Mercurial > sdl-ios-xcode
view src/audio/SDL_mixer_MMX.c @ 761:c5b2b6d2d1fe
Date: Wed, 31 Dec 2003 21:55:30 +0100
From: Max Horn
Subject: SDL: video/quartz cleanup
while doing some experimental changes in the quartz code, I was annoyed
by having to recompile that one big .o file over and over again. So I
decided to finally realize one TODO: properly splitting the code over
multiple files :-).
With two exceptions, I didn't make code changes, only rearranged files
and added new headers. Since there are several new files, making a
patch didn't work out so well, so I decided to just send you all the
new & modified files.
The one source change I made is related to showing/hiding the mouse. I
renamed cursor_visible to cursor_should_be_visible and cursor_hidden to
cursor_visible; I think that makes reading the code easier.
Then I added two new functions: QZ_ShowMouse and QZ_HideMouse. They
help manage cursor_visible (the former 'cursor_hidden'). Finally I
replaced the Carbon ShowCursor/HiderCuror calls by [NSCursor hide] and
[NSCursor unhide]. The API docs are not conclusive, but it might be
that with those the "cursor_visible" (former 'cursor_hidden') hack may
not be necessary anymore; however so far I didn't test this hypothesis,
so I left that in.
The other change was to remove in_foreground and use [NSApp isActive]
instead: Manually keeping track of whether we are in the foreground is
error prone. This should work better in some corner cases.
| author | Sam Lantinga <slouken@libsdl.org> |
|---|---|
| date | Sun, 04 Jan 2004 14:55:35 +0000 |
| parents | a9e38f3b8e4d |
| children | b4b64bb88f2f |
line wrap: on
line source
// MMX assembler version of SDL_MixAudio for signed little endian 16 bit samples and signed 8 bit samples // Copyright 2002 Stephane Marchesin (stephane.marchesin@wanadoo.fr) // This code is licensed under the LGPL (see COPYING for details) // // Assumes buffer size in bytes is a multiple of 16 // Assumes SDL_MIX_MAXVOLUME = 128 //////////////////////////////////////////////// // Mixing for 16 bit signed buffers //////////////////////////////////////////////// #if defined(i386) && defined(__GNUC__) && defined(USE_ASMBLIT) void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume) { __asm__ __volatile__ ( " movl %0,%%edi\n" // edi = dst " movl %1,%%esi\n" // esi = src " movl %3,%%eax\n" // eax = volume " movl %2,%%ebx\n" // ebx = size " shrl $4,%%ebx\n" // process 16 bytes per iteration = 8 samples " 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 ".align 16\n" " .mixloopS16:\n" " movq (%%esi),%%mm1\n" // mm1 = a|b|c|d " movq %%mm1,%%mm2\n" // mm2 = a|b|c|d " movq 8(%%esi),%%mm4\n" // mm4 = e|f|g|h // pré charger le buffer dst dans mm7 " movq (%%edi),%%mm7\n" // mm7 = dst[0]" // multiplier par le volume " pmullw %%mm0,%%mm1\n" // mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) " pmulhw %%mm0,%%mm2\n" // mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) " movq %%mm4,%%mm5\n" // mm5 = e|f|g|h " pmullw %%mm0,%%mm4\n" // mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) " pmulhw %%mm0,%%mm5\n" // mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v) " movq %%mm1,%%mm3\n" // mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v) " punpckhwd %%mm2,%%mm1\n" // mm1 = a*v|b*v " movq %%mm4,%%mm6\n" // mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v) " punpcklwd %%mm2,%%mm3\n" // mm3 = c*v|d*v " punpckhwd %%mm5,%%mm4\n" // mm4 = e*f|f*v " punpcklwd %%mm5,%%mm6\n" // mm6 = g*v|h*v // pré charger le buffer dst dans mm5 " movq 8(%%edi),%%mm5\n" // mm5 = dst[1] // diviser par 128 " psrad $7,%%mm1\n" // mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME " addl $16,%%esi\n" " psrad $7,%%mm3\n" // mm3 = c*v/128|d*v/128 " psrad $7,%%mm4\n" // mm4 = e*v/128|f*v/128 // mm1 = le sample avec le volume modifié " packssdw %%mm1,%%mm3\n" // mm3 = s(a*v|b*v|c*v|d*v) " psrad $7,%%mm6\n" // mm6= g*v/128|h*v/128 " paddsw %%mm7,%%mm3\n" // mm3 = adjust_volume(src)+dst // mm4 = le sample avec le volume modifié " packssdw %%mm4,%%mm6\n" // mm6 = s(e*v|f*v|g*v|h*v) " movq %%mm3,(%%edi)\n" " paddsw %%mm5,%%mm6\n" // mm6 = adjust_volume(src)+dst " movq %%mm6,8(%%edi)\n" " addl $16,%%edi\n" " dec %%ebx\n" " jnz .mixloopS16\n" " emms\n" ".endS16:\n" : : "m" (dst), "m"(src),"m"(size), "m"(volume) : "eax","ebx", "esi", "edi","memory" ); } //////////////////////////////////////////////// // Mixing for 8 bit signed buffers //////////////////////////////////////////////// void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume) { __asm__ __volatile__ ( " movl %0,%%edi\n" // edi = dst " movl %1,%%esi\n" // esi = src " movl %3,%%eax\n" // eax = volume " movd %%ebx,%%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,%%ebx\n" // ebx = size " shr $3,%%ebx\n" // process 8 bytes per iteration = 8 samples " cmp $0,%%ebx\n" " je .endS8\n" ".align 16\n" " .mixloopS8:\n" " pxor %%mm2,%%mm2\n" // mm2 = 0 " movq (%%esi),%%mm1\n" // mm1 = a|b|c|d|e|f|g|h " movq %%mm1,%%mm3\n" // mm3 = a|b|c|d|e|f|g|h // on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 " pcmpgtb %%mm1,%%mm2\n" // mm2 = 11111111|00000000|00000000.... " punpckhbw %%mm2,%%mm1\n" // mm1 = 0|a|0|b|0|c|0|d " punpcklbw %%mm2,%%mm3\n" // mm3 = 0|e|0|f|0|g|0|h " movq (%%edi),%%mm2\n" // mm2 = destination " pmullw %%mm0,%%mm1\n" // mm1 = v*a|v*b|v*c|v*d " addl $8,%%esi\n" " pmullw %%mm0,%%mm3\n" // mm3 = v*e|v*f|v*g|v*h " psraw $7,%%mm1\n" // mm1 = v*a/128|v*b/128|v*c/128|v*d/128 " psraw $7,%%mm3\n" // mm3 = v*e/128|v*f/128|v*g/128|v*h/128 " 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 " paddsb %%mm2,%%mm3\n" // add to destination buffer " movq %%mm3,(%%edi)\n" // store back to ram " addl $8,%%edi\n" " dec %%ebx\n" " jnz .mixloopS8\n" ".endS8:\n" " emms\n" : : "m" (dst), "m"(src),"m"(size), "m"(volume) : "eax","ebx", "esi", "edi","memory" ); } #endif