src/audio/SDL_mixer_MMX.c
author Sam Lantinga <slouken@libsdl.org>
Sun, 02 Aug 2009 18:39:57 +0000
changeset 3227 458e53d8662c
parent 2859 99210400e8b9
child 3628 4d46850be3f6
permissions -rw-r--r--
Clarified API documentation
     1 /*
     2     SDL - Simple DirectMedia Layer
     3     Copyright (C) 1997-2009 Sam Lantinga
     4 
     5     This library is free software; you can redistribute it and/or
     6     modify it under the terms of the GNU Lesser General Public
     7     License as published by the Free Software Foundation; either
     8     version 2.1 of the License, or (at your option) any later version.
     9 
    10     This library is distributed in the hope that it will be useful,
    11     but WITHOUT ANY WARRANTY; without even the implied warranty of
    12     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
    13     Lesser General Public License for more details.
    14 
    15     You should have received a copy of the GNU Lesser General Public
    16     License along with this library; if not, write to the Free Software
    17     Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
    18 
    19     Sam Lantinga
    20     slouken@libsdl.org
    21 */
    22 #include "SDL_config.h"
    23 
    24 /*
    25     MMX assembler version of SDL_MixAudio for signed little endian 16 bit samples and signed 8 bit samples
    26     Copyright 2002 Stephane Marchesin (stephane.marchesin@wanadoo.fr)
    27     This code is licensed under the LGPL (see COPYING for details)
    28  
    29     Assumes buffer size in bytes is a multiple of 16
    30     Assumes SDL_MIX_MAXVOLUME = 128
    31 */
    32 
    33 
    34 /***********************************************
    35 *   Mixing for 16 bit signed buffers
    36 ***********************************************/
    37 
    38 #if defined(__GNUC__) && defined(__i386__) && defined(SDL_ASSEMBLY_ROUTINES)
    39 void
    40 SDL_MixAudio_MMX_S16(char *dst, char *src, unsigned int size, int volume)
    41 {
    42     __asm__ __volatile__("	movl %3,%%eax\n"        /* eax = volume */
    43                          "	movl %2,%%edx\n"        /* edx = size */
    44                          "	shrl $4,%%edx\n"        /* process 16 bytes per iteration = 8 samples */
    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 */
    46                          ".align 8\n" "	.mixloopS16:\n" "	movq (%1),%%mm1\n"      /* mm1 = a|b|c|d */
    47                          "	movq %%mm1,%%mm2\n"     /* mm2 = a|b|c|d */
    48                          "	movq 8(%1),%%mm4\n"     /* mm4 = e|f|g|h */
    49                          /* pré charger le buffer dst dans mm7 */
    50                          "	movq (%0),%%mm7\n"      /* mm7 = dst[0] */
    51                          /* multiplier par le volume */
    52                          "	pmullw %%mm0,%%mm1\n"   /* mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */
    53                          "	pmulhw %%mm0,%%mm2\n"   /* mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) */
    54                          "	movq %%mm4,%%mm5\n"     /* mm5 = e|f|g|h */
    55                          "	pmullw %%mm0,%%mm4\n"   /* mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */
    56                          "	pmulhw %%mm0,%%mm5\n"   /* mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v) */
    57                          "	movq %%mm1,%%mm3\n"     /* mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */
    58                          "	punpckhwd %%mm2,%%mm1\n"        /* mm1 = a*v|b*v */
    59                          "	movq %%mm4,%%mm6\n"     /* mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */
    60                          "	punpcklwd %%mm2,%%mm3\n"        /* mm3 = c*v|d*v */
    61                          "	punpckhwd %%mm5,%%mm4\n"        /* mm4 = e*f|f*v */
    62                          "	punpcklwd %%mm5,%%mm6\n"        /* mm6 = g*v|h*v */
    63                          /* pré charger le buffer dst dans mm5 */
    64                          "	movq 8(%0),%%mm5\n"     /* mm5 = dst[1] */
    65                          /* diviser par 128 */
    66                          "	psrad $7,%%mm1\n"       /* mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME */
    67                          "	add $16,%1\n" "	psrad $7,%%mm3\n"       /* mm3 = c*v/128|d*v/128 */
    68                          "	psrad $7,%%mm4\n"       /* mm4 = e*v/128|f*v/128 */
    69                          /* mm1 = le sample avec le volume modifié */
    70                          "	packssdw %%mm1,%%mm3\n" /* mm3 = s(a*v|b*v|c*v|d*v) */
    71                          "	psrad $7,%%mm6\n"       /* mm6= g*v/128|h*v/128 */
    72                          "	paddsw %%mm7,%%mm3\n"   /* mm3 = adjust_volume(src)+dst */
    73                          /* mm4 = le sample avec le volume modifié */
    74                          "	packssdw %%mm4,%%mm6\n" /* mm6 = s(e*v|f*v|g*v|h*v) */
    75                          "	movq %%mm3,(%0)\n" "	paddsw %%mm5,%%mm6\n"   /* mm6 = adjust_volume(src)+dst */
    76                          "	movq %%mm6,8(%0)\n"
    77                          "	add $16,%0\n"
    78                          "	dec %%edx\n"
    79                          "	jnz .mixloopS16\n"
    80                          "	emms\n"
    81                          ".endS16:\n"::"r"(dst), "r"(src),
    82                          "m"(size), "m"(volume):"eax", "edx", "memory");
    83 }
    84 
    85 
    86 
    87 /*////////////////////////////////////////////// */
    88 /* Mixing for 8 bit signed buffers */
    89 /*////////////////////////////////////////////// */
    90 
    91 void
    92 SDL_MixAudio_MMX_S8(char *dst, char *src, unsigned int size, int volume)
    93 {
    94     __asm__ __volatile__("	movl %3,%%eax\n"        /* eax = volume */
    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 */
    96                          "	shr $3,%%edx\n" /* process 8 bytes per iteration = 8 samples */
    97                          "	cmp $0,%%edx\n" "	je .endS8\n" ".align 8\n" "	.mixloopS8:\n" "	pxor %%mm2,%%mm2\n"     /* mm2 = 0 */
    98                          "	movq (%1),%%mm1\n"      /* mm1 = a|b|c|d|e|f|g|h */
    99                          "	movq %%mm1,%%mm3\n"     /* mm3 = a|b|c|d|e|f|g|h */
   100                          /* on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 */
   101                          "	pcmpgtb %%mm1,%%mm2\n"  /* mm2 = 11111111|00000000|00000000.... */
   102                          "	punpckhbw %%mm2,%%mm1\n"        /* mm1 = 0|a|0|b|0|c|0|d */
   103                          "	punpcklbw %%mm2,%%mm3\n"        /* mm3 = 0|e|0|f|0|g|0|h */
   104                          "	movq (%0),%%mm2\n"      /* mm2 = destination */
   105                          "	pmullw %%mm0,%%mm1\n"   /* mm1 = v*a|v*b|v*c|v*d */
   106                          "	add $8,%1\n" "	pmullw %%mm0,%%mm3\n"   /* mm3 = v*e|v*f|v*g|v*h */
   107                          "	psraw $7,%%mm1\n"       /* mm1 = v*a/128|v*b/128|v*c/128|v*d/128  */
   108                          "	psraw $7,%%mm3\n"       /* mm3 = v*e/128|v*f/128|v*g/128|v*h/128 */
   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 */
   110                          "	paddsb %%mm2,%%mm3\n"   /* add to destination buffer */
   111                          "	movq %%mm3,(%0)\n"      /* store back to ram */
   112                          "	add $8,%0\n"
   113                          "	dec %%edx\n"
   114                          "	jnz .mixloopS8\n"
   115                          ".endS8:\n"
   116                          "	emms\n"::"r"(dst), "r"(src), "m"(size),
   117                          "m"(volume):"eax", "edx", "memory");
   118 }
   119 #endif
   120 /* vi: set ts=4 sw=4 expandtab: */