src/video/SDL_yuv_mmx.c
author Sam Lantinga <slouken@libsdl.org>
Sun, 13 Feb 2005 07:10:02 +0000
changeset 1038 29d7db09776e
parent 949 e0d96eb0af19
child 1148 63fb2da89a4b
permissions -rw-r--r--
Ugly hack to make this work with gcc 2.x and 3.x Thanks to Stephane Marchesin for the dirty dirty work.

/*
    SDL - Simple DirectMedia Layer
    Copyright (C) 1997-2004 Sam Lantinga

    This library is free software; you can redistribute it and/or
    modify it under the terms of the GNU Library 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
    Library General Public License for more details.

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

    Sam Lantinga
    slouken@libsdl.org
*/

#ifdef SAVE_RCSID
static char rcsid =
 "@(#) $Id$";
#endif


#if defined(i386) && defined(__GNUC__) && defined(USE_ASMBLIT)

#include "SDL_types.h"

#if __GNUC__ > 2
#    undef GCC2_HACK
#else
#    define GCC2_HACK
#endif

 
#if defined(GCC2_HACK) && defined (__ELF__)
#define ASM_VAR(X) _##X
#else
#define ASM_VAR(X) X
#endif
 
static volatile unsigned int  ASM_VAR(MMX_0080w)[]    = {0x00800080, 0x00800080};
static volatile unsigned int  ASM_VAR(MMX_00FFw)[]    = {0x00ff00ff, 0x00ff00ff}; 
static volatile unsigned int  ASM_VAR(MMX_FF00w)[]    = {0xff00ff00, 0xff00ff00}; 

static volatile unsigned short ASM_VAR(MMX_Ycoeff)[]  = {0x004a, 0x004a, 0x004a, 0x004a}; 

static volatile unsigned short ASM_VAR(MMX_UbluRGB)[] = {0x0072, 0x0072, 0x0072, 0x0072};    
static volatile unsigned short ASM_VAR(MMX_VredRGB)[] = {0x0059, 0x0059, 0x0059, 0x0059};  
static volatile unsigned short ASM_VAR(MMX_UgrnRGB)[] = {0xffea, 0xffea, 0xffea, 0xffea}; 
static volatile unsigned short ASM_VAR(MMX_VgrnRGB)[] = {0xffd2, 0xffd2, 0xffd2, 0xffd2};  

static volatile unsigned short ASM_VAR(MMX_Ublu5x5)[] = {0x0081, 0x0081, 0x0081, 0x0081};
static volatile unsigned short ASM_VAR(MMX_Vred5x5)[] = {0x0066, 0x0066, 0x0066, 0x0066};
static volatile unsigned short ASM_VAR(MMX_Ugrn555)[] = {0xffe7, 0xffe7, 0xffe7, 0xffe7};
static volatile unsigned short ASM_VAR(MMX_Vgrn555)[] = {0xffcc, 0xffcc, 0xffcc, 0xffcc};
static volatile unsigned short ASM_VAR(MMX_Ugrn565)[] = {0xffe8, 0xffe8, 0xffe8, 0xffe8};
static volatile unsigned short ASM_VAR(MMX_Vgrn565)[] = {0xffcd, 0xffcd, 0xffcd, 0xffcd};

static volatile unsigned short ASM_VAR(MMX_red555)[]  = {0x7c00, 0x7c00, 0x7c00, 0x7c00};
static volatile unsigned short ASM_VAR(MMX_red565)[]  = {0xf800, 0xf800, 0xf800, 0xf800};
static volatile unsigned short ASM_VAR(MMX_grn555)[]  = {0x03e0, 0x03e0, 0x03e0, 0x03e0};
static volatile unsigned short ASM_VAR(MMX_grn565)[]  = {0x07e0, 0x07e0, 0x07e0, 0x07e0};
static volatile unsigned short ASM_VAR(MMX_blu5x5)[]  = {0x001f, 0x001f, 0x001f, 0x001f};

/**
   This MMX assembler is my first assembler/MMX program ever.
   Thus it maybe buggy.
   Send patches to:
   mvogt@rhrk.uni-kl.de

   After it worked fine I have "obfuscated" the code a bit to have
   more parallism in the MMX units. This means I moved
   initilisation around and delayed other instruction.
   Performance measurement did not show that this brought any advantage
   but in theory it _should_ be faster this way.

   The overall performanve gain to the C based dither was 30%-40%.
   The MMX routine calculates 256bit=8RGB values in each cycle
   (4 for row1 & 4 for row2)

   The red/green/blue.. coefficents are taken from the mpeg_play 
   player. They look nice, but I dont know if you can have
   better values, to avoid integer rounding errors.
   

   IMPORTANT:
   ==========

   It is a requirement that the cr/cb/lum are 8 byte aligned and
   the out are 16byte aligned or you will/may get segfaults

*/

void ColorRGBDitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
                              unsigned char *lum, unsigned char *cr,
                              unsigned char *cb, unsigned char *out,
                              int rows, int cols, int mod )
{
    Uint32 *row1;
    Uint32 *row2;

    unsigned char* y = lum +cols*rows;    // Pointer to the end
    int x=0;
    row1 = (Uint32 *)out;                 // 32 bit target
    row2 = (Uint32 *)out+cols+mod;        // start of second row 
    mod = (mod+cols+mod)*4;               // increment for row1 in byte

    __asm__ __volatile__ (
/* We don't really care about PIC - the code should be rewritten to use
   relative addressing for the static tables, so right now we take the
   COW hit on the pages this code resides. Big deal.
   This spill is just to reduce register pressure in the PIC case. */
		 "pushl %%ebx\n"
		 "movl %0, %%ebx\n"

	         ".align 8\n"
		 "1:\n"
		
		 // create Cr (result in mm1)
		 "movd (%%ebx), %%mm1\n"   //         0  0  0  0  v3 v2 v1 v0
		 "pxor %%mm7,%%mm7\n"      //         00 00 00 00 00 00 00 00
		 "movd (%2), %%mm2\n"           //    0  0  0  0 l3 l2 l1 l0
		 "punpcklbw %%mm7,%%mm1\n" //         0  v3 0  v2 00 v1 00 v0
		 "punpckldq %%mm1,%%mm1\n" //         00 v1 00 v0 00 v1 00 v0
#ifdef GCC2_HACK
		 "psubw _MMX_0080w,%%mm1\n"  // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 
#else
		 "psubw %[_MMX_0080w],%%mm1\n"  // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 
#endif

		 // create Cr_g (result in mm0)
		 "movq %%mm1,%%mm0\n"           // r1 r1 r0 r0 r1 r1 r0 r0
#ifdef GCC2_HACK
		 "pmullw _MMX_VgrnRGB,%%mm0\n"// red*-46dec=0.7136*64
		 "pmullw _MMX_VredRGB,%%mm1\n"// red*89dec=1.4013*64
#else
		 "pmullw %[_MMX_VgrnRGB],%%mm0\n"// red*-46dec=0.7136*64
		 "pmullw %[_MMX_VredRGB],%%mm1\n"// red*89dec=1.4013*64
#endif
		 "psraw  $6, %%mm0\n"           // red=red/64
		 "psraw  $6, %%mm1\n"           // red=red/64
		 
		 // create L1 L2 (result in mm2,mm4)
		 // L2=lum+cols
		 "movq (%2,%4),%%mm3\n"         //    0  0  0  0 L3 L2 L1 L0
		 "punpckldq %%mm3,%%mm2\n"      //   L3 L2 L1 L0 l3 l2 l1 l0
		 "movq %%mm2,%%mm4\n"           //   L3 L2 L1 L0 l3 l2 l1 l0
#ifdef GCC2_HACK
		 "pand _MMX_FF00w,%%mm2\n"      //   L3 0  L1  0 l3  0 l1  0
		 "pand _MMX_00FFw,%%mm4\n"      //   0  L2  0 L0  0 l2  0 l0
#else
		 "pand %[_MMX_FF00w],%%mm2\n"      //   L3 0  L1  0 l3  0 l1  0
		 "pand %[_MMX_00FFw],%%mm4\n"      //   0  L2  0 L0  0 l2  0 l0
#endif
		 "psrlw $8,%%mm2\n"             //   0  L3  0 L1  0 l3  0 l1

		 // create R (result in mm6)
		 "movq %%mm2,%%mm5\n"           //   0 L3  0 L1  0 l3  0 l1
		 "movq %%mm4,%%mm6\n"           //   0 L2  0 L0  0 l2  0 l0
		 "paddsw  %%mm1, %%mm5\n"       // lum1+red:x R3 x R1 x r3 x r1
		 "paddsw  %%mm1, %%mm6\n"       // lum1+red:x R2 x R0 x r2 x r0
		 "packuswb %%mm5,%%mm5\n"       //  R3 R1 r3 r1 R3 R1 r3 r1
		 "packuswb %%mm6,%%mm6\n"       //  R2 R0 r2 r0 R2 R0 r2 r0
		 "pxor %%mm7,%%mm7\n"      //         00 00 00 00 00 00 00 00
		 "punpcklbw %%mm5,%%mm6\n"      //  R3 R2 R1 R0 r3 r2 r1 r0

		 // create Cb (result in mm1)
		 "movd (%1), %%mm1\n"      //         0  0  0  0  u3 u2 u1 u0
		 "punpcklbw %%mm7,%%mm1\n" //         0  u3 0  u2 00 u1 00 u0
		 "punpckldq %%mm1,%%mm1\n" //         00 u1 00 u0 00 u1 00 u0
#ifdef GCC2_HACK
		 "psubw _MMX_0080w,%%mm1\n"  // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 
#else
		 "psubw %[_MMX_0080w],%%mm1\n"  // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 
#endif
		 // create Cb_g (result in mm5)
		 "movq %%mm1,%%mm5\n"            // u1 u1 u0 u0 u1 u1 u0 u0
#ifdef GCC2_HACK
		 "pmullw _MMX_UgrnRGB,%%mm5\n"    // blue*-109dec=1.7129*64
		 "pmullw _MMX_UbluRGB,%%mm1\n"    // blue*114dec=1.78125*64
#else
		 "pmullw %[_MMX_UgrnRGB],%%mm5\n"    // blue*-109dec=1.7129*64
		 "pmullw %[_MMX_UbluRGB],%%mm1\n"    // blue*114dec=1.78125*64
#endif
		 "psraw  $6, %%mm5\n"            // blue=red/64
		 "psraw  $6, %%mm1\n"            // blue=blue/64

		 // create G (result in mm7)
		 "movq %%mm2,%%mm3\n"      //   0  L3  0 L1  0 l3  0 l1
		 "movq %%mm4,%%mm7\n"      //   0  L2  0 L0  0 l2  0 l1
		 "paddsw  %%mm5, %%mm3\n"  // lum1+Cb_g:x G3t x G1t x g3t x g1t
		 "paddsw  %%mm5, %%mm7\n"  // lum1+Cb_g:x G2t x G0t x g2t x g0t
		 "paddsw  %%mm0, %%mm3\n"  // lum1+Cr_g:x G3  x G1  x g3  x g1
		 "paddsw  %%mm0, %%mm7\n"  // lum1+blue:x G2  x G0  x g2  x g0
		 "packuswb %%mm3,%%mm3\n"  // G3 G1 g3 g1 G3 G1 g3 g1
		 "packuswb %%mm7,%%mm7\n"  // G2 G0 g2 g0 G2 G0 g2 g0
		 "punpcklbw %%mm3,%%mm7\n" // G3 G2 G1 G0 g3 g2 g1 g0
		 
		 // create B (result in mm5)
		 "movq %%mm2,%%mm3\n"         //   0  L3  0 L1  0 l3  0 l1
		 "movq %%mm4,%%mm5\n"         //   0  L2  0 L0  0 l2  0 l1
		 "paddsw  %%mm1, %%mm3\n"     // lum1+blue:x B3 x B1 x b3 x b1
		 "paddsw  %%mm1, %%mm5\n"     // lum1+blue:x B2 x B0 x b2 x b0
		 "packuswb %%mm3,%%mm3\n"     // B3 B1 b3 b1 B3 B1 b3 b1
		 "packuswb %%mm5,%%mm5\n"     // B2 B0 b2 b0 B2 B0 b2 b0
		 "punpcklbw %%mm3,%%mm5\n"    // B3 B2 B1 B0 b3 b2 b1 b0

		 // fill destination row1 (needed are mm6=Rr,mm7=Gg,mm5=Bb)

		 "pxor %%mm2,%%mm2\n"           //  0  0  0  0  0  0  0  0
		 "pxor %%mm4,%%mm4\n"           //  0  0  0  0  0  0  0  0
		 "movq %%mm6,%%mm1\n"           // R3 R2 R1 R0 r3 r2 r1 r0
		 "movq %%mm5,%%mm3\n"           // B3 B2 B1 B0 b3 b2 b1 b0
		 // process lower lum
		 "punpcklbw %%mm4,%%mm1\n"      //  0 r3  0 r2  0 r1  0 r0
		 "punpcklbw %%mm4,%%mm3\n"      //  0 b3  0 b2  0 b1  0 b0
		 "movq %%mm1,%%mm2\n"           //  0 r3  0 r2  0 r1  0 r0
		 "movq %%mm3,%%mm0\n"           //  0 b3  0 b2  0 b1  0 b0
		 "punpcklwd %%mm1,%%mm3\n"      //  0 r1  0 b1  0 r0  0 b0
		 "punpckhwd %%mm2,%%mm0\n"      //  0 r3  0 b3  0 r2  0 b2

		 "pxor %%mm2,%%mm2\n"           //  0  0  0  0  0  0  0  0
		 "movq %%mm7,%%mm1\n"           // G3 G2 G1 G0 g3 g2 g1 g0
		 "punpcklbw %%mm1,%%mm2\n"      // g3  0 g2  0 g1  0 g0  0
		 "punpcklwd %%mm4,%%mm2\n"      //  0  0 g1  0  0  0 g0  0 
		 "por  %%mm3, %%mm2\n"          //  0 r1 g1 b1  0 r0 g0 b0
		 "movq   %%mm2,(%3)\n"          // wrote out ! row1

		 "pxor %%mm2,%%mm2\n"           //  0  0  0  0  0  0  0  0
		 "punpcklbw %%mm1,%%mm4\n"      // g3  0 g2  0 g1  0 g0  0
		 "punpckhwd %%mm2,%%mm4\n"      //  0  0 g3  0  0  0 g2  0 
		 "por  %%mm0, %%mm4\n"          //  0 r3 g3 b3  0 r2 g2 b2
		 "movq   %%mm4,8(%3)\n"         // wrote out ! row1
		 
		 // fill destination row2 (needed are mm6=Rr,mm7=Gg,mm5=Bb)
		 // this can be done "destructive"
		 "pxor %%mm2,%%mm2\n"           //  0  0  0  0  0  0  0  0
		 "punpckhbw %%mm2,%%mm6\n"      //  0 R3  0 R2  0 R1  0 R0
		 "punpckhbw %%mm1,%%mm5\n"      // G3 B3 G2 B2 G1 B1 G0 B0
		 "movq %%mm5,%%mm1\n"           // G3 B3 G2 B2 G1 B1 G0 B0
		 "punpcklwd %%mm6,%%mm1\n"      //  0 R1 G1 B1  0 R0 G0 B0
		 "movq   %%mm1,(%5)\n"          // wrote out ! row2
		 "punpckhwd %%mm6,%%mm5\n"      //  0 R3 G3 B3  0 R2 G2 B2
		 "movq   %%mm5,8(%5)\n"         // wrote out ! row2
		 
		 "addl  $4,%2\n"            // lum+4
		 "leal  16(%3),%3\n"        // row1+16
		 "leal  16(%5),%5\n"        // row2+16
		 "addl  $2, %%ebx\n"        // cr+2
		 "addl  $2, %1\n"           // cb+2

		 "addl  $4,%6\n"            // x+4
		 "cmpl  %4,%6\n"

		 "jl    1b\n"
		 "addl           %4,     %2\n" // lum += cols 
		 "addl           %8,     %3\n" // row1+= mod
		 "addl           %8,     %5\n" // row2+= mod
		 "movl           $0,     %6\n" // x=0
		 "cmpl           %7,     %2\n"
		 "jl             1b\n"
		 "emms\n"
		 "popl %%ebx\n"
		 :
		 : "m" (cr), "r"(cb),"r"(lum),
		 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)
#ifndef GCC2_HACK
		 ,[_MMX_0080w] "m" (*MMX_0080w),
		 [_MMX_00FFw] "m" (*MMX_00FFw),
		 [_MMX_FF00w] "m" (*MMX_FF00w),
		 [_MMX_VgrnRGB] "m" (*MMX_VgrnRGB),
		 [_MMX_VredRGB] "m" (*MMX_VredRGB),
		 [_MMX_UgrnRGB] "m" (*MMX_UgrnRGB),
		 [_MMX_UbluRGB] "m" (*MMX_UbluRGB)
#endif
			 );
}

void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
                             unsigned char *lum, unsigned char *cr,
                             unsigned char *cb, unsigned char *out,
                             int rows, int cols, int mod )
{
    Uint16 *row1;
    Uint16 *row2;

    unsigned char* y = lum +cols*rows;    /* Pointer to the end */
    int x=0;
    row1 = (Uint16 *)out;                 /* 16 bit target */
    row2 = (Uint16 *)out+cols+mod;        /* start of second row  */
    mod = (mod+cols+mod)*2;               /* increment for row1 in byte */


      __asm__ __volatile__(
         "pushl %%ebx\n"
	 "movl %0, %%ebx\n"

         ".align 8\n"
         "1:\n"
         "movd           (%1),                   %%mm0\n" // 4 Cb         0  0  0  0 u3 u2 u1 u0
         "pxor           %%mm7,                  %%mm7\n"
         "movd           (%%ebx),                %%mm1\n" // 4 Cr                0  0  0  0 v3 v2 v1 v0
         "punpcklbw      %%mm7,                  %%mm0\n" // 4 W cb   0 u3  0 u2  0 u1  0 u0
         "punpcklbw      %%mm7,                  %%mm1\n" // 4 W cr   0 v3  0 v2  0 v1  0 v0
#ifdef GCC2_HACK
         "psubw          _MMX_0080w,             %%mm0\n"
         "psubw          _MMX_0080w,             %%mm1\n"
#else
         "psubw          %[_MMX_0080w],             %%mm0\n"
         "psubw          %[_MMX_0080w],             %%mm1\n"
#endif
         "movq           %%mm0,                  %%mm2\n" // Cb                   0 u3  0 u2  0 u1  0 u0
         "movq           %%mm1,                  %%mm3\n" // Cr
#ifdef GCC2_HACK
         "pmullw         _MMX_Ugrn565,           %%mm2\n" // Cb2green 0 R3  0 R2  0 R1  0 R0
#else
         "pmullw         %[_MMX_Ugrn565],           %%mm2\n" // Cb2green 0 R3  0 R2  0 R1  0 R0
#endif
         "movq           (%2),                   %%mm6\n" // L1      l7 L6 L5 L4 L3 L2 L1 L0
#ifdef GCC2_HACK
         "pmullw         _MMX_Ublu5x5,           %%mm0\n" // Cb2blue
         "pand           _MMX_00FFw,             %%mm6\n" // L1      00 L6 00 L4 00 L2 00 L0
         "pmullw         _MMX_Vgrn565,           %%mm3\n" // Cr2green
#else
         "pmullw         %[_MMX_Ublu5x5],           %%mm0\n" // Cb2blue
         "pand           %[_MMX_00FFw],             %%mm6\n" // L1      00 L6 00 L4 00 L2 00 L0
         "pmullw         %[_MMX_Vgrn565],           %%mm3\n" // Cr2green
#endif
         "movq           (%2),                   %%mm7\n" // L2
#ifdef GCC2_HACK
         "pmullw         _MMX_Vred5x5,           %%mm1\n" // Cr2red
#else
         "pmullw         %[_MMX_Vred5x5],           %%mm1\n" // Cr2red
#endif
         "psrlw          $8,                     %%mm7\n"        // L2           00 L7 00 L5 00 L3 00 L1
#ifdef GCC2_HACK
         "pmullw         _MMX_Ycoeff,            %%mm6\n" // lum1
#else
         "pmullw         %[_MMX_Ycoeff],            %%mm6\n" // lum1
#endif
         "paddw          %%mm3,                  %%mm2\n" // Cb2green + Cr2green == green
#ifdef GCC2_HACK
         "pmullw         _MMX_Ycoeff,            %%mm7\n" // lum2
#else
         "pmullw         %[_MMX_Ycoeff],            %%mm7\n" // lum2
#endif

         "movq           %%mm6,                  %%mm4\n" // lum1
         "paddw          %%mm0,                  %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0
         "movq           %%mm4,                  %%mm5\n" // lum1
         "paddw          %%mm1,                  %%mm4\n" // lum1 +red  00 R6 00 R4 00 R2 00 R0
         "paddw          %%mm2,                  %%mm5\n" // lum1 +green 00 G6 00 G4 00 G2 00 G0
         "psraw          $6,                     %%mm4\n" // R1 0 .. 64
         "movq           %%mm7,                  %%mm3\n" // lum2                       00 L7 00 L5 00 L3 00 L1
         "psraw          $6,                     %%mm5\n" // G1  - .. +
         "paddw          %%mm0,                  %%mm7\n" // Lum2 +blue 00 B7 00 B5 00 B3 00 B1
         "psraw          $6,                     %%mm6\n" // B1         0 .. 64
         "packuswb       %%mm4,                  %%mm4\n" // R1 R1
         "packuswb       %%mm5,                  %%mm5\n" // G1 G1
         "packuswb       %%mm6,                  %%mm6\n" // B1 B1
         "punpcklbw      %%mm4,                  %%mm4\n"
         "punpcklbw      %%mm5,                  %%mm5\n"

#ifdef GCC2_HACK
         "pand           _MMX_red565,            %%mm4\n"
#else
         "pand           %[_MMX_red565],            %%mm4\n"
#endif
         "psllw          $3,                     %%mm5\n" // GREEN       1
         "punpcklbw      %%mm6,                  %%mm6\n"
#ifdef GCC2_HACK
         "pand           _MMX_grn565,            %%mm5\n"
         "pand           _MMX_red565,            %%mm6\n"
#else
         "pand           %[_MMX_grn565],            %%mm5\n"
         "pand           %[_MMX_red565],            %%mm6\n"
#endif
         "por            %%mm5,                  %%mm4\n" //
         "psrlw          $11,                    %%mm6\n" // BLUE        1
         "movq           %%mm3,                  %%mm5\n" // lum2
         "paddw          %%mm1,                  %%mm3\n" // lum2 +red      00 R7 00 R5 00 R3 00 R1
         "paddw          %%mm2,                  %%mm5\n" // lum2 +green 00 G7 00 G5 00 G3 00 G1
         "psraw          $6,                     %%mm3\n" // R2
         "por            %%mm6,                  %%mm4\n" // MM4
         "psraw          $6,                     %%mm5\n" // G2
         "movq           (%2, %4),               %%mm6\n" // L3 load lum2
         "psraw          $6,                     %%mm7\n"
         "packuswb       %%mm3,                  %%mm3\n"
         "packuswb       %%mm5,                  %%mm5\n"
         "packuswb       %%mm7,                  %%mm7\n"
#ifdef GCC2_HACK
         "pand           _MMX_00FFw,             %%mm6\n" // L3
#else
         "pand           %[_MMX_00FFw],             %%mm6\n" // L3
#endif
         "punpcklbw      %%mm3,                  %%mm3\n"
         "punpcklbw      %%mm5,                  %%mm5\n"
#ifdef GCC2_HACK
         "pmullw         _MMX_Ycoeff,            %%mm6\n" // lum3
#else
         "pmullw         %[_MMX_Ycoeff],            %%mm6\n" // lum3
#endif
         "punpcklbw      %%mm7,                  %%mm7\n"
         "psllw          $3,                     %%mm5\n" // GREEN 2
#ifdef GCC2_HACK
         "pand           _MMX_red565,            %%mm7\n"
         "pand           _MMX_red565,            %%mm3\n"
#else
         "pand           %[_MMX_red565],            %%mm7\n"
         "pand           %[_MMX_red565],            %%mm3\n"
#endif
         "psrlw          $11,                    %%mm7\n" // BLUE  2
#ifdef GCC2_HACK
         "pand           _MMX_grn565,            %%mm5\n"
#else
         "pand           %[_MMX_grn565],            %%mm5\n"
#endif
         "por            %%mm7,                  %%mm3\n"
         "movq           (%2,%4),                %%mm7\n" // L4 load lum2
         "por            %%mm5,                  %%mm3\n" //
         "psrlw          $8,                     %%mm7\n" // L4
         "movq           %%mm4,                  %%mm5\n"
         "punpcklwd      %%mm3,                  %%mm4\n"
#ifdef GCC2_HACK
         "pmullw         _MMX_Ycoeff,            %%mm7\n" // lum4
#else
         "pmullw         %[_MMX_Ycoeff],            %%mm7\n" // lum4
#endif
         "punpckhwd      %%mm3,                  %%mm5\n"

         "movq           %%mm4,                  (%3)\n"  // write row1
         "movq           %%mm5,                  8(%3)\n" // write row1

         "movq           %%mm6,                  %%mm4\n" // Lum3
         "paddw          %%mm0,                  %%mm6\n" // Lum3 +blue

         "movq           %%mm4,                  %%mm5\n" // Lum3
         "paddw          %%mm1,                  %%mm4\n" // Lum3 +red
         "paddw          %%mm2,                  %%mm5\n" // Lum3 +green
         "psraw          $6,                     %%mm4\n"
         "movq           %%mm7,                  %%mm3\n" // Lum4
         "psraw          $6,                     %%mm5\n"
         "paddw          %%mm0,                  %%mm7\n" // Lum4 +blue
         "psraw          $6,                     %%mm6\n" // Lum3 +blue
         "movq           %%mm3,                  %%mm0\n" // Lum4
         "packuswb       %%mm4,                  %%mm4\n"
         "paddw          %%mm1,                  %%mm3\n" // Lum4 +red
         "packuswb       %%mm5,                  %%mm5\n"
         "paddw          %%mm2,                  %%mm0\n" // Lum4 +green
         "packuswb       %%mm6,                  %%mm6\n"
         "punpcklbw      %%mm4,                  %%mm4\n"
         "punpcklbw      %%mm5,                  %%mm5\n"
         "punpcklbw      %%mm6,                  %%mm6\n"
         "psllw          $3,                     %%mm5\n" // GREEN 3
#ifdef GCC2_HACK
         "pand           _MMX_red565,            %%mm4\n"
#else
         "pand           %[_MMX_red565],            %%mm4\n"
#endif
         "psraw          $6,                     %%mm3\n" // psr 6
         "psraw          $6,                     %%mm0\n"
#ifdef GCC2_HACK
         "pand           _MMX_red565,            %%mm6\n" // BLUE
         "pand           _MMX_grn565,            %%mm5\n"
#else
         "pand           %[_MMX_red565],            %%mm6\n" // BLUE
         "pand           %[_MMX_grn565],            %%mm5\n"
#endif
         "psrlw          $11,                    %%mm6\n" // BLUE  3
         "por            %%mm5,                  %%mm4\n"
         "psraw          $6,                     %%mm7\n"
         "por            %%mm6,                  %%mm4\n"
         "packuswb       %%mm3,                  %%mm3\n"
         "packuswb       %%mm0,                  %%mm0\n"
         "packuswb       %%mm7,                  %%mm7\n"
         "punpcklbw      %%mm3,                  %%mm3\n"
         "punpcklbw      %%mm0,                  %%mm0\n"
         "punpcklbw      %%mm7,                  %%mm7\n"
#ifdef GCC2_HACK
         "pand           _MMX_red565,            %%mm3\n"
         "pand           _MMX_red565,            %%mm7\n" // BLUE
#else
         "pand           %[_MMX_red565],            %%mm3\n"
         "pand           %[_MMX_red565],            %%mm7\n" // BLUE
#endif
         "psllw          $3,                     %%mm0\n" // GREEN 4
         "psrlw          $11,                    %%mm7\n"
#ifdef GCC2_HACK
         "pand           _MMX_grn565,            %%mm0\n"
#else
         "pand           %[_MMX_grn565],            %%mm0\n"
#endif
         "por            %%mm7,                  %%mm3\n"
         "por            %%mm0,                  %%mm3\n"

         "movq           %%mm4,                  %%mm5\n"

         "punpcklwd      %%mm3,                  %%mm4\n"
         "punpckhwd      %%mm3,                  %%mm5\n"

         "movq           %%mm4,                  (%5)\n"
	 "movq           %%mm5,                  8(%5)\n"

         "addl           $8,                     %6\n"
         "addl           $8,                     %2\n"
         "addl           $4,                     %%ebx\n"
         "addl           $4,                     %1\n"
         "cmpl           %4,                     %6\n"
         "leal           16(%3),                 %3\n"
	 "leal           16(%5),%5\n" // row2+16


         "jl             1b\n"
	 "addl           %4,     %2\n" // lum += cols 
	 "addl           %8,     %3\n" // row1+= mod
	 "addl           %8,     %5\n" // row2+= mod
	 "movl           $0,     %6\n" // x=0
	 "cmpl           %7,     %2\n"
	 "jl             1b\n"
         "emms\n"
	 "popl %%ebx\n"
         :
         :"m" (cr), "r"(cb),"r"(lum),
	 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)
#ifndef GCC2_HACK
	 ,[_MMX_0080w] "m" (*MMX_0080w),
	 [_MMX_Ugrn565] "m" (*MMX_Ugrn565),
	 [_MMX_Ublu5x5] "m" (*MMX_Ublu5x5),
	 [_MMX_00FFw] "m" (*MMX_00FFw),
	 [_MMX_Vgrn565] "m" (*MMX_Vgrn565),
	 [_MMX_Vred5x5] "m" (*MMX_Vred5x5),
	 [_MMX_Ycoeff] "m" (*MMX_Ycoeff),
	 [_MMX_red565] "m" (*MMX_red565),
	 [_MMX_grn565] "m" (*MMX_grn565)
#endif
		 );
}

#undef GCC2_HACK

#endif /* GCC i386 inline assembly */