Mercurial > libavcodec.hg
annotate i386/flacdsp_mmx.c @ 7110:f9ebb1553982 libavcodec
More comments removal/fixes
author | vitor |
---|---|
date | Mon, 23 Jun 2008 20:28:00 +0000 |
parents | f7cbb7733146 |
children | 30e3a8465436 |
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 | |
6763 | 22 #include "libavutil/x86_cpu.h" |
5946
55251379b5b1
make ff_p* vars extern so that they can be used in various *_mmx.c files
aurel
parents:
5933
diff
changeset
|
23 #include "dsputil_mmx.h" |
2967 | 24 |
5737 | 25 static void apply_welch_window_sse2(const int32_t *data, int len, double *w_data) |
26 { | |
27 double c = 2.0 / (len-1.0); | |
28 int n2 = len>>1; | |
6755
33896780c612
Do not misuse long as the size of a register in x86.
ramiro
parents:
6339
diff
changeset
|
29 x86_reg i = -n2*sizeof(int32_t); |
33896780c612
Do not misuse long as the size of a register in x86.
ramiro
parents:
6339
diff
changeset
|
30 x86_reg j = n2*sizeof(int32_t); |
5737 | 31 asm volatile( |
32 "movsd %0, %%xmm7 \n\t" | |
33 "movapd %1, %%xmm6 \n\t" | |
34 "movapd %2, %%xmm5 \n\t" | |
35 "movlhps %%xmm7, %%xmm7 \n\t" | |
36 "subpd %%xmm5, %%xmm7 \n\t" | |
37 "addsd %%xmm6, %%xmm7 \n\t" | |
38 ::"m"(c), "m"(*ff_pd_1), "m"(*ff_pd_2) | |
39 ); | |
6339 | 40 #define WELCH(MOVPD, offset)\ |
5737 | 41 asm volatile(\ |
42 "1: \n\t"\ | |
43 "movapd %%xmm7, %%xmm1 \n\t"\ | |
44 "mulpd %%xmm1, %%xmm1 \n\t"\ | |
45 "movapd %%xmm6, %%xmm0 \n\t"\ | |
46 "subpd %%xmm1, %%xmm0 \n\t"\ | |
47 "pshufd $0x4e, %%xmm0, %%xmm1 \n\t"\ | |
6339 | 48 "cvtpi2pd (%3,%0), %%xmm2 \n\t"\ |
49 "cvtpi2pd "#offset"*4(%3,%1), %%xmm3 \n\t"\ | |
5737 | 50 "mulpd %%xmm0, %%xmm2 \n\t"\ |
51 "mulpd %%xmm1, %%xmm3 \n\t"\ | |
52 "movapd %%xmm2, (%2,%0,2) \n\t"\ | |
6339 | 53 MOVPD" %%xmm3, "#offset"*8(%2,%1,2) \n\t"\ |
5737 | 54 "subpd %%xmm5, %%xmm7 \n\t"\ |
55 "sub $8, %1 \n\t"\ | |
56 "add $8, %0 \n\t"\ | |
57 "jl 1b \n\t"\ | |
58 :"+&r"(i), "+&r"(j)\ | |
6339 | 59 :"r"(w_data+n2), "r"(data+n2)\ |
5737 | 60 ); |
61 if(len&1) | |
6339 | 62 WELCH("movupd", -1) |
5737 | 63 else |
6339 | 64 WELCH("movapd", -2) |
5737 | 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 | |
6755
33896780c612
Do not misuse long as the size of a register in x86.
ramiro
parents:
6339
diff
changeset
|
75 if((x86_reg)data1 & 15) |
5737 | 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){ | |
6755
33896780c612
Do not misuse long as the size of a register in x86.
ramiro
parents:
6339
diff
changeset
|
85 x86_reg i = -len*sizeof(double); |
5737 | 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 } |