code.delx.au - pulseaudio/blob - src/pulsecore/remap_mmx.c

   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, write to the Free Software
  19   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
  20   USA.
  21 ***/
  22
  23 #ifdef HAVE_CONFIG_H
  24 #include <config.h>
  25 #endif
  26
  27 #include <string.h>
  28
  29 #include <pulse/sample.h>
  30 #include <pulsecore/log.h>
  31 #include <pulsecore/macro.h>
  32
  33 #include "cpu-x86.h"
  34 #include "remap.h"
  35
  36 #define LOAD_SAMPLES                                   \
  37                 " movq (%1), %%mm0              \n\t"  \
  38                 " movq 8(%1), %%mm2             \n\t"  \
  39                 " movq 16(%1), %%mm4            \n\t"  \
  40                 " movq 24(%1), %%mm6            \n\t"  \
  41                 " movq %%mm0, %%mm1             \n\t"  \
  42                 " movq %%mm2, %%mm3             \n\t"  \
  43                 " movq %%mm4, %%mm5             \n\t"  \
  44                 " movq %%mm6, %%mm7             \n\t"
  45
  46 #define UNPACK_SAMPLES(s)                              \
  47                 " punpckl"#s" %%mm0, %%mm0      \n\t"  \
  48                 " punpckh"#s" %%mm1, %%mm1      \n\t"  \
  49                 " punpckl"#s" %%mm2, %%mm2      \n\t"  \
  50                 " punpckh"#s" %%mm3, %%mm3      \n\t"  \
  51                 " punpckl"#s" %%mm4, %%mm4      \n\t"  \
  52                 " punpckh"#s" %%mm5, %%mm5      \n\t"  \
  53                 " punpckl"#s" %%mm6, %%mm6      \n\t"  \
  54                 " punpckh"#s" %%mm7, %%mm7      \n\t"  \
  55
  56 #define STORE_SAMPLES                                  \
  57                 " movq %%mm0, (%0)              \n\t"  \
  58                 " movq %%mm1, 8(%0)             \n\t"  \
  59                 " movq %%mm2, 16(%0)            \n\t"  \
  60                 " movq %%mm3, 24(%0)            \n\t"  \
  61                 " movq %%mm4, 32(%0)            \n\t"  \
  62                 " movq %%mm5, 40(%0)            \n\t"  \
  63                 " movq %%mm6, 48(%0)            \n\t"  \
  64                 " movq %%mm7, 56(%0)            \n\t"  \
  65                 " add $32, %1                   \n\t"  \
  66                 " add $64, %0                   \n\t"
  67
  68 #define HANDLE_SINGLE(s)                               \
  69                 " movd (%1), %%mm0              \n\t"  \
  70                 " movq %%mm0, %%mm1             \n\t"  \
  71                 " punpckl"#s" %%mm0, %%mm0      \n\t"  \
  72                 " movq %%mm0, (%0)              \n\t"  \
  73                 " add $4, %1                    \n\t"  \
  74                 " add $8, %0                    \n\t"
  75
  76 static void remap_mono_to_stereo_mmx (pa_remap_t *m, void *dst, const void *src, unsigned n) {
  77     pa_reg_x86 temp;
  78
  79     switch (*m->format) {
  80         case PA_SAMPLE_FLOAT32NE:
  81         {
  82             __asm__ __volatile__ (
  83                 " mov %3, %2                    \n\t"
  84                 " sar $3, %2                    \n\t" /* prepare for processing 8 samples at a time */
  85                 " cmp $0, %2                    \n\t"
  86                 " je 2f                         \n\t"
  87
  88                 "1:                             \n\t" /* do samples in groups of 8 */
  89                 LOAD_SAMPLES
  90                 UNPACK_SAMPLES(dq)
  91                 STORE_SAMPLES
  92                 " dec %2                        \n\t"
  93                 " jne 1b                        \n\t"
  94
  95                 "2:                             \n\t"
  96                 " mov %3, %2                    \n\t"
  97                 " and $7, %2                    \n\t" /* prepare for processing the remaining samples */
  98                 " je 4f                         \n\t"
  99
 100                 "3:                             \n\t"
 101                 HANDLE_SINGLE(dq)
 102                 " dec %2                        \n\t"
 103                 " jne 3b                        \n\t"
 104
 105                 "4:                             \n\t"
 106                 " emms                          \n\t"
 107
 108                 : "+r" (dst), "+r" (src), "=&r" (temp)
 109                 : "r" ((pa_reg_x86)n)
 110                 : "cc"
 111             );
 112             break;
 113         }
 114         case PA_SAMPLE_S16NE:
 115         {
 116             __asm__ __volatile__ (
 117                 " mov %3, %2                    \n\t"
 118                 " sar $3, %2                    \n\t" /* prepare for processing 8 samples at a time */
 119                 " cmp $0, %2                    \n\t"
 120                 " je 2f                         \n\t"
 121
 122                 "1:                             \n\t" /* do samples in groups of 16 */
 123                 LOAD_SAMPLES
 124                 UNPACK_SAMPLES(wd)
 125                 STORE_SAMPLES
 126                 " dec %2                        \n\t"
 127                 " jne 1b                        \n\t"
 128
 129                 "2:                             \n\t"
 130                 " mov %3, %2                    \n\t"
 131                 " and $7, %2                    \n\t" /* prepare for processing the remaining samples */
 132                 " je 4f                         \n\t"
 133
 134                 "3:                             \n\t"
 135                 HANDLE_SINGLE(wd)
 136                 " dec %2                        \n\t"
 137                 " jne 3b                        \n\t"
 138
 139                 "4:                             \n\t"
 140                 " emms                          \n\t"
 141
 142                 : "+r" (dst), "+r" (src), "=&r" (temp)
 143                 : "r" ((pa_reg_x86)n)
 144                 : "cc"
 145             );
 146             break;
 147         }
 148         default:
 149             pa_assert_not_reached();
 150     }
 151 }
 152
 153 /* set the function that will execute the remapping based on the matrices */
 154 static void init_remap_mmx (pa_remap_t *m) {
 155     unsigned n_oc, n_ic;
 156
 157     n_oc = m->o_ss->channels;
 158     n_ic = m->i_ss->channels;
 159
 160     /* find some common channel remappings, fall back to full matrix operation. */
 161     if (n_ic == 1 && n_oc == 2 &&
 162             m->map_table_f[0][0] >= 1.0 && m->map_table_f[1][0] >= 1.0) {
 163         m->do_remap = (pa_do_remap_func_t) remap_mono_to_stereo_mmx;
 164         pa_log_info("Using MMX mono to stereo remapping");
 165     }
 166 }
 167
 168 void pa_remap_func_init_mmx (pa_cpu_x86_flag_t flags) {
 169 #if defined (__i386__) || defined (__amd64__)
 170     pa_log_info("Initialising MMX optimized remappers.");
 171
 172     pa_set_init_remap_func ((pa_init_remap_func_t) init_remap_mmx);
 173 #endif /* defined (__i386__) || defined (__amd64__) */
 174 }