Mercurial > libavcodec.hg
annotate i386/flacdsp_mmx.c @ 6243:930ac362be5a libavcodec
const
author | michael |
---|---|
date | Fri, 01 Feb 2008 13:55:15 +0000 |
parents | fb99890ee609 |
children | 0ea2b97aa9f6 |
rev | line source |
---|---|
0 | 1 /* |
6030 | 2 * MMX optimized FLAC DSP utils |
3 * Copyright (c) 2007 Loren Merritt | |
0 | 4 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3932
diff
changeset
|
5 * This file is part of FFmpeg. |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3932
diff
changeset
|
6 * |
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3932
diff
changeset
|
7 * FFmpeg is free software; you can redistribute it and/or |
429 | 8 * modify it under the terms of the GNU Lesser General Public |
9 * 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:
3932
diff
changeset
|
10 * version 2.1 of the License, or (at your option) any later version. |
0 | 11 * |
3947
c8c591fe26f8
Change license headers to say 'FFmpeg' instead of 'this program/this library'
diego
parents:
3932
diff
changeset
|
12 * FFmpeg is distributed in the hope that it will be useful, |
0 | 13 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
429 | 14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
15 * Lesser General Public License for more details. | |
0 | 16 * |
429 | 17 * 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:
3932
diff
changeset
|
18 * 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
|
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
0 | 20 */ |
21 | |
5946
55251379b5b1
make ff_p* vars extern so that they can be used in various *_mmx.c files
aurel
parents:
5933
diff
changeset
|
22 #include "dsputil_mmx.h" |
2967 | 23 |
5737 | 24 static void apply_welch_window_sse2(const int32_t *data, int len, double *w_data) |
25 { | |
26 double c = 2.0 / (len-1.0); | |
27 int n2 = len>>1; | |
28 long i = -n2*sizeof(int32_t); | |
29 long j = n2*sizeof(int32_t); | |
30 asm volatile( | |
31 "movsd %0, %%xmm7 \n\t" | |
32 "movapd %1, %%xmm6 \n\t" | |
33 "movapd %2, %%xmm5 \n\t" | |
34 "movlhps %%xmm7, %%xmm7 \n\t" | |
35 "subpd %%xmm5, %%xmm7 \n\t" | |
36 "addsd %%xmm6, %%xmm7 \n\t" | |
37 ::"m"(c), "m"(*ff_pd_1), "m"(*ff_pd_2) | |
38 ); | |
39 #define WELCH(MOVPD)\ | |
40 asm volatile(\ | |
41 "1: \n\t"\ | |
42 "movapd %%xmm7, %%xmm1 \n\t"\ | |
43 "mulpd %%xmm1, %%xmm1 \n\t"\ | |
44 "movapd %%xmm6, %%xmm0 \n\t"\ | |
45 "subpd %%xmm1, %%xmm0 \n\t"\ | |
46 "pshufd $0x4e, %%xmm0, %%xmm1 \n\t"\ | |
47 "cvtpi2pd (%4,%0), %%xmm2 \n\t"\ | |
48 "cvtpi2pd (%5,%1), %%xmm3 \n\t"\ | |
49 "mulpd %%xmm0, %%xmm2 \n\t"\ | |
50 "mulpd %%xmm1, %%xmm3 \n\t"\ | |
51 "movapd %%xmm2, (%2,%0,2) \n\t"\ | |
52 MOVPD" %%xmm3, (%3,%1,2) \n\t"\ | |
53 "subpd %%xmm5, %%xmm7 \n\t"\ | |
54 "sub $8, %1 \n\t"\ | |
55 "add $8, %0 \n\t"\ | |
56 "jl 1b \n\t"\ | |
57 :"+&r"(i), "+&r"(j)\ | |
58 :"r"(w_data+n2), "r"(w_data+len-2-n2),\ | |
59 "r"(data+n2), "r"(data+len-2-n2)\ | |
60 ); | |
61 if(len&1) | |
62 WELCH("movupd") | |
63 else | |
64 WELCH("movapd") | |
65 #undef WELCH | |
66 } | |
67 | |
6030 | 68 void ff_flac_compute_autocorr_sse2(const int32_t *data, int len, int lag, |
69 double *autoc) | |
5737 | 70 { |
71 double tmp[len + lag + 2]; | |
72 double *data1 = tmp + lag; | |
73 int j; | |
74 | |
75 if((long)data1 & 15) | |
76 data1++; | |
77 | |
78 apply_welch_window_sse2(data, len, data1); | |
79 | |
80 for(j=0; j<lag; j++) | |
81 data1[j-lag]= 0.0; | |
82 data1[len] = 0.0; | |
83 | |
84 for(j=0; j<lag; j+=2){ | |
85 long i = -len*sizeof(double); | |
86 if(j == lag-2) { | |
87 asm volatile( | |
88 "movsd %6, %%xmm0 \n\t" | |
89 "movsd %6, %%xmm1 \n\t" | |
90 "movsd %6, %%xmm2 \n\t" | |
91 "1: \n\t" | |
92 "movapd (%4,%0), %%xmm3 \n\t" | |
93 "movupd -8(%5,%0), %%xmm4 \n\t" | |
94 "movapd (%5,%0), %%xmm5 \n\t" | |
95 "mulpd %%xmm3, %%xmm4 \n\t" | |
96 "mulpd %%xmm3, %%xmm5 \n\t" | |
97 "mulpd -16(%5,%0), %%xmm3 \n\t" | |
98 "addpd %%xmm4, %%xmm1 \n\t" | |
99 "addpd %%xmm5, %%xmm0 \n\t" | |
100 "addpd %%xmm3, %%xmm2 \n\t" | |
101 "add $16, %0 \n\t" | |
102 "jl 1b \n\t" | |
103 "movhlps %%xmm0, %%xmm3 \n\t" | |
104 "movhlps %%xmm1, %%xmm4 \n\t" | |
105 "movhlps %%xmm2, %%xmm5 \n\t" | |
106 "addsd %%xmm3, %%xmm0 \n\t" | |
107 "addsd %%xmm4, %%xmm1 \n\t" | |
108 "addsd %%xmm5, %%xmm2 \n\t" | |
109 "movsd %%xmm0, %1 \n\t" | |
110 "movsd %%xmm1, %2 \n\t" | |
111 "movsd %%xmm2, %3 \n\t" | |
112 :"+&r"(i), "=m"(autoc[j]), "=m"(autoc[j+1]), "=m"(autoc[j+2]) | |
113 :"r"(data1+len), "r"(data1+len-j), "m"(*ff_pd_1) | |
114 ); | |
115 } else { | |
116 asm volatile( | |
117 "movsd %5, %%xmm0 \n\t" | |
118 "movsd %5, %%xmm1 \n\t" | |
119 "1: \n\t" | |
120 "movapd (%3,%0), %%xmm3 \n\t" | |
121 "movupd -8(%4,%0), %%xmm4 \n\t" | |
122 "mulpd %%xmm3, %%xmm4 \n\t" | |
123 "mulpd (%4,%0), %%xmm3 \n\t" | |
124 "addpd %%xmm4, %%xmm1 \n\t" | |
125 "addpd %%xmm3, %%xmm0 \n\t" | |
126 "add $16, %0 \n\t" | |
127 "jl 1b \n\t" | |
128 "movhlps %%xmm0, %%xmm3 \n\t" | |
129 "movhlps %%xmm1, %%xmm4 \n\t" | |
130 "addsd %%xmm3, %%xmm0 \n\t" | |
131 "addsd %%xmm4, %%xmm1 \n\t" | |
132 "movsd %%xmm0, %1 \n\t" | |
133 "movsd %%xmm1, %2 \n\t" | |
134 :"+&r"(i), "=m"(autoc[j]), "=m"(autoc[j+1]) | |
135 :"r"(data1+len), "r"(data1+len-j), "m"(*ff_pd_1) | |
136 ); | |
137 } | |
138 } | |
139 } |