1 /***
2 This file is part of PulseAudio.
3
4 Copyright 2004-2006 Lennart Poettering
5 Copyright 2009 Wim Taymans <wim.taymans@collabora.co.uk.com>
6
7 PulseAudio is free software; you can redistribute it and/or modify
8 it under the terms of the GNU Lesser General Public License as published
9 by the Free Software Foundation; either version 2.1 of the License,
10 or (at your option) any later version.
11
12 PulseAudio is distributed in the hope that it will be useful, but
13 WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 General Public License for more details.
16
17 You should have received a copy of the GNU Lesser General Public License
18 along with PulseAudio; if not, see <http://www.gnu.org/licenses/>.
19 ***/
20
21 #ifdef HAVE_CONFIG_H
22 #include <config.h>
23 #endif
24
25 #include <pulse/sample.h>
26 #include <pulse/volume.h>
27 #include <pulsecore/log.h>
28 #include <pulsecore/macro.h>
29
30 #include "cpu-x86.h"
31 #include "remap.h"
32
33 #define LOAD_SAMPLES \
34 " movq (%1), %%mm0 \n\t" \
35 " movq 8(%1), %%mm2 \n\t" \
36 " movq 16(%1), %%mm4 \n\t" \
37 " movq 24(%1), %%mm6 \n\t" \
38 " movq %%mm0, %%mm1 \n\t" \
39 " movq %%mm2, %%mm3 \n\t" \
40 " movq %%mm4, %%mm5 \n\t" \
41 " movq %%mm6, %%mm7 \n\t"
42
43 #define UNPACK_SAMPLES(s) \
44 " punpckl"#s" %%mm0, %%mm0 \n\t" \
45 " punpckh"#s" %%mm1, %%mm1 \n\t" \
46 " punpckl"#s" %%mm2, %%mm2 \n\t" \
47 " punpckh"#s" %%mm3, %%mm3 \n\t" \
48 " punpckl"#s" %%mm4, %%mm4 \n\t" \
49 " punpckh"#s" %%mm5, %%mm5 \n\t" \
50 " punpckl"#s" %%mm6, %%mm6 \n\t" \
51 " punpckh"#s" %%mm7, %%mm7 \n\t"
52
53 #define STORE_SAMPLES \
54 " movq %%mm0, (%0) \n\t" \
55 " movq %%mm1, 8(%0) \n\t" \
56 " movq %%mm2, 16(%0) \n\t" \
57 " movq %%mm3, 24(%0) \n\t" \
58 " movq %%mm4, 32(%0) \n\t" \
59 " movq %%mm5, 40(%0) \n\t" \
60 " movq %%mm6, 48(%0) \n\t" \
61 " movq %%mm7, 56(%0) \n\t" \
62 " add $32, %1 \n\t" \
63 " add $64, %0 \n\t"
64
65 #define HANDLE_SINGLE_dq() \
66 " movd (%1), %%mm0 \n\t" \
67 " punpckldq %%mm0, %%mm0 \n\t" \
68 " movq %%mm0, (%0) \n\t" \
69 " add $4, %1 \n\t" \
70 " add $8, %0 \n\t"
71
72 #define HANDLE_SINGLE_wd() \
73 " movw (%1), %w3 \n\t" \
74 " movd %3, %%mm0 \n\t" \
75 " punpcklwd %%mm0, %%mm0 \n\t" \
76 " movd %%mm0, (%0) \n\t" \
77 " add $2, %1 \n\t" \
78 " add $4, %0 \n\t"
79
80 #define MONO_TO_STEREO(s,shift,mask) \
81 " mov %4, %2 \n\t" \
82 " sar $"#shift", %2 \n\t" \
83 " cmp $0, %2 \n\t" \
84 " je 2f \n\t" \
85 "1: \n\t" \
86 LOAD_SAMPLES \
87 UNPACK_SAMPLES(s) \
88 STORE_SAMPLES \
89 " dec %2 \n\t" \
90 " jne 1b \n\t" \
91 "2: \n\t" \
92 " mov %4, %2 \n\t" \
93 " and $"#mask", %2 \n\t" \
94 " je 4f \n\t" \
95 "3: \n\t" \
96 HANDLE_SINGLE_##s() \
97 " dec %2 \n\t" \
98 " jne 3b \n\t" \
99 "4: \n\t" \
100 " emms \n\t"
101
102 #if defined (__i386__) || defined (__amd64__)
remap_mono_to_stereo_s16ne_mmx(pa_remap_t * m,int16_t * dst,const int16_t * src,unsigned n)103 static void remap_mono_to_stereo_s16ne_mmx(pa_remap_t *m, int16_t *dst, const int16_t *src, unsigned n) {
104 pa_reg_x86 temp, temp2;
105
106 __asm__ __volatile__ (
107 MONO_TO_STEREO(wd,4,15) /* do words to doubles */
108 : "+r" (dst), "+r" (src), "=&r" (temp), "=&r" (temp2)
109 : "r" ((pa_reg_x86)n)
110 : "cc"
111 );
112 }
113
114 /* Works for both S32NE and FLOAT32NE */
remap_mono_to_stereo_any32ne_mmx(pa_remap_t * m,float * dst,const float * src,unsigned n)115 static void remap_mono_to_stereo_any32ne_mmx(pa_remap_t *m, float *dst, const float *src, unsigned n) {
116 pa_reg_x86 temp, temp2;
117
118 __asm__ __volatile__ (
119 MONO_TO_STEREO(dq,3,7) /* do doubles to quads */
120 : "+r" (dst), "+r" (src), "=&r" (temp), "=&r" (temp2)
121 : "r" ((pa_reg_x86)n)
122 : "cc"
123 );
124 }
125
126 /* set the function that will execute the remapping based on the matrices */
init_remap_mmx(pa_remap_t * m)127 static void init_remap_mmx(pa_remap_t *m) {
128 unsigned n_oc, n_ic;
129
130 n_oc = m->o_ss.channels;
131 n_ic = m->i_ss.channels;
132
133 /* find some common channel remappings, fall back to full matrix operation. */
134 if (n_ic == 1 && n_oc == 2 &&
135 m->map_table_i[0][0] == 0x10000 && m->map_table_i[1][0] == 0x10000) {
136
137 pa_log_info("Using MMX mono to stereo remapping");
138 pa_set_remap_func(m, (pa_do_remap_func_t) remap_mono_to_stereo_s16ne_mmx,
139 (pa_do_remap_func_t) remap_mono_to_stereo_any32ne_mmx,
140 (pa_do_remap_func_t) remap_mono_to_stereo_any32ne_mmx);
141 }
142 }
143 #endif /* defined (__i386__) || defined (__amd64__) */
144
pa_remap_func_init_mmx(pa_cpu_x86_flag_t flags)145 void pa_remap_func_init_mmx(pa_cpu_x86_flag_t flags) {
146 #if defined (__i386__) || defined (__amd64__)
147
148 if (flags & PA_CPU_X86_MMX) {
149 pa_log_info("Initialising MMX optimized remappers.");
150
151 pa_set_init_remap_func((pa_init_remap_func_t) init_remap_mmx);
152 }
153
154 #endif /* defined (__i386__) || defined (__amd64__) */
155 }
156