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 <pulsecore/log.h>
  29 #include <pulsecore/macro.h>
  30
  31 #include "cpu-x86.h"
  32 #include "remap.h"
  33
  34 #define LOAD_SAMPLES                                   \
  35                 " movq (%1), %%mm0              \n\t"  \
  36                 " movq 8(%1), %%mm2             \n\t"  \
  37                 " movq 16(%1), %%mm4            \n\t"  \
  38                 " movq 24(%1), %%mm6            \n\t"  \
  39                 " movq %%mm0, %%mm1             \n\t"  \
  40                 " movq %%mm2, %%mm3             \n\t"  \
  41                 " movq %%mm4, %%mm5             \n\t"  \
  42                 " movq %%mm6, %%mm7             \n\t"
  43
  44 #define UNPACK_SAMPLES(s)                              \
  45                 " punpckl"#s" %%mm0, %%mm0      \n\t"  \
  46                 " punpckh"#s" %%mm1, %%mm1      \n\t"  \
  47                 " punpckl"#s" %%mm2, %%mm2      \n\t"  \
  48                 " punpckh"#s" %%mm3, %%mm3      \n\t"  \
  49                 " punpckl"#s" %%mm4, %%mm4      \n\t"  \
  50                 " punpckh"#s" %%mm5, %%mm5      \n\t"  \
  51                 " punpckl"#s" %%mm6, %%mm6      \n\t"  \
  52                 " punpckh"#s" %%mm7, %%mm7      \n\t"
  53
  54 #define STORE_SAMPLES                                  \
  55                 " movq %%mm0, (%0)              \n\t"  \
  56                 " movq %%mm1, 8(%0)             \n\t"  \
  57                 " movq %%mm2, 16(%0)            \n\t"  \
  58                 " movq %%mm3, 24(%0)            \n\t"  \
  59                 " movq %%mm4, 32(%0)            \n\t"  \
  60                 " movq %%mm5, 40(%0)            \n\t"  \
  61                 " movq %%mm6, 48(%0)            \n\t"  \
  62                 " movq %%mm7, 56(%0)            \n\t"  \
  63                 " add $32, %1                   \n\t"  \
  64                 " add $64, %0                   \n\t"
  65
  66 #define HANDLE_SINGLE_dq()                            \
  67                 " movd (%1), %%mm0              \n\t"  \
  68                 " punpckldq %%mm0, %%mm0        \n\t"  \
  69                 " movq %%mm0, (%0)              \n\t"  \
  70                 " add $4, %1                    \n\t"  \
  71                 " add $8, %0                    \n\t"
  72
  73 #define HANDLE_SINGLE_wd()                             \
  74                 " movw (%1), %w3                \n\t"  \
  75                 " movd %3,  %%mm0               \n\t"  \
  76                 " punpcklwd %%mm0, %%mm0        \n\t"  \
  77                 " movd %%mm0, (%0)              \n\t"  \
  78                 " add $2, %1                    \n\t"  \
  79                 " add $4, %0                    \n\t"
  80
  81 #define MONO_TO_STEREO(s,shift,mask)                   \
  82                 " mov %4, %2                    \n\t"  \
  83                 " sar $"#shift", %2             \n\t"  \
  84                 " cmp $0, %2                    \n\t"  \
  85                 " je 2f                         \n\t"  \
  86                 "1:                             \n\t"  \
  87                 LOAD_SAMPLES                           \
  88                 UNPACK_SAMPLES(s)                      \
  89                 STORE_SAMPLES                          \
  90                 " dec %2                        \n\t"  \
  91                 " jne 1b                        \n\t"  \
  92                 "2:                             \n\t"  \
  93                 " mov %4, %2                    \n\t"  \
  94                 " and $"#mask", %2              \n\t"  \
  95                 " je 4f                         \n\t"  \
  96                 "3:                             \n\t"  \
  97                 HANDLE_SINGLE_##s()                    \
  98                 " dec %2                        \n\t"  \
  99                 " jne 3b                        \n\t"  \
 100                 "4:                             \n\t"  \
 101                 " emms                          \n\t"
 102
 103 #if defined (__i386__) || defined (__amd64__)
 104 static void remap_mono_to_stereo_mmx(pa_remap_t *m, void *dst, const void *src, unsigned n) {
 105     pa_reg_x86 temp, temp2;
 106
 107     switch (*m->format) {
 108         case PA_SAMPLE_FLOAT32NE:
 109         {
 110             __asm__ __volatile__ (
 111                 MONO_TO_STEREO(dq,3,7) /* do doubles to quads */
 112                 : "+r" (dst), "+r" (src), "=&r" (temp), "=&r" (temp2)
 113                 : "r" ((pa_reg_x86)n)
 114                 : "cc"
 115             );
 116             break;
 117         }
 118         case PA_SAMPLE_S16NE:
 119         {
 120             __asm__ __volatile__ (
 121                 MONO_TO_STEREO(wd,4,15) /* do words to doubles */
 122                 : "+r" (dst), "+r" (src), "=&r" (temp), "=&r" (temp2)
 123                 : "r" ((pa_reg_x86)n)
 124                 : "cc"
 125             );
 126             break;
 127         }
 128         default:
 129             pa_assert_not_reached();
 130     }
 131 }
 132
 133 /* set the function that will execute the remapping based on the matrices */
 134 static void init_remap_mmx(pa_remap_t *m) {
 135     unsigned n_oc, n_ic;
 136
 137     n_oc = m->o_ss->channels;
 138     n_ic = m->i_ss->channels;
 139
 140     /* find some common channel remappings, fall back to full matrix operation. */
 141     if (n_ic == 1 && n_oc == 2 &&
 142             m->map_table_f[0][0] >= 1.0 && m->map_table_f[1][0] >= 1.0) {
 143         m->do_remap = (pa_do_remap_func_t) remap_mono_to_stereo_mmx;
 144         pa_log_info("Using MMX mono to stereo remapping");
 145     }
 146 }
 147 #endif /* defined (__i386__) || defined (__amd64__) */
 148
 149 void pa_remap_func_init_mmx(pa_cpu_x86_flag_t flags) {
 150 #if defined (__i386__) || defined (__amd64__)
 151
 152     if (flags & PA_CPU_X86_MMX) {
 153         pa_log_info("Initialising MMX optimized remappers.");
 154
 155         pa_set_init_remap_func((pa_init_remap_func_t) init_remap_mmx);
 156     }
 157
 158 #endif /* defined (__i386__) || defined (__amd64__) */
 159 }