Reduce number of ASM constraints for ff_lpc_compute_autocorr_sse2 since it
[libav.git] / libavcodec / x86 / lpc_mmx.c
1 /*
2 * MMX optimized LPC DSP utils
3 * Copyright (c) 2007 Loren Merritt
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #include "libavutil/x86_cpu.h"
23 #include "dsputil_mmx.h"
24
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;
29 x86_reg i = -n2*sizeof(int32_t);
30 x86_reg j = n2*sizeof(int32_t);
31 __asm__ volatile(
32 "movsd %0, %%xmm7 \n\t"
33 "movapd "MANGLE(ff_pd_1)", %%xmm6 \n\t"
34 "movapd "MANGLE(ff_pd_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)
39 );
40 #define WELCH(MOVPD, offset)\
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"\
48 "cvtpi2pd (%3,%0), %%xmm2 \n\t"\
49 "cvtpi2pd "#offset"*4(%3,%1), %%xmm3 \n\t"\
50 "mulpd %%xmm0, %%xmm2 \n\t"\
51 "mulpd %%xmm1, %%xmm3 \n\t"\
52 "movapd %%xmm2, (%2,%0,2) \n\t"\
53 MOVPD" %%xmm3, "#offset"*8(%2,%1,2) \n\t"\
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)\
59 :"r"(w_data+n2), "r"(data+n2)\
60 );
61 if(len&1)
62 WELCH("movupd", -1)
63 else
64 WELCH("movapd", -2)
65 #undef WELCH
66 }
67
68 void ff_lpc_compute_autocorr_sse2(const int32_t *data, int len, int lag,
69 double *autoc)
70 {
71 double tmp[len + lag + 2];
72 double *data1 = tmp + lag;
73 int j;
74
75 if((x86_reg)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 x86_reg i = -len*sizeof(double);
86 if(j == lag-2) {
87 __asm__ volatile(
88 "movsd "MANGLE(ff_pd_1)", %%xmm0 \n\t"
89 "movsd "MANGLE(ff_pd_1)", %%xmm1 \n\t"
90 "movsd "MANGLE(ff_pd_1)", %%xmm2 \n\t"
91 "1: \n\t"
92 "movapd (%2,%0), %%xmm3 \n\t"
93 "movupd -8(%3,%0), %%xmm4 \n\t"
94 "movapd (%3,%0), %%xmm5 \n\t"
95 "mulpd %%xmm3, %%xmm4 \n\t"
96 "mulpd %%xmm3, %%xmm5 \n\t"
97 "mulpd -16(%3,%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, 8(%1) \n\t"
111 "movsd %%xmm2, 16(%1) \n\t"
112 :"+&r"(i)
113 :"r"(autoc+j), "r"(data1+len), "r"(data1+len-j)
114 :"memory"
115 );
116 } else {
117 __asm__ volatile(
118 "movsd "MANGLE(ff_pd_1)", %%xmm0 \n\t"
119 "movsd "MANGLE(ff_pd_1)", %%xmm1 \n\t"
120 "1: \n\t"
121 "movapd (%3,%0), %%xmm3 \n\t"
122 "movupd -8(%4,%0), %%xmm4 \n\t"
123 "mulpd %%xmm3, %%xmm4 \n\t"
124 "mulpd (%4,%0), %%xmm3 \n\t"
125 "addpd %%xmm4, %%xmm1 \n\t"
126 "addpd %%xmm3, %%xmm0 \n\t"
127 "add $16, %0 \n\t"
128 "jl 1b \n\t"
129 "movhlps %%xmm0, %%xmm3 \n\t"
130 "movhlps %%xmm1, %%xmm4 \n\t"
131 "addsd %%xmm3, %%xmm0 \n\t"
132 "addsd %%xmm4, %%xmm1 \n\t"
133 "movsd %%xmm0, %1 \n\t"
134 "movsd %%xmm1, %2 \n\t"
135 :"+&r"(i), "=m"(autoc[j]), "=m"(autoc[j+1])
136 :"r"(data1+len), "r"(data1+len-j)
137 );
138 }
139 }
140 }