Originally committed as revision 2415 to svn://svn.ffmpeg.org/ffmpeg/trunk
Michael Niedermayer authored on 2003/10/23 03:26:38... | ... |
@@ -18,7 +18,7 @@ OBJS= common.o utils.o mem.o allcodecs.o \ |
18 | 18 |
fft.o mdct.o mace.o huffyuv.o cyuv.o opts.o raw.o h264.o golomb.o \ |
19 | 19 |
vp3.o asv1.o 4xm.o cabac.o ffv1.o ra144.o ra288.o vcr1.o cljr.o \ |
20 | 20 |
roqvideo.o dpcm.o interplayvideo.o xan.o rpza.o cinepak.o msrle.o \ |
21 |
- msvideo1.o vqavideo.o idcinvideo.o adx.o rational.o |
|
21 |
+ msvideo1.o vqavideo.o idcinvideo.o adx.o rational.o faandct.o |
|
22 | 22 |
|
23 | 23 |
ifeq ($(AMR_NB),yes) |
24 | 24 |
ifeq ($(AMR_NB_FIXED),yes) |
... | ... |
@@ -14,6 +14,7 @@ |
14 | 14 |
|
15 | 15 |
#include "i386/mmx.h" |
16 | 16 |
#include "simple_idct.h" |
17 |
+#include "faandct.h" |
|
17 | 18 |
|
18 | 19 |
#ifndef MAX |
19 | 20 |
#define MAX(a, b) (((a) > (b)) ? (a) : (b)) |
... | ... |
@@ -479,6 +480,7 @@ int main(int argc, char **argv) |
479 | 479 |
dct_error("IJG-AAN-INT", 0, fdct_ifast, fdct, test); |
480 | 480 |
dct_error("IJG-LLM-INT", 0, ff_jpeg_fdct_islow, fdct, test); |
481 | 481 |
dct_error("MMX", 0, ff_fdct_mmx, fdct, test); |
482 |
+ dct_error("FAAN", 0, ff_faandct, fdct, test); |
|
482 | 483 |
} else { |
483 | 484 |
dct_error("REF-DBL", 1, idct, idct, test); |
484 | 485 |
dct_error("INT", 1, j_rev_dct, idct, test); |
... | ... |
@@ -28,7 +28,7 @@ |
28 | 28 |
#include "dsputil.h" |
29 | 29 |
#include "mpegvideo.h" |
30 | 30 |
#include "simple_idct.h" |
31 |
- |
|
31 |
+#include "faandct.h" |
|
32 | 32 |
|
33 | 33 |
uint8_t cropTbl[256 + 2 * MAX_NEG_CROP]; |
34 | 34 |
uint32_t squareTbl[512]; |
... | ... |
@@ -2871,6 +2871,8 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx) |
2871 | 2871 |
#ifdef CONFIG_ENCODERS |
2872 | 2872 |
if(avctx->dct_algo==FF_DCT_FASTINT) |
2873 | 2873 |
c->fdct = fdct_ifast; |
2874 |
+ else if(avctx->dct_algo==FF_DCT_FAAN) |
|
2875 |
+ c->fdct = ff_faandct; |
|
2874 | 2876 |
else |
2875 | 2877 |
c->fdct = ff_jpeg_fdct_islow; //slow/accurate/default |
2876 | 2878 |
#endif //CONFIG_ENCODERS |
2877 | 2879 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,156 @@ |
0 |
+/* |
|
1 |
+ * Floating point AAN DCT |
|
2 |
+ * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at> |
|
3 |
+ * |
|
4 |
+ * This library is free software; you can redistribute it and/or |
|
5 |
+ * modify it under the terms of the GNU Lesser General Public |
|
6 |
+ * License as published by the Free Software Foundation; either |
|
7 |
+ * version 2 of the License, or (at your option) any later version. |
|
8 |
+ * |
|
9 |
+ * This library is distributed in the hope that it will be useful, |
|
10 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
11 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
12 |
+ * Lesser General Public License for more details. |
|
13 |
+ * |
|
14 |
+ * You should have received a copy of the GNU Lesser General Public |
|
15 |
+ * License along with this library; if not, write to the Free Software |
|
16 |
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
|
17 |
+ * |
|
18 |
+ */ |
|
19 |
+ |
|
20 |
+/** |
|
21 |
+ * @file faandct.c |
|
22 |
+ * @brief |
|
23 |
+ * Floating point AAN DCT |
|
24 |
+ * @author Michael Niedermayer <michaelni@gmx.at> |
|
25 |
+ */ |
|
26 |
+ |
|
27 |
+#include <math.h> |
|
28 |
+ |
|
29 |
+#include "dsputil.h" |
|
30 |
+#include "faandct.h" |
|
31 |
+ |
|
32 |
+#define FLOAT float |
|
33 |
+ |
|
34 |
+//numbers generated by simple c code (not as accurate as they could be) |
|
35 |
+/* |
|
36 |
+for(i=0; i<8; i++){ |
|
37 |
+ printf("#define B%d %1.20llf\n", i, (long double)1.0/(cosl(i*acosl(-1.0)/(long double)16.0)*sqrtl(2))); |
|
38 |
+} |
|
39 |
+*/ |
|
40 |
+#define B0 1.00000000000000000000 |
|
41 |
+#define B1 0.72095982200694791383 // (cos(pi*1/16)sqrt(2))^-1 |
|
42 |
+#define B2 0.76536686473017954350 // (cos(pi*2/16)sqrt(2))^-1 |
|
43 |
+#define B3 0.85043009476725644878 // (cos(pi*3/16)sqrt(2))^-1 |
|
44 |
+#define B4 1.00000000000000000000 // (cos(pi*4/16)sqrt(2))^-1 |
|
45 |
+#define B5 1.27275858057283393842 // (cos(pi*5/16)sqrt(2))^-1 |
|
46 |
+#define B6 1.84775906502257351242 // (cos(pi*6/16)sqrt(2))^-1 |
|
47 |
+#define B7 3.62450978541155137218 // (cos(pi*7/16)sqrt(2))^-1 |
|
48 |
+ |
|
49 |
+ |
|
50 |
+#define A1 0.70710678118654752438 // cos(pi*4/16) |
|
51 |
+#define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2) |
|
52 |
+#define A5 0.38268343236508977170 // cos(pi*6/16) |
|
53 |
+#define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2) |
|
54 |
+ |
|
55 |
+static FLOAT postscale[64]={ |
|
56 |
+B0*B0, B0*B1, B0*B2, B0*B3, B0*B4, B0*B5, B0*B6, B0*B7, |
|
57 |
+B1*B0, B1*B1, B1*B2, B1*B3, B1*B4, B1*B5, B1*B6, B1*B7, |
|
58 |
+B2*B0, B2*B1, B2*B2, B2*B3, B2*B4, B2*B5, B2*B6, B2*B7, |
|
59 |
+B3*B0, B3*B1, B3*B2, B3*B3, B3*B4, B3*B5, B3*B6, B3*B7, |
|
60 |
+B4*B0, B4*B1, B4*B2, B4*B3, B4*B4, B4*B5, B4*B6, B4*B7, |
|
61 |
+B5*B0, B5*B1, B5*B2, B5*B3, B5*B4, B5*B5, B5*B6, B5*B7, |
|
62 |
+B6*B0, B6*B1, B6*B2, B6*B3, B6*B4, B6*B5, B6*B6, B6*B7, |
|
63 |
+B7*B0, B7*B1, B7*B2, B7*B3, B7*B4, B7*B5, B7*B6, B7*B7, |
|
64 |
+}; |
|
65 |
+ |
|
66 |
+void ff_faandct(DCTELEM * data) |
|
67 |
+{ |
|
68 |
+ FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; |
|
69 |
+ FLOAT tmp10, tmp11, tmp12, tmp13; |
|
70 |
+ FLOAT z1, z2, z3, z4, z5, z11, z13; |
|
71 |
+ FLOAT temp[64]; |
|
72 |
+ int i; |
|
73 |
+ |
|
74 |
+ emms(); |
|
75 |
+ |
|
76 |
+ for (i=0; i<8*8; i+=8) { |
|
77 |
+ tmp0= data[0 + i] + data[7 + i]; |
|
78 |
+ tmp7= data[0 + i] - data[7 + i]; |
|
79 |
+ tmp1= data[1 + i] + data[6 + i]; |
|
80 |
+ tmp6= data[1 + i] - data[6 + i]; |
|
81 |
+ tmp2= data[2 + i] + data[5 + i]; |
|
82 |
+ tmp5= data[2 + i] - data[5 + i]; |
|
83 |
+ tmp3= data[3 + i] + data[4 + i]; |
|
84 |
+ tmp4= data[3 + i] - data[4 + i]; |
|
85 |
+ |
|
86 |
+ tmp10= tmp0 + tmp3; |
|
87 |
+ tmp13= tmp0 - tmp3; |
|
88 |
+ tmp11= tmp1 + tmp2; |
|
89 |
+ tmp12= tmp1 - tmp2; |
|
90 |
+ |
|
91 |
+ temp[0 + i]= tmp10 + tmp11; |
|
92 |
+ temp[4 + i]= tmp10 - tmp11; |
|
93 |
+ |
|
94 |
+ z1= (tmp12 + tmp13)*A1; |
|
95 |
+ temp[2 + i]= tmp13 + z1; |
|
96 |
+ temp[6 + i]= tmp13 - z1; |
|
97 |
+ |
|
98 |
+ tmp10= tmp4 + tmp5; |
|
99 |
+ tmp11= tmp5 + tmp6; |
|
100 |
+ tmp12= tmp6 + tmp7; |
|
101 |
+ |
|
102 |
+ z5= (tmp10 - tmp12) * A5; |
|
103 |
+ z2= tmp10*A2 + z5; |
|
104 |
+ z4= tmp12*A4 + z5; |
|
105 |
+ z3= tmp11*A1; |
|
106 |
+ |
|
107 |
+ z11= tmp7 + z3; |
|
108 |
+ z13= tmp7 - z3; |
|
109 |
+ |
|
110 |
+ temp[5 + i]= z13 + z2; |
|
111 |
+ temp[3 + i]= z13 - z2; |
|
112 |
+ temp[1 + i]= z11 + z4; |
|
113 |
+ temp[7 + i]= z11 - z4; |
|
114 |
+ } |
|
115 |
+ |
|
116 |
+ for (i=0; i<8; i++) { |
|
117 |
+ tmp0= temp[8*0 + i] + temp[8*7 + i]; |
|
118 |
+ tmp7= temp[8*0 + i] - temp[8*7 + i]; |
|
119 |
+ tmp1= temp[8*1 + i] + temp[8*6 + i]; |
|
120 |
+ tmp6= temp[8*1 + i] - temp[8*6 + i]; |
|
121 |
+ tmp2= temp[8*2 + i] + temp[8*5 + i]; |
|
122 |
+ tmp5= temp[8*2 + i] - temp[8*5 + i]; |
|
123 |
+ tmp3= temp[8*3 + i] + temp[8*4 + i]; |
|
124 |
+ tmp4= temp[8*3 + i] - temp[8*4 + i]; |
|
125 |
+ |
|
126 |
+ tmp10= tmp0 + tmp3; |
|
127 |
+ tmp13= tmp0 - tmp3; |
|
128 |
+ tmp11= tmp1 + tmp2; |
|
129 |
+ tmp12= tmp1 - tmp2; |
|
130 |
+ |
|
131 |
+ data[8*0 + i]= lrint(postscale[8*0 + i] * (tmp10 + tmp11)); |
|
132 |
+ data[8*4 + i]= lrint(postscale[8*4 + i] * (tmp10 - tmp11)); |
|
133 |
+ |
|
134 |
+ z1= (tmp12 + tmp13)* A1; |
|
135 |
+ data[8*2 + i]= lrint(postscale[8*2 + i] * (tmp13 + z1)); |
|
136 |
+ data[8*6 + i]= lrint(postscale[8*6 + i] * (tmp13 - z1)); |
|
137 |
+ |
|
138 |
+ tmp10= tmp4 + tmp5; |
|
139 |
+ tmp11= tmp5 + tmp6; |
|
140 |
+ tmp12= tmp6 + tmp7; |
|
141 |
+ |
|
142 |
+ z5= (tmp10 - tmp12) * A5; |
|
143 |
+ z2= tmp10*A2 + z5; |
|
144 |
+ z4= tmp12*A4 + z5; |
|
145 |
+ z3= tmp11*A1; |
|
146 |
+ |
|
147 |
+ z11= tmp7 + z3; |
|
148 |
+ z13= tmp7 - z3; |
|
149 |
+ |
|
150 |
+ data[8*5 + i]= lrint(postscale[8*5 + i] * (z13 + z2)); |
|
151 |
+ data[8*3 + i]= lrint(postscale[8*3 + i] * (z13 - z2)); |
|
152 |
+ data[8*1 + i]= lrint(postscale[8*1 + i] * (z11 + z4)); |
|
153 |
+ data[8*7 + i]= lrint(postscale[8*7 + i] * (z11 - z4)); |
|
154 |
+ } |
|
155 |
+} |
0 | 156 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,28 @@ |
0 |
+/* |
|
1 |
+ * Floating point AAN DCT |
|
2 |
+ * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at> |
|
3 |
+ * |
|
4 |
+ * This library is free software; you can redistribute it and/or |
|
5 |
+ * modify it under the terms of the GNU Lesser General Public |
|
6 |
+ * License as published by the Free Software Foundation; either |
|
7 |
+ * version 2 of the License, or (at your option) any later version. |
|
8 |
+ * |
|
9 |
+ * This library is distributed in the hope that it will be useful, |
|
10 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
11 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
12 |
+ * Lesser General Public License for more details. |
|
13 |
+ * |
|
14 |
+ * You should have received a copy of the GNU Lesser General Public |
|
15 |
+ * License along with this library; if not, write to the Free Software |
|
16 |
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
|
17 |
+ * |
|
18 |
+ */ |
|
19 |
+ |
|
20 |
+/** |
|
21 |
+ * @file faandct.h |
|
22 |
+ * @brief |
|
23 |
+ * Floating point AAN DCT |
|
24 |
+ * @author Michael Niedermayer <michaelni@gmx.at> |
|
25 |
+ */ |
|
26 |
+ |
|
27 |
+void ff_faandct(DCTELEM * data); |
... | ... |
@@ -28,6 +28,7 @@ |
28 | 28 |
#include "avcodec.h" |
29 | 29 |
#include "dsputil.h" |
30 | 30 |
#include "mpegvideo.h" |
31 |
+#include "faandct.h" |
|
31 | 32 |
|
32 | 33 |
#ifdef USE_FASTMEMCPY |
33 | 34 |
#include "fastmemcpy.h" |
... | ... |
@@ -100,7 +101,7 @@ static void convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16 |
100 | 100 |
|
101 | 101 |
for(qscale=qmin; qscale<=qmax; qscale++){ |
102 | 102 |
int i; |
103 |
- if (s->dsp.fdct == ff_jpeg_fdct_islow) { |
|
103 |
+ if (s->dsp.fdct == ff_jpeg_fdct_islow || s->dsp.fdct == ff_faandct) { |
|
104 | 104 |
for(i=0;i<64;i++) { |
105 | 105 |
const int j= s->dsp.idct_permutation[i]; |
106 | 106 |
/* 16 <= qscale * quant_matrix[i] <= 7905 */ |