should be needed to | |
444 compensate the extra butterfly stage - I don't have the full DV | |
445 specification */ | |
446 void simple_idct248_put(UINT8 *dest, int line_size, INT16 *block) | |
447 { | |
448 int i; | |
449 INT16 *ptr; | |
450 | |
451 /* butterfly */ | |
452 ptr = block; | |
453 for(i=0;i<4;i++) { | |
454 BF(0); | |
455 BF(1); | |
456 BF(2); | |
457 BF(3); | |
458 BF(4); | |
459 BF(5); | |
460 BF(6); | |
461 BF(7); | |
462 ptr += 2 * 8; | |
463 } | |
464 | |
465 /* IDCT8 on each line */ | |
466 for(i=0; i<8; i++) { | |
467 idctRowCondDC(block + i*8); | |
468 } | |
469 | |
470 /* IDCT4 and store */ | |
471 for(i=0;i<8;i++) { | |
472 idct4col(dest + i, 2 * line_size, block + i); | |
473 idct4col(dest + line_size + i, 2 * line_size, block + 8 + i); | |
474 } | |
475 } |