src/audio/SDL_mixer_MMX.c
author Sam Lantinga <slouken@libsdl.org>
Mon, 29 May 2006 04:04:35 +0000
branchSDL-1.3
changeset 1668 4da1ee79c9af
parent 1662 782fd950bd46
permissions -rw-r--r--
more tweaking indent options
slouken@1402
     1
/*
slouken@1402
     2
    SDL - Simple DirectMedia Layer
slouken@1402
     3
    Copyright (C) 1997-2006 Sam Lantinga
slouken@1402
     4
slouken@1402
     5
    This library is free software; you can redistribute it and/or
slouken@1402
     6
    modify it under the terms of the GNU Lesser General Public
slouken@1402
     7
    License as published by the Free Software Foundation; either
slouken@1402
     8
    version 2.1 of the License, or (at your option) any later version.
slouken@1402
     9
slouken@1402
    10
    This library is distributed in the hope that it will be useful,
slouken@1402
    11
    but WITHOUT ANY WARRANTY; without even the implied warranty of
slouken@1402
    12
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
slouken@1402
    13
    Lesser General Public License for more details.
slouken@1402
    14
slouken@1402
    15
    You should have received a copy of the GNU Lesser General Public
slouken@1402
    16
    License along with this library; if not, write to the Free Software
slouken@1402
    17
    Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
slouken@1402
    18
slouken@1402
    19
    Sam Lantinga
slouken@1402
    20
    slouken@libsdl.org
slouken@1402
    21
*/
slouken@1402
    22
#include "SDL_config.h"
slouken@1402
    23
slouken@1019
    24
/*
slouken@1019
    25
    MMX assembler version of SDL_MixAudio for signed little endian 16 bit samples and signed 8 bit samples
slouken@1019
    26
    Copyright 2002 Stephane Marchesin (stephane.marchesin@wanadoo.fr)
slouken@1019
    27
    This code is licensed under the LGPL (see COPYING for details)
slouken@1019
    28
 
slouken@1019
    29
    Assumes buffer size in bytes is a multiple of 16
slouken@1019
    30
    Assumes SDL_MIX_MAXVOLUME = 128
slouken@1019
    31
*/
slouken@539
    32
slouken@539
    33
slouken@1019
    34
/***********************************************
slouken@1019
    35
*   Mixing for 16 bit signed buffers
slouken@1019
    36
***********************************************/
slouken@539
    37
slouken@1402
    38
#if defined(__GNUC__) && defined(__i386__) && defined(SDL_ASSEMBLY_ROUTINES)
slouken@1662
    39
void
slouken@1668
    40
SDL_MixAudio_MMX_S16(char *dst, char *src, unsigned int size, int volume)
slouken@539
    41
{
slouken@1668
    42
    __asm__ __volatile__("	movl %3,%%eax\n"        /* eax = volume */
slouken@1668
    43
                         "	movl %2,%%edx\n"        /* edx = size */
slouken@1668
    44
                         "	shrl $4,%%edx\n"        /* process 16 bytes per iteration = 8 samples */
slouken@1668
    45
                         "	jz .endS16\n" "	pxor %%mm0,%%mm0\n" "	movd %%eax,%%mm0\n" "	movq %%mm0,%%mm1\n" "	psllq $16,%%mm0\n" "	por %%mm1,%%mm0\n" "	psllq $16,%%mm0\n" "	por %%mm1,%%mm0\n" "	psllq $16,%%mm0\n" "	por %%mm1,%%mm0\n"      /* mm0 = vol|vol|vol|vol */
slouken@1668
    46
                         ".align 8\n" "	.mixloopS16:\n" "	movq (%1),%%mm1\n"      /* mm1 = a|b|c|d */
slouken@1668
    47
                         "	movq %%mm1,%%mm2\n"     /* mm2 = a|b|c|d */
slouken@1668
    48
                         "	movq 8(%1),%%mm4\n"     /* mm4 = e|f|g|h */
slouken@1668
    49
                         /* pré charger le buffer dst dans mm7 */
slouken@1668
    50
                         "	movq (%0),%%mm7\n"      /* mm7 = dst[0] */
slouken@1668
    51
                         /* multiplier par le volume */
slouken@1668
    52
                         "	pmullw %%mm0,%%mm1\n"   /* mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */
slouken@1668
    53
                         "	pmulhw %%mm0,%%mm2\n"   /* mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) */
slouken@1668
    54
                         "	movq %%mm4,%%mm5\n"     /* mm5 = e|f|g|h */
slouken@1668
    55
                         "	pmullw %%mm0,%%mm4\n"   /* mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */
slouken@1668
    56
                         "	pmulhw %%mm0,%%mm5\n"   /* mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v) */
slouken@1668
    57
                         "	movq %%mm1,%%mm3\n"     /* mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */
slouken@1668
    58
                         "	punpckhwd %%mm2,%%mm1\n"        /* mm1 = a*v|b*v */
slouken@1668
    59
                         "	movq %%mm4,%%mm6\n"     /* mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */
slouken@1668
    60
                         "	punpcklwd %%mm2,%%mm3\n"        /* mm3 = c*v|d*v */
slouken@1668
    61
                         "	punpckhwd %%mm5,%%mm4\n"        /* mm4 = e*f|f*v */
slouken@1668
    62
                         "	punpcklwd %%mm5,%%mm6\n"        /* mm6 = g*v|h*v */
slouken@1668
    63
                         /* pré charger le buffer dst dans mm5 */
slouken@1668
    64
                         "	movq 8(%0),%%mm5\n"     /* mm5 = dst[1] */
slouken@1668
    65
                         /* diviser par 128 */
slouken@1668
    66
                         "	psrad $7,%%mm1\n"       /* mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME */
slouken@1668
    67
                         "	add $16,%1\n" "	psrad $7,%%mm3\n"       /* mm3 = c*v/128|d*v/128 */
slouken@1668
    68
                         "	psrad $7,%%mm4\n"       /* mm4 = e*v/128|f*v/128 */
slouken@1668
    69
                         /* mm1 = le sample avec le volume modifié */
slouken@1668
    70
                         "	packssdw %%mm1,%%mm3\n" /* mm3 = s(a*v|b*v|c*v|d*v) */
slouken@1668
    71
                         "	psrad $7,%%mm6\n"       /* mm6= g*v/128|h*v/128 */
slouken@1668
    72
                         "	paddsw %%mm7,%%mm3\n"   /* mm3 = adjust_volume(src)+dst */
slouken@1668
    73
                         /* mm4 = le sample avec le volume modifié */
slouken@1668
    74
                         "	packssdw %%mm4,%%mm6\n" /* mm6 = s(e*v|f*v|g*v|h*v) */
slouken@1668
    75
                         "	movq %%mm3,(%0)\n" "	paddsw %%mm5,%%mm6\n"   /* mm6 = adjust_volume(src)+dst */
slouken@1668
    76
                         "	movq %%mm6,8(%0)\n"
slouken@1668
    77
                         "	add $16,%0\n"
slouken@1668
    78
                         "	dec %%edx\n"
slouken@1668
    79
                         "	jnz .mixloopS16\n"
slouken@1668
    80
                         "	emms\n"
slouken@1668
    81
                         ".endS16:\n"::"r"(dst), "r"(src),
slouken@1668
    82
                         "m"(size), "m"(volume):"eax", "edx", "memory");
slouken@539
    83
}
slouken@539
    84
slouken@539
    85
slouken@539
    86
slouken@1612
    87
/*////////////////////////////////////////////// */
slouken@1612
    88
/* Mixing for 8 bit signed buffers */
slouken@1612
    89
/*////////////////////////////////////////////// */
slouken@539
    90
slouken@1662
    91
void
slouken@1668
    92
SDL_MixAudio_MMX_S8(char *dst, char *src, unsigned int size, int volume)
slouken@539
    93
{
slouken@1668
    94
    __asm__ __volatile__("	movl %3,%%eax\n"        /* eax = volume */
slouken@1668
    95
                         "	movd %%eax,%%mm0\n" "	movq %%mm0,%%mm1\n" "	psllq $16,%%mm0\n" "	por %%mm1,%%mm0\n" "	psllq $16,%%mm0\n" "	por %%mm1,%%mm0\n" "	psllq $16,%%mm0\n" "	por %%mm1,%%mm0\n" "	movl %2,%%edx\n"        /* edx = size */
slouken@1668
    96
                         "	shr $3,%%edx\n" /* process 8 bytes per iteration = 8 samples */
slouken@1668
    97
                         "	cmp $0,%%edx\n" "	je .endS8\n" ".align 8\n" "	.mixloopS8:\n" "	pxor %%mm2,%%mm2\n"     /* mm2 = 0 */
slouken@1668
    98
                         "	movq (%1),%%mm1\n"      /* mm1 = a|b|c|d|e|f|g|h */
slouken@1668
    99
                         "	movq %%mm1,%%mm3\n"     /* mm3 = a|b|c|d|e|f|g|h */
slouken@1668
   100
                         /* on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 */
slouken@1668
   101
                         "	pcmpgtb %%mm1,%%mm2\n"  /* mm2 = 11111111|00000000|00000000.... */
slouken@1668
   102
                         "	punpckhbw %%mm2,%%mm1\n"        /* mm1 = 0|a|0|b|0|c|0|d */
slouken@1668
   103
                         "	punpcklbw %%mm2,%%mm3\n"        /* mm3 = 0|e|0|f|0|g|0|h */
slouken@1668
   104
                         "	movq (%0),%%mm2\n"      /* mm2 = destination */
slouken@1668
   105
                         "	pmullw %%mm0,%%mm1\n"   /* mm1 = v*a|v*b|v*c|v*d */
slouken@1668
   106
                         "	add $8,%1\n" "	pmullw %%mm0,%%mm3\n"   /* mm3 = v*e|v*f|v*g|v*h */
slouken@1668
   107
                         "	psraw $7,%%mm1\n"       /* mm1 = v*a/128|v*b/128|v*c/128|v*d/128  */
slouken@1668
   108
                         "	psraw $7,%%mm3\n"       /* mm3 = v*e/128|v*f/128|v*g/128|v*h/128 */
slouken@1668
   109
                         "	packsswb %%mm1,%%mm3\n" /* mm1 = v*a/128|v*b/128|v*c/128|v*d/128|v*e/128|v*f/128|v*g/128|v*h/128 */
slouken@1668
   110
                         "	paddsb %%mm2,%%mm3\n"   /* add to destination buffer */
slouken@1668
   111
                         "	movq %%mm3,(%0)\n"      /* store back to ram */
slouken@1668
   112
                         "	add $8,%0\n"
slouken@1668
   113
                         "	dec %%edx\n"
slouken@1668
   114
                         "	jnz .mixloopS8\n"
slouken@1668
   115
                         ".endS8:\n"
slouken@1668
   116
                         "	emms\n"::"r"(dst), "r"(src), "m"(size),
slouken@1668
   117
                         "m"(volume):"eax", "edx", "memory");
slouken@539
   118
}
slouken@539
   119
#endif
slouken@1662
   120
/* vi: set ts=4 sw=4 expandtab: */