generic_mmx.h

DirectFB的源码包:用于图形加速和应用开发的好东东,而且是稳定版本的,已经有应用在嵌入式平台的先例,比如NXP平台.

/*
   (c) Copyright 2001-2007  The DirectFB Organization (directfb.org)
   (c) Copyright 2000-2004  Convergence (integrated media) GmbH

   All rights reserved.

   Written by Denis Oliver Kropp <[e-mail]>,
              Andreas Hundt <[e-mail]>,
              Sven Neumann <[e-mail]>,
              Ville Syrjälä <[e-mail]> and
              Claudio Ciccani <[e-mail]>.

   This library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2 of the License, or (at your option) any later version.

   This library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with this library; if not, write to the
   Free Software Foundation, Inc., 59 Temple Place - Suite 330,
   Boston, MA 02111-1307, USA.
*/


#define __aligned( n )  __attribute__ ((aligned((n))))


static void SCacc_add_to_Dacc_MMX( GenefxState *gfxs )
{
     __asm__ __volatile__ (
               "    movq     %2, %%mm0\n"
               ".align   16\n"
               "1:\n"
               "    movq     (%0), %%mm1\n"
               "    paddw    %%mm0, %%mm1\n"
               "    movq     %%mm1, (%0)\n"
               "    add      $8, %0\n"
               "    dec      %1\n"
               "    jnz      1b\n"
               "    emms"
               : /* no outputs */
               : "D" (gfxs->Dacc), "c" (gfxs->length), "m" (gfxs->SCacc)
               : "%st", "memory");
}

static void Dacc_modulate_argb_MMX( GenefxState *gfxs )
{
     __asm__ __volatile__ (
               "movq     %2, %%mm0\n\t"
               ".align   16\n"
               "1:\n\t"
               "testw    $0xF000, 6(%0)\n\t"
               "jnz      2f\n\t"
               "movq     (%0), %%mm1\n\t"
               "pmullw   %%mm0, %%mm1\n\t"
               "psrlw    $8, %%mm1\n\t"
               "movq     %%mm1, (%0)\n"
               ".align   16\n"
               "2:\n\t"
               "add      $8, %0\n\t"
               "dec      %1\n\t"
               "jnz      1b\n\t"
               "emms"
               : /* no outputs */
               : "D" (gfxs->Dacc), "c" (gfxs->length), "m" (gfxs->Cacc)
               : "%st", "memory");
}

static void Sacc_add_to_Dacc_MMX( GenefxState *gfxs )
{
     __asm__ __volatile__ (
               ".align   16\n"
               "1:\n\t"
               "movq     (%2), %%mm0\n\t"
               "movq     (%0), %%mm1\n\t"
               "paddw    %%mm1, %%mm0\n\t"
               "movq     %%mm0, (%0)\n\t"
               "add      $8, %0\n\t"
               "add      $8, %2\n\t"
               "dec      %1\n\t"
               "jnz      1b\n\t"
               "emms"
               : /* no outputs */
               : "D" (gfxs->Dacc), "c" (gfxs->length), "S" (gfxs->Sacc)
               : "%st", "memory");
}

static void Sacc_to_Aop_rgb16_MMX( GenefxState *gfxs )
{
     static const u32 preload[] = { 0xFF00FF00, 0x0000FF00 };
     static const u32 mask[]    = { 0x00FC00F8, 0x000000F8 };
     static const u32 pm[]      = { 0x01000004, 0x00000004 };

     __asm__ __volatile__ (
	       "movq     %3, %%mm7\n\t"
	       "movq     %4, %%mm5\n\t"
	       "movq     %5, %%mm4\n\t"
               ".align   16\n"
               "1:\n\t"
               "testw    $0xF000, 6(%2)\n\t"
               "jnz      2f\n\t"
               "movq     (%2), %%mm0\n\t"
               "paddusw  %%mm7, %%mm0\n\t"
               "pand     %%mm5, %%mm0\n\t"
               "pmaddwd  %%mm4, %%mm0\n\t"
               "psrlq    $5, %%mm0\n\t"
               "movq     %%mm0, %%mm1\n\t"
               "psrlq    $21, %%mm0\n\t"
               "por      %%mm1, %%mm0\n\t"
               "movd     %%mm0, %%eax\n\t"
               "movw     %%ax, (%0)\n\t"
               ".align 16\n"
               "2:\n\t"
               "add      $8, %2\n\t"
	       "add      $2, %0\n\t"
               "dec      %1\n\t"
               "jnz      1b\n\t"
               "emms"
               : /* no outputs */
               : "D" (gfxs->Aop[0]), "c" (gfxs->length), "S" (gfxs->Sacc),
                 "m" (*preload), "m" (*mask), "m" (*pm)
               : "%eax", "%st", "memory");
}

static void Sacc_to_Aop_rgb32_MMX( GenefxState *gfxs )
{
     static const u32 preload[]  = { 0xFF00FF00, 0x0000FF00 };
     static const u32 postload[] = { 0x00FF00FF, 0x000000FF };
     static const u32 pm[]       = { 0x01000001, 0x00000001 };

     __asm__ __volatile__ (
	       "movq     %3, %%mm1\n\t"
	       "movq     %4, %%mm2\n\t"
	       "movq     %5, %%mm3\n\t"
               ".align   16\n"
               "1:\n\t"
               "testw    $0xF000, 6(%2)\n\t"
               "jnz      2f\n\t"
               "movq     (%2), %%mm0\n\t"
               "paddusw  %%mm1, %%mm0\n\t"
               "pand     %%mm2, %%mm0\n\t"
               "pmaddwd  %%mm3, %%mm0\n\t"
               "movq     %%mm0, %%mm4\n\t"
               "psrlq    $16, %%mm0\n\t"
               "por      %%mm0, %%mm4\n\t"
               "movd     %%mm4, (%0)\n\t"
               ".align 16\n"
               "2:\n\t"
               "add      $8, %2\n\t"
	       "add      $4, %0\n\t"
               "dec      %1\n\t"
               "jnz      1b\n\t"
               "emms"
               : /* no outputs */
               : "D" (gfxs->Aop[0]), "c" (gfxs->length), "S" (gfxs->Sacc),
                 "m" (*preload), "m" (*postload), "m" (*pm)
               : "%st", "memory");
}

__attribute__((no_instrument_function))
static void Sop_argb_Sto_Dacc_MMX( GenefxState *gfxs )
{
     static const u32 zeros[]  = { 0, 0 };
     int i = 0;

     __asm__ __volatile__ (
	       "movq     %5, %%mm0\n\t"
               ".align   16\n"
               "1:\n\t"
               "movd     (%3), %%mm1\n\t"
               "punpcklbw %%mm0, %%mm1\n\t"
               ".align   16\n"
               "2:\n\t"
               "movq     %%mm1, (%1)\n\t"
               "dec      %2\n\t"
               "jz       3f\n\t"
               "add      $8, %1\n\t"
               "add      %4, %0\n\t"
               "testl    $0xFFFF0000, %0\n\t"
               "jz       2b\n\t"
               "movl     %0, %%edx\n\t"
               "andl     $0xFFFF0000, %%edx\n\t"
               "shrl     $14, %%edx\n\t"
#ifdef ARCH_X86_64
               "addq     %%rdx, %3\n\t"
#else
               "addl     %%edx, %3\n\t"
#endif
               "andl     $0xFFFF, %0\n\t"
               "jmp      1b\n"
               "3:\n\t"
               "emms"
               : "=r" (i)
               : "D" (gfxs->Dacc), "c" (gfxs->length), "S" (gfxs->Sop[0]),
                 "a" (gfxs->SperD), "m" (*zeros), "0" (i)
               : "%edx", "%st", "memory");
}

static void Sop_argb_to_Dacc_MMX( GenefxState *gfxs )
{
     static const u32 zeros[]  = { 0, 0 };

     __asm__ __volatile__ (
	       "movq     %3, %%mm0\n\t"
               ".align   16\n"
               "1:\n\t"
               "movd     (%2), %%mm1\n\t"
               "punpcklbw %%mm0, %%mm1\n\t"
               "movq     %%mm1, (%0)\n\t"
               "add      $4, %2\n\t"
               "add      $8, %0\n\t"
               "dec      %1\n\t"
               "jnz      1b\n\t"
               "emms"
               : /* no outputs */
               : "D" (gfxs->Dacc), "c" (gfxs->length),
                 "S" (gfxs->Sop[0]), "m" (*zeros)
               : "%st", "memory");
}

static void Sop_rgb16_to_Dacc_MMX( GenefxState *gfxs )
{
     static const u32 mask[]  = { 0x07E0001F, 0x0000F800 };
     static const u32 smul[]  = { 0x00200800, 0x00000001 };
     static const u32 alpha[] = { 0x00000000, 0x00FF0000 };

     __asm__ __volatile__ (
	       "movq     %3, %%mm4\n\t"
	       "movq     %4, %%mm5\n\t"
	       "movq     %5, %%mm7\n\t"
               ".align   16\n"
               "1:\n\t"
               "movq     (%2), %%mm0\n\t"
               /* 1. Konvertierung nach 24 bit interleaved */
	       "movq     %%mm0, %%mm3\n\t"
               "punpcklwd %%mm3, %%mm3\n\t"
               "punpckldq %%mm3, %%mm3\n\t"
               "pand     %%mm4, %%mm3\n\t"
               "pmullw   %%mm5, %%mm3\n\t"
               "psrlw    $8, %%mm3\n\t"
               /* mm3 enthaelt jetzt: 0000 00rr 00gg 00bb des alten pixels */
               "por      %%mm7, %%mm3\n\t"
               "movq     %%mm3, (%0)\n\t"
               "dec      %1\n\t"
               "jz       2f\n\t"
               "psrlq    $16, %%mm0\n\t"
	       "add      $8, %0\n\t"
               /* 2. Konvertierung nach 24 bit interleaved */
	       "movq     %%mm0, %%mm3\n\t"
               "punpcklwd %%mm3, %%mm3\n\t"
               "punpckldq %%mm3, %%mm3\n\t"
               "pand     %%mm4, %%mm3\n\t"
               "pmullw   %%mm5, %%mm3\n\t"
               "psrlw    $8, %%mm3\n\t"
               /* mm3 enthaelt jetzt: 0000 00rr 00gg 00bb des alten pixels */
               "por      %%mm7, %%mm3\n\t"
               "movq     %%mm3, (%0)\n\t"
               "dec      %1\n\t"
               "jz       2f\n\t"
               "psrlq    $16, %%mm0\n\t"
	       "add      $8, %0\n\t"
               /* 3. Konvertierung nach 24 bit interleaved */
	       "movq     %%mm0, %%mm3\n\t"
               "punpcklwd %%mm3, %%mm3\n\t"
               "punpckldq %%mm3, %%mm3\n\t"
               "pand     %%mm4, %%mm3\n\t"
               "pmullw   %%mm5, %%mm3\n\t"
               "psrlw    $8, %%mm3\n\t"
               /* mm3 enthaelt jetzt: 0000 00rr 00gg 00bb des alten pixels */
               "por      %%mm7, %%mm3\n\t"
               "movq     %%mm3, (%0)\n\t"
               "dec      %1\n\t"
               "jz       2f\n\t"
               "psrlq    $16, %%mm0\n\t"
	       "add      $8, %0\n\t"
               /* 4. Konvertierung nach 24 bit interleaved */
	       "movq     %%mm0, %%mm3\n\t"
               "punpcklwd %%mm3, %%mm3\n\t"
               "punpckldq %%mm3, %%mm3\n\t"
               "pand     %%mm4, %%mm3\n\t"
               "pmullw   %%mm5, %%mm3\n\t"
               "psrlw    $8, %%mm3\n\t"
               /* mm3 enthaelt jetzt: 0000 00rr 00gg 00bb des alten pixels */
               "por      %%mm7, %%mm3\n\t"
               "movq     %%mm3, (%0)\n\t"
               "dec      %1\n\t"
               "jz       2f\n\t"
	       "add      $8, %0\n\t"
	       "add      $8, %2\n\t"
               "jmp      1b\n"
               "2:\n\t"
               "emms"
               : /* no outputs */
               : "D" (gfxs->Dacc), "c" (gfxs->length), "S" (gfxs->Sop[0]),
                 "m" (*mask), "m" (*smul), "m" (*alpha)
               : "%st", "memory");
}

static void Sop_rgb32_to_Dacc_MMX( GenefxState *gfxs )
{
     static const u32 alpha[]  = { 0, 0x00FF0000 };
     static const u32 zeros[]  = { 0, 0 };

     __asm__ __volatile__ (
	       "movq     %3, %%mm7\n\t"
	       "movq     %4, %%mm6\n\t"
               ".align   16\n"
               "1:\n\t"
               "movd     (%2), %%mm0\n\t"
               "punpcklbw %%mm6, %%mm0\n\t"
               "por      %%mm7, %%mm0\n\t"
               "movq     %%mm0, (%0)\n\t"
	       "add      $4, %2\n\t"
	       "add      $8, %0\n\t"
               "dec      %1\n\t"
               "jnz      1b\n\t"
               "emms"
               : /* no outputs */
               : "D" (gfxs->Dacc), "c" (gfxs->length), "S" (gfxs->Sop[0]),
                 "m" (*alpha), "m" (*zeros)
               : "%st", "memory");
}

static void Xacc_blend_invsrcalpha_MMX( GenefxState *gfxs )
{