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 <pulse/sample.h>
  28 #include <pulse/volume.h>
  29 #include <pulsecore/log.h>
  30 #include <pulsecore/macro.h>
  31
  32 #include "cpu-x86.h"
  33 #include "remap.h"
  34
  35 #define LOAD_SAMPLES                                   \
  36                 " movq (%1), %%mm0              \n\t"  \
  37                 " movq 8(%1), %%mm2             \n\t"  \
  38                 " movq 16(%1), %%mm4            \n\t"  \
  39                 " movq 24(%1), %%mm6            \n\t"  \
  40                 " movq %%mm0, %%mm1             \n\t"  \
  41                 " movq %%mm2, %%mm3             \n\t"  \
  42                 " movq %%mm4, %%mm5             \n\t"  \
  43                 " movq %%mm6, %%mm7             \n\t"
  44
  45 #define UNPACK_SAMPLES(s)                              \
  46                 " punpckl"#s" %%mm0, %%mm0      \n\t"  \
  47                 " punpckh"#s" %%mm1, %%mm1      \n\t"  \
  48                 " punpckl"#s" %%mm2, %%mm2      \n\t"  \
  49                 " punpckh"#s" %%mm3, %%mm3      \n\t"  \
  50                 " punpckl"#s" %%mm4, %%mm4      \n\t"  \
  51                 " punpckh"#s" %%mm5, %%mm5      \n\t"  \
  52                 " punpckl"#s" %%mm6, %%mm6      \n\t"  \
  53                 " punpckh"#s" %%mm7, %%mm7      \n\t"
  54
  55 #define STORE_SAMPLES                                  \
  56                 " movq %%mm0, (%0)              \n\t"  \
  57                 " movq %%mm1, 8(%0)             \n\t"  \
  58                 " movq %%mm2, 16(%0)            \n\t"  \
  59                 " movq %%mm3, 24(%0)            \n\t"  \
  60                 " movq %%mm4, 32(%0)            \n\t"  \
  61                 " movq %%mm5, 40(%0)            \n\t"  \
  62                 " movq %%mm6, 48(%0)            \n\t"  \
  63                 " movq %%mm7, 56(%0)            \n\t"  \
  64                 " add $32, %1                   \n\t"  \
  65                 " add $64, %0                   \n\t"
  66
  67 #define HANDLE_SINGLE_dq()                            \
  68                 " movd (%1), %%mm0              \n\t"  \
  69                 " punpckldq %%mm0, %%mm0        \n\t"  \
  70                 " movq %%mm0, (%0)              \n\t"  \
  71                 " add $4, %1                    \n\t"  \
  72                 " add $8, %0                    \n\t"
  73
  74 #define HANDLE_SINGLE_wd()                             \
  75                 " movw (%1), %w3                \n\t"  \
  76                 " movd %3,  %%mm0               \n\t"  \
  77                 " punpcklwd %%mm0, %%mm0        \n\t"  \
  78                 " movd %%mm0, (%0)              \n\t"  \
  79                 " add $2, %1                    \n\t"  \
  80                 " add $4, %0                    \n\t"
  81
  82 #define MONO_TO_STEREO(s,shift,mask)                   \
  83                 " mov %4, %2                    \n\t"  \
  84                 " sar $"#shift", %2             \n\t"  \
  85                 " cmp $0, %2                    \n\t"  \
  86                 " je 2f                         \n\t"  \
  87                 "1:                             \n\t"  \
  88                 LOAD_SAMPLES                           \
  89                 UNPACK_SAMPLES(s)                      \
  90                 STORE_SAMPLES                          \
  91                 " dec %2                        \n\t"  \
  92                 " jne 1b                        \n\t"  \
  93                 "2:                             \n\t"  \
  94                 " mov %4, %2                    \n\t"  \
  95                 " and $"#mask", %2              \n\t"  \
  96                 " je 4f                         \n\t"  \
  97                 "3:                             \n\t"  \
  98                 HANDLE_SINGLE_##s()                    \
  99                 " dec %2                        \n\t"  \
 100                 " jne 3b                        \n\t"  \
 101                 "4:                             \n\t"  \
 102                 " emms                          \n\t"
 103
 104 #if defined (__i386__) || defined (__amd64__)
 105 static void remap_mono_to_stereo_s16ne_mmx(pa_remap_t *m, void *dst, const void *src, unsigned n) {
 106     pa_reg_x86 temp, temp2;
 107
 108     __asm__ __volatile__ (
 109         MONO_TO_STEREO(wd,4,15) /* do words to doubles */
 110         : "+r" (dst), "+r" (src), "=&r" (temp), "=&r" (temp2)
 111         : "r" ((pa_reg_x86)n)
 112         : "cc"
 113     );
 114 }
 115
 116 static void remap_mono_to_stereo_float32ne_mmx(pa_remap_t *m, void *dst, const void *src, unsigned n) {
 117     pa_reg_x86 temp, temp2;
 118
 119     __asm__ __volatile__ (
 120         MONO_TO_STEREO(dq,3,7) /* do doubles to quads */
 121         : "+r" (dst), "+r" (src), "=&r" (temp), "=&r" (temp2)
 122         : "r" ((pa_reg_x86)n)
 123         : "cc"
 124     );
 125 }
 126
 127 /* set the function that will execute the remapping based on the matrices */
 128 static void init_remap_mmx(pa_remap_t *m) {
 129     unsigned n_oc, n_ic;
 130
 131     n_oc = m->o_ss.channels;
 132     n_ic = m->i_ss.channels;
 133
 134     /* find some common channel remappings, fall back to full matrix operation. */
 135     if (n_ic == 1 && n_oc == 2 &&
 136             m->map_table_i[0][0] == 0x10000 && m->map_table_i[1][0] == 0x10000) {
 137
 138         pa_log_info("Using MMX mono to stereo remapping");
 139         pa_set_remap_func(m, remap_mono_to_stereo_s16ne_mmx, remap_mono_to_stereo_float32ne_mmx);
 140     }
 141 }
 142 #endif /* defined (__i386__) || defined (__amd64__) */
 143
 144 void pa_remap_func_init_mmx(pa_cpu_x86_flag_t flags) {
 145 #if defined (__i386__) || defined (__amd64__)
 146
 147     if (flags & PA_CPU_X86_MMX) {
 148         pa_log_info("Initialising MMX optimized remappers.");
 149
 150         pa_set_init_remap_func((pa_init_remap_func_t) init_remap_mmx);
 151     }
 152
 153 #endif /* defined (__i386__) || defined (__amd64__) */
 154 }