sl@0: /*
sl@0:  * LIBOIL - Library of Optimized Inner Loops
sl@0:  * Copyright (c) 2005 David A. Schleef <ds@schleef.org>
sl@0:  * All rights reserved.
sl@0:  *
sl@0:  * Redistribution and use in source and binary forms, with or without
sl@0:  * modification, are permitted provided that the following conditions
sl@0:  * are met:
sl@0:  * 1. Redistributions of source code must retain the above copyright
sl@0:  *    notice, this list of conditions and the following disclaimer.
sl@0:  * 2. Redistributions in binary form must reproduce the above copyright
sl@0:  *    notice, this list of conditions and the following disclaimer in the
sl@0:  *    documentation and/or other materials provided with the distribution.
sl@0:  * 
sl@0:  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
sl@0:  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
sl@0:  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
sl@0:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
sl@0:  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
sl@0:  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
sl@0:  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
sl@0:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
sl@0:  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
sl@0:  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
sl@0:  * POSSIBILITY OF SUCH DAMAGE.
sl@0:  */
sl@0: //Portions Copyright (c)  2008-2009 Nokia Corporation and/or its subsidiary(-ies). All rights reserved. 
sl@0: 
sl@0: #ifdef HAVE_CONFIG_H
sl@0: #include "config.h"
sl@0: #endif
sl@0: 
sl@0: #include <liboil/liboil.h>
sl@0: #include <liboil/liboilfunction.h>
sl@0: 
sl@0: OIL_DECLARE_CLASS (composite_in_argb);
sl@0: OIL_DECLARE_CLASS (composite_in_argb_const_src);
sl@0: OIL_DECLARE_CLASS (composite_in_argb_const_mask);
sl@0: OIL_DECLARE_CLASS (composite_over_argb);
sl@0: OIL_DECLARE_CLASS (composite_over_argb_const_src);
sl@0: OIL_DECLARE_CLASS (composite_add_argb);
sl@0: OIL_DECLARE_CLASS (composite_add_argb_const_src);
sl@0: OIL_DECLARE_CLASS (composite_in_over_argb);
sl@0: OIL_DECLARE_CLASS (composite_in_over_argb_const_src);
sl@0: OIL_DECLARE_CLASS (composite_in_over_argb_const_mask);
sl@0: 
sl@0: #if 0
sl@0: static void
sl@0: composite_in_argb_mmx (uint32_t *dest, uint32_t *src, uint8_t *mask, int n)
sl@0: {
sl@0:   int i;
sl@0: 
sl@0:   for(i=0;i<n;i++){
sl@0:     dest[i] = ARGB(
sl@0:         COMPOSITE_IN(ARGB_A(src[i]), mask[i]),
sl@0:         COMPOSITE_IN(ARGB_R(src[i]), mask[i]),
sl@0:         COMPOSITE_IN(ARGB_G(src[i]), mask[i]),
sl@0:         COMPOSITE_IN(ARGB_B(src[i]), mask[i]));
sl@0:   }
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_in_argb_mmx, composite_in_argb);
sl@0: #endif
sl@0: 
sl@0: /*
sl@0:  * This macro loads the constants:
sl@0:  * mm7 = { 0, 0, 0, 0 }
sl@0:  * mm6 = { 128, 128, 128, 128 }
sl@0:  * mm5 = { 255, 255, 255, 255 }
sl@0:  */
sl@0: #define MMX_LOAD_CONSTANTS \
sl@0:       "  pxor %%mm7, %%mm7\n" \
sl@0:       "  movl $0x80808080, %%eax\n" \
sl@0:       "  movd %%eax, %%mm6\n" \
sl@0:       "  punpcklbw %%mm7, %%mm6\n" \
sl@0:       "  movl $0xffffffff, %%eax\n" \
sl@0:       "  movd %%eax, %%mm5\n" \
sl@0:       "  punpcklbw %%mm7, %%mm5\n"
sl@0: 
sl@0: /*
sl@0:  * a = muldiv255(a, b)
sl@0:  *   a, b are unpacked
sl@0:  *   destroys both registers
sl@0:  *   requires mm6 set up as above
sl@0:  */
sl@0: #define MMX_MULDIV255(a,b) \
sl@0:       "  pmullw %%" #b ", %%" #a "\n" \
sl@0:       "  paddw %%mm6, %%" #a "\n" \
sl@0:       "  movq %%" #a ", %%" #b "\n" \
sl@0:       "  psrlw $8, %%" #b "\n" \
sl@0:       "  paddw %%" #b ", %%" #a "\n" \
sl@0:       "  psrlw $8, %%" #a "\n"
sl@0: 
sl@0: static void
sl@0: composite_in_argb_mmx (uint32_t *dest, uint32_t *src, const uint8_t *mask, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "1:\n"
sl@0:       "  movd (%2), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0x00, %%mm0, %%mm1\n"
sl@0: 
sl@0:       "  movd (%1), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  addl $1, %2\n"
sl@0:       "  decl %3\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (mask), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_in_argb_mmx, composite_in_argb, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: /* 
sl@0:  * This is a different style than the others.  Should be moved elsewhere.
sl@0:  */
sl@0: static void
sl@0: composite_in_argb_mmx2 (uint32_t *dest, uint32_t *src, const uint8_t *mask, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "1:\n"
sl@0:       "  movl (%2), %%eax\n"
sl@0:       /* if alpha == 0, write a 0 */
sl@0:       "  testl $0x000000ff, %%eax\n"
sl@0:       "  je 2f\n"
sl@0:       /* if alpha == 0xff, write src value */
sl@0:       "  cmp $0xff, %%al\n"
sl@0:       "  je 3f\n"
sl@0: 
sl@0:       "  movd %%eax, %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0x00, %%mm0, %%mm1\n"
sl@0: 
sl@0:       "  movd (%1), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  jmp 4f\n"
sl@0:       "2:\n"
sl@0:       "  movl $0, (%0)\n"
sl@0:       "  jmp 4f\n"
sl@0:       "3:\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  movl %%eax, (%0)\n"
sl@0:       "4:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  addl $1, %2\n"
sl@0:       "  decl %3\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (mask), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_in_argb_mmx2, composite_in_argb, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: static void
sl@0: composite_in_argb_const_src_mmx (uint32_t *dest, uint32_t *src, const uint8_t *mask, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "  movd (%1), %%mm3\n"
sl@0:       "  punpcklbw %%mm7, %%mm3\n"
sl@0:       "1:\n"
sl@0:       "  movd (%2), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0x00, %%mm0, %%mm1\n"
sl@0: 
sl@0:       "  movq %%mm3, %%mm2\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $1, %2\n"
sl@0:       "  decl %3\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (mask), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_in_argb_const_src_mmx, composite_in_argb_const_src, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: static void
sl@0: composite_in_argb_const_mask_mmx (uint32_t *dest, uint32_t *src, const uint8_t *mask, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "  movd (%2), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0x00, %%mm0, %%mm3\n"
sl@0:       "1:\n"
sl@0:       "  movq %%mm3, %%mm1\n"
sl@0:       "  movd (%1), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  decl %3\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (mask), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_in_argb_const_mask_mmx, composite_in_argb_const_mask, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: static void
sl@0: composite_over_argb_mmx (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "1:\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 2f\n"
sl@0: 
sl@0:       "  movd %%eax, %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0xff, %%mm0, %%mm1\n"
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0: 
sl@0:       "  movd (%0), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "2:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  decl %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_mmx, composite_over_argb, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: /* unroll 2 */
sl@0: static void
sl@0: composite_over_argb_mmx_2 (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0: 
sl@0:       "  testl $0x1, %2\n"
sl@0:       "  jz 2f\n"
sl@0: 
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 1f\n"
sl@0: 
sl@0:       "  movd %%eax, %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0xff, %%mm0, %%mm1\n"
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0: 
sl@0:       "  movd (%0), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0:       "  pmullw %%mm1, %%mm2\n"
sl@0:       "  paddw %%mm6, %%mm2\n"
sl@0:       "  movq %%mm2, %%mm1\n"
sl@0:       "  psrlw $8, %%mm1\n"
sl@0:       "  paddw %%mm1, %%mm2\n"
sl@0:       "  psrlw $8, %%mm2\n"
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0: 
sl@0:       "1:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0: 
sl@0:       "2:\n"
sl@0:       "  shr $1, %2\n"
sl@0:       "  jz 5f\n"
sl@0:       "3:\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  orl 4(%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 4f\n"
sl@0: 
sl@0:       "  movd (%1), %%mm0\n"
sl@0:       "  movd (%0), %%mm2\n"
sl@0: 
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "   movd 4(%1), %%mm3\n"
sl@0: 
sl@0:       "  pshufw $0xff, %%mm0, %%mm1\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0: 
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0:       "   movd 4(%0), %%mm4\n"
sl@0: 
sl@0:       "  pmullw %%mm1, %%mm2\n"
sl@0:       "   punpcklbw %%mm7, %%mm3\n"
sl@0: 
sl@0:       "  paddw %%mm6, %%mm2\n"
sl@0:       "   punpcklbw %%mm7, %%mm4\n"
sl@0: 
sl@0:       "  movq %%mm2, %%mm1\n"
sl@0:       "   pshufw $0xff, %%mm3, %%mm7\n"
sl@0: 
sl@0:       "  psrlw $8, %%mm1\n"
sl@0:       "   pxor %%mm5, %%mm7\n"
sl@0: 
sl@0:       "  paddw %%mm1, %%mm2\n"
sl@0:       "   pmullw %%mm7, %%mm4\n"
sl@0: 
sl@0:       "  psrlw $8, %%mm2\n"
sl@0:       "   paddw %%mm6, %%mm4\n"
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "   movq %%mm4, %%mm7\n"
sl@0: 
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0:       "   psrlw $8, %%mm7\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "   paddw %%mm7, %%mm4\n"
sl@0: 
sl@0:       "   psrlw $8, %%mm4\n"
sl@0:       "   paddw %%mm3, %%mm4\n"
sl@0:       "   packuswb %%mm4, %%mm4\n"
sl@0:       "   movd %%mm4, 4(%0)\n"
sl@0: 
sl@0:       "  pxor %%mm7, %%mm7\n"
sl@0:       "4:\n"
sl@0:       "  addl $8, %0\n"
sl@0:       "  addl $8, %1\n"
sl@0:       "  decl %2\n"
sl@0:       "  jnz 3b\n"
sl@0:       "5:\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_mmx_2, composite_over_argb, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: /* replace pshufw with punpck */
sl@0: static void
sl@0: composite_over_argb_mmx_3 (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "1:\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 2f\n"
sl@0: 
sl@0:       "  movd %%eax, %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  movq %%mm0, %%mm1\n"
sl@0:       "  punpckhwd %%mm1, %%mm1\n"
sl@0:       "  punpckhdq %%mm1, %%mm1\n"
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0: 
sl@0:       "  movd (%0), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0:       "  pmullw %%mm1, %%mm2\n"
sl@0:       "  paddw %%mm6, %%mm2\n"
sl@0:       "  movq %%mm2, %%mm1\n"
sl@0:       "  psrlw $8, %%mm1\n"
sl@0:       "  paddw %%mm1, %%mm2\n"
sl@0:       "  psrlw $8, %%mm2\n"
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0:       "  movd %%mm2, (%0)\n"
sl@0: 
sl@0:       "2:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  decl %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_mmx_3, composite_over_argb, OIL_IMPL_FLAG_MMX);
sl@0: 
sl@0: /* written for gromit */
sl@0: static void
sl@0: composite_over_argb_mmx_4 (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ ("  pxor %%mm7, %%mm7\n"   // mm7 = { 0, 0, 0, 0 }
sl@0:       "  movl $0x80808080, %%eax\n"
sl@0:       "  movd %%eax, %%mm6\n"  // mm6 = { 128, 128, 128, 128 }
sl@0:       "  punpcklbw %%mm7, %%mm6\n"
sl@0:       "  movl $0xffffffff, %%eax\n" // mm5 = { 255, 255, 255, 255 }
sl@0:       "  movd %%eax, %%mm5\n"
sl@0:       "  punpcklbw %%mm7, %%mm5\n"
sl@0:       "  movl $0x02020202, %%eax\n"
sl@0:       "  movd %%eax, %%mm4\n"
sl@0:       "  punpcklbw %%mm7, %%mm4\n"
sl@0:       "  paddw %%mm5, %%mm4\n" // mm5 = { 257, 257, 257, 257 }
sl@0:       "1:\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 2f\n"
sl@0: 
sl@0:       "  movd %%eax, %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0xff, %%mm0, %%mm1\n"
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0: 
sl@0:       "  movd (%0), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0:       "  pmullw %%mm1, %%mm2\n"
sl@0:       "  paddw %%mm6, %%mm2\n"
sl@0:       "  pmulhuw %%mm4, %%mm2\n"
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "2:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  subl $1, %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_mmx_4, composite_over_argb, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: static void
sl@0: composite_over_argb_mmx_5 (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ ("  pxor %%mm7, %%mm7\n"   // mm7 = { 0, 0, 0, 0 }
sl@0:       "  movl $0x80808080, %%eax\n"
sl@0:       "  movd %%eax, %%mm6\n"  // mm6 = { 128, 128, 128, 128 }
sl@0:       "  punpcklbw %%mm7, %%mm6\n"
sl@0: #if 0
sl@0:       "  movl $0xffffffff, %%eax\n" // mm5 = { 255, 255, 255, 255 }
sl@0:       "  movd %%eax, %%mm5\n"
sl@0:       "  punpcklbw %%mm7, %%mm5\n"
sl@0: #else
sl@0:       "  pcmpeqw %%mm5, %%mm5\n"
sl@0:       "  psrlw $8, %%mm5\n" // mm5 = { 255, 255, 255, 255 }
sl@0: #endif
sl@0:       "  movl $0x02020202, %%eax\n"
sl@0:       "  movd %%eax, %%mm4\n"
sl@0:       "  punpcklbw %%mm7, %%mm4\n"
sl@0:       "  paddw %%mm5, %%mm4\n" // mm5 = { 257, 257, 257, 257 }
sl@0:       "1:\n"
sl@0:       "  movd (%1), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  xor %%eax, %%eax\n"
sl@0:       "  pextrw $3, %%mm0, %%eax\n"
sl@0:       "  test %%eax, %%eax\n"
sl@0:       "  jz 2f\n"
sl@0: 
sl@0:       "  pshufw $0xff, %%mm0, %%mm1\n"
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0: 
sl@0:       "  movd (%0), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0:       "  pmullw %%mm1, %%mm2\n"
sl@0:       "  paddw %%mm6, %%mm2\n"
sl@0:       "  pmulhuw %%mm4, %%mm2\n"
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "2:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  subl $1, %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_mmx_5, composite_over_argb, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: static void
sl@0: composite_over_argb_sse2 (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ ("  pxor %%xmm7, %%xmm7\n"   // mm7 = { 0, 0, 0, 0 }
sl@0:       "  movl $0x80808080, %%eax\n"
sl@0:       "  movd %%eax, %%xmm6\n"  // mm6 = { 128, 128, 128, 128 }
sl@0:       "  punpcklbw %%xmm7, %%xmm6\n"
sl@0:       "  movl $0xffffffff, %%eax\n" // mm5 = { 255, 255, 255, 255 }
sl@0:       "  movd %%eax, %%xmm5\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm5\n"
sl@0:       "  movl $0x02020202, %%eax\n"
sl@0:       "  movd %%eax, %%xmm4\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm4\n"
sl@0:       "  paddw %%xmm5, %%xmm4\n" // mm4 = { 255, 255, 255, 255 }
sl@0:       "1:\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 2f\n"
sl@0: 
sl@0:       "  movd (%1), %%xmm1\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm1\n"
sl@0:       "  pshuflw $0xff, %%xmm1, %%xmm0\n"
sl@0:       "  pxor %%xmm5, %%xmm0\n"
sl@0: 
sl@0:       "  movd (%0), %%xmm3\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm3\n"
sl@0:       "  pmullw %%xmm0, %%xmm3\n"
sl@0:       "  paddw %%xmm6, %%xmm3\n"
sl@0:       "  pmulhuw %%xmm4, %%xmm3\n"
sl@0: 
sl@0:       "  paddw %%xmm1, %%xmm3\n"
sl@0:       "  packuswb %%xmm3, %%xmm3\n"
sl@0:       "  movd %%xmm3, (%0)\n"
sl@0:       "2:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  decl %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_sse2, composite_over_argb, OIL_IMPL_FLAG_SSE2);
sl@0: 
sl@0: /* written for shaun */
sl@0: static void
sl@0: composite_over_argb_sse2_2 (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ ("  pxor %%xmm7, %%xmm7\n"   // mm7 = { 0, 0, 0, 0 }
sl@0:       "  movl $0x80808080, %%eax\n"
sl@0:       "  movd %%eax, %%xmm6\n"  // mm6 = { 128, 128, 128, 128 }
sl@0:       "  punpcklbw %%xmm7, %%xmm6\n"
sl@0:       "  punpcklwd %%xmm6, %%xmm6\n"
sl@0:       "  movl $0xffffffff, %%eax\n" // mm5 = { 255, 255, 255, 255 }
sl@0:       "  movd %%eax, %%xmm5\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm5\n"
sl@0:       "  punpcklwd %%xmm5, %%xmm5\n"
sl@0:       "  movl $0x02020202, %%eax\n"
sl@0:       "  movd %%eax, %%xmm4\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm4\n"
sl@0:       "  paddw %%xmm5, %%xmm4\n" // mm4 = { 257, 257, 257, 257 }
sl@0:       "  punpcklwd %%xmm4, %%xmm4\n"
sl@0:       :
sl@0:       :
sl@0:       :"eax");
sl@0: 
sl@0:   if (n&1) {
sl@0:     __asm__ __volatile__ (
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 1f\n"
sl@0: 
sl@0:       "  movd (%1), %%xmm1\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm1\n"
sl@0:       "  pshuflw $0xff, %%xmm1, %%xmm0\n"
sl@0:       "  pxor %%xmm5, %%xmm0\n"
sl@0: 
sl@0:       "  movd (%0), %%xmm3\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm3\n"
sl@0:       "  pmullw %%xmm0, %%xmm3\n"
sl@0:       "  paddw %%xmm6, %%xmm3\n"
sl@0:       "  pmulhuw %%xmm4, %%xmm3\n"
sl@0: 
sl@0:       "  paddw %%xmm1, %%xmm3\n"
sl@0:       "  packuswb %%xmm3, %%xmm3\n"
sl@0:       "  movd %%xmm3, (%0)\n"
sl@0: 
sl@0:       "1:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       :"+r" (dest), "+r" (src)
sl@0:       :
sl@0:       :"eax");
sl@0:   }
sl@0:   n>>=1;
sl@0: 
sl@0:   if (n>0){
sl@0:     __asm__ __volatile__ ("\n"
sl@0:       "3:\n"
sl@0: #if 0
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  orl 4(%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 4f\n"
sl@0: #endif
sl@0: 
sl@0:       "  movq (%1), %%xmm1\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm1\n"
sl@0:       "  pshuflw $0xff, %%xmm1, %%xmm0\n"
sl@0:       "  pshufhw $0xff, %%xmm0, %%xmm0\n"
sl@0:       "  pxor %%xmm5, %%xmm0\n"
sl@0: 
sl@0:       "  movq (%0), %%xmm3\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm3\n"
sl@0:       "  pmullw %%xmm0, %%xmm3\n"
sl@0:       "  paddw %%xmm6, %%xmm3\n"
sl@0:       "  pmulhuw %%xmm4, %%xmm3\n"
sl@0:       "  paddw %%xmm1, %%xmm3\n"
sl@0:       "  packuswb %%xmm3, %%xmm3\n"
sl@0:       "  movq %%xmm3, (%0)\n"
sl@0:       "4:\n"
sl@0:       "  addl $8, %0\n"
sl@0:       "  addl $8, %1\n"
sl@0:       "  subl $1, %2\n"
sl@0:       "  jnz 3b\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0:   }
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_sse2_2, composite_over_argb, OIL_IMPL_FLAG_SSE2);
sl@0: 
sl@0: /* written for shaun */
sl@0: static void
sl@0: composite_over_argb_sse2_3 (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0:   int begin;
sl@0:   int middle;
sl@0:   int end;
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ ("  pxor %%xmm7, %%xmm7\n"   // mm7 = { 0, 0, 0, 0 }
sl@0:       "  movl $0x80808080, %%eax\n"
sl@0:       "  movd %%eax, %%xmm6\n"  // mm6 = { 128, 128, 128, 128 }
sl@0:       "  punpcklbw %%xmm7, %%xmm6\n"
sl@0:       "  punpcklwd %%xmm6, %%xmm6\n"
sl@0:       "  movl $0xffffffff, %%eax\n" // mm5 = { 255, 255, 255, 255 }
sl@0:       "  movd %%eax, %%xmm5\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm5\n"
sl@0:       "  punpcklwd %%xmm5, %%xmm5\n"
sl@0:       "  movl $0x02020202, %%eax\n"
sl@0:       "  movd %%eax, %%xmm4\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm4\n"
sl@0:       "  paddw %%xmm5, %%xmm4\n" // mm4 = { 257, 257, 257, 257 }
sl@0:       "  punpcklwd %%xmm4, %%xmm4\n"
sl@0:       :
sl@0:       :
sl@0:       :"eax");
sl@0: 
sl@0:   begin = 0x3 & (4 - (((unsigned long)dest & 0xf) >> 2));
sl@0:   if (begin>n) {
sl@0:     begin = n;
sl@0:     middle = 0;
sl@0:     end = 0;
sl@0:   } else {
sl@0:     middle = (n-begin)>>2;
sl@0:     end = n - begin - middle*4;
sl@0:   }
sl@0: 
sl@0:   if (begin>0) {
sl@0:     __asm__ __volatile__ ("\n"
sl@0:       "1:\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 2f\n"
sl@0: 
sl@0:       "  movd (%1), %%xmm1\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm1\n"
sl@0:       "  pshuflw $0xff, %%xmm1, %%xmm0\n"
sl@0:       "  pxor %%xmm5, %%xmm0\n"
sl@0: 
sl@0:       "  movd (%0), %%xmm3\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm3\n"
sl@0:       "  pmullw %%xmm0, %%xmm3\n"
sl@0:       "  paddw %%xmm6, %%xmm3\n"
sl@0:       "  pmulhuw %%xmm4, %%xmm3\n"
sl@0: 
sl@0:       "  paddw %%xmm1, %%xmm3\n"
sl@0:       "  packuswb %%xmm3, %%xmm3\n"
sl@0:       "  movd %%xmm3, (%0)\n"
sl@0: 
sl@0:       "2:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  subl $1, %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (begin)
sl@0:       :
sl@0:       :"eax");
sl@0:   }
sl@0: 
sl@0:   if (middle>0){
sl@0:     __asm__ __volatile__ ("\n"
sl@0:       "1:\n"
sl@0:       "  movq (%1), %%xmm1\n"
sl@0:       "  movq 8(%1), %%xmm0\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  orl 4(%1), %%eax\n"
sl@0:       "  orl 8(%1), %%eax\n"
sl@0:       "  orl 12(%1), %%eax\n"
sl@0:       "  test $0xff000000, %%eax\n"
sl@0:       "  jz 2f\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm1\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm0\n"
sl@0:       "  pshuflw $0xff, %%xmm1, %%xmm1\n"
sl@0:       "  pshuflw $0xff, %%xmm0, %%xmm0\n"
sl@0:       "  pshufhw $0xff, %%xmm1, %%xmm1\n"
sl@0:       "  pshufhw $0xff, %%xmm0, %%xmm0\n"
sl@0: 
sl@0:       "  pxor %%xmm5, %%xmm1\n"
sl@0:       "  pxor %%xmm5, %%xmm0\n"
sl@0: 
sl@0:       "  movq (%0), %%xmm3\n"
sl@0:       "  movq 8(%0), %%xmm2\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm3\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm2\n"
sl@0: 
sl@0:       "  pmullw %%xmm1, %%xmm3\n"
sl@0:       "  paddw %%xmm6, %%xmm3\n"
sl@0:       "  pmulhuw %%xmm4, %%xmm3\n"
sl@0:       "  pmullw %%xmm0, %%xmm2\n"
sl@0:       "  paddw %%xmm6, %%xmm2\n"
sl@0:       "  pmulhuw %%xmm4, %%xmm2\n"
sl@0:       "  packuswb %%xmm2, %%xmm3\n"
sl@0: 
sl@0:       "  movdqu (%1), %%xmm1\n"
sl@0:       "  paddb %%xmm1, %%xmm3\n"
sl@0:       "  movdqa %%xmm3, (%0)\n"
sl@0:       "2:\n"
sl@0:       "  addl $16, %0\n"
sl@0:       "  addl $16, %1\n"
sl@0:       "  subl $1, %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (middle)
sl@0:       :
sl@0:       :"eax");
sl@0:   }
sl@0:   if (end>0) {
sl@0:     __asm__ __volatile__ ("\n"
sl@0:       "1:\n"
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  testl $0xff000000, %%eax\n"
sl@0:       "  jz 2f\n"
sl@0: 
sl@0:       "  movd (%1), %%xmm1\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm1\n"
sl@0:       "  pshuflw $0xff, %%xmm1, %%xmm0\n"
sl@0:       "  pxor %%xmm5, %%xmm0\n"
sl@0: 
sl@0:       "  movd (%0), %%xmm3\n"
sl@0:       "  punpcklbw %%xmm7, %%xmm3\n"
sl@0:       "  pmullw %%xmm0, %%xmm3\n"
sl@0:       "  paddw %%xmm6, %%xmm3\n"
sl@0:       "  pmulhuw %%xmm4, %%xmm3\n"
sl@0: 
sl@0:       "  paddw %%xmm1, %%xmm3\n"
sl@0:       "  packuswb %%xmm3, %%xmm3\n"
sl@0:       "  movd %%xmm3, (%0)\n"
sl@0: 
sl@0:       "2:\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  subl $1, %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (end)
sl@0:       :
sl@0:       :"eax");
sl@0:   }
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_sse2_3, composite_over_argb, OIL_IMPL_FLAG_SSE2);
sl@0: 
sl@0: 
sl@0: static void
sl@0: composite_over_argb_const_src_mmx (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "  movl (%1), %%eax\n"
sl@0:       "  movd %%eax, %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0xff, %%mm0, %%mm3\n"
sl@0:       "  pxor %%mm5, %%mm3\n"
sl@0:       "1:\n"
sl@0:       "  movq %%mm3, %%mm1\n"
sl@0:       "  movd (%0), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  decl %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_over_argb_const_src_mmx, composite_over_argb_const_src, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: static void
sl@0: composite_add_argb_mmx (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       "1:\n"
sl@0:       "  movd (%1), %%mm0\n"
sl@0:       "  movd (%0), %%mm2\n"
sl@0:       "  paddusb %%mm0, %%mm2\n"
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  decl %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_add_argb_mmx, composite_add_argb, OIL_IMPL_FLAG_MMX);
sl@0: 
sl@0: static void
sl@0: composite_add_argb_const_src_mmx (uint32_t *dest, uint32_t *src, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       "  movd (%1), %%mm0\n"
sl@0:       "1:\n"
sl@0:       "  movd (%0), %%mm2\n"
sl@0:       "  paddusb %%mm0, %%mm2\n"
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  decl %2\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_add_argb_const_src_mmx, composite_add_argb_const_src, OIL_IMPL_FLAG_MMX);
sl@0: 
sl@0: static void
sl@0: composite_in_over_argb_mmx (uint32_t *dest, uint32_t *src, uint8_t *mask, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "1:\n"
sl@0:       "  movd (%2), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0x00, %%mm0, %%mm1\n"
sl@0: 
sl@0:       "  movd (%1), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  movd (%0), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0: 
sl@0:       "  pshufw $0xff, %%mm2, %%mm1\n"
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm0, mm1)
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  addl $1, %2\n"
sl@0:       "  decl %3\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (mask), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_in_over_argb_mmx, composite_in_over_argb, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: static void
sl@0: composite_in_over_argb_const_src_mmx (uint32_t *dest, uint32_t *src, uint8_t *mask, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0: 
sl@0:       "  movd (%1), %%mm3\n"
sl@0:       "  punpcklbw %%mm7, %%mm3\n"
sl@0:       "1:\n"
sl@0:       "  movd (%2), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0x00, %%mm0, %%mm1\n"
sl@0: 
sl@0:       "  movq %%mm3, %%mm2\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  movd (%0), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0: 
sl@0:       "  pshufw $0xff, %%mm2, %%mm1\n"
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm0, mm1)
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $1, %2\n"
sl@0:       "  decl %3\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (mask), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_in_over_argb_const_src_mmx, composite_in_over_argb_const_src, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: static void
sl@0: composite_in_over_argb_const_mask_mmx (uint32_t *dest, uint32_t *src, uint8_t *mask, int n)
sl@0: {
sl@0: #if !defined(__WINSCW__) && !defined(__WINS__)      
sl@0:   __asm__ __volatile__ (
sl@0:       MMX_LOAD_CONSTANTS
sl@0:       "  movd (%2), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0:       "  pshufw $0x00, %%mm0, %%mm3\n"
sl@0: 
sl@0:       "1:\n"
sl@0:       "  movd (%1), %%mm2\n"
sl@0:       "  punpcklbw %%mm7, %%mm2\n"
sl@0:       "  movq %%mm3, %%mm1\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm2, mm1)
sl@0: 
sl@0:       "  movd (%0), %%mm0\n"
sl@0:       "  punpcklbw %%mm7, %%mm0\n"
sl@0: 
sl@0:       "  pshufw $0xff, %%mm2, %%mm1\n"
sl@0:       "  pxor %%mm5, %%mm1\n"
sl@0: 
sl@0:       MMX_MULDIV255(mm0, mm1)
sl@0: 
sl@0:       "  paddw %%mm0, %%mm2\n"
sl@0:       "  packuswb %%mm2, %%mm2\n"
sl@0: 
sl@0:       "  movd %%mm2, (%0)\n"
sl@0:       "  addl $4, %0\n"
sl@0:       "  addl $4, %1\n"
sl@0:       "  decl %3\n"
sl@0:       "  jnz 1b\n"
sl@0:       "  emms\n"
sl@0:       :"+r" (dest), "+r" (src), "+r" (mask), "+r" (n)
sl@0:       :
sl@0:       :"eax");
sl@0: #endif
sl@0: }
sl@0: OIL_DEFINE_IMPL_FULL (composite_in_over_argb_const_mask_mmx, composite_in_over_argb_const_mask, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_MMXEXT);
sl@0: 
sl@0: 
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_in_argb_mmx() {
sl@0: 		return &_oil_function_impl_composite_in_argb_mmx;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_in_argb_mmx, composite_in_argb() {
sl@0: 		return &_oil_function_impl_composite_in_argb_mmx, composite_in_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_in_argb_mmx2, composite_in_argb() {
sl@0: 		return &_oil_function_impl_composite_in_argb_mmx2, composite_in_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_in_argb_const_src_mmx, composite_in_argb_const_src() {
sl@0: 		return &_oil_function_impl_composite_in_argb_const_src_mmx, composite_in_argb_const_src;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_in_argb_const_mask_mmx, composite_in_argb_const_mask() {
sl@0: 		return &_oil_function_impl_composite_in_argb_const_mask_mmx, composite_in_argb_const_mask;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_mmx, composite_over_argb() {
sl@0: 		return &_oil_function_impl_composite_over_argb_mmx, composite_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_mmx_2, composite_over_argb() {
sl@0: 		return &_oil_function_impl_composite_over_argb_mmx_2, composite_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_mmx_3, composite_over_argb() {
sl@0: 		return &_oil_function_impl_composite_over_argb_mmx_3, composite_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_mmx_4, composite_over_argb() {
sl@0: 		return &_oil_function_impl_composite_over_argb_mmx_4, composite_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_mmx_5, composite_over_argb() {
sl@0: 		return &_oil_function_impl_composite_over_argb_mmx_5, composite_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_sse2, composite_over_argb() {
sl@0: 		return &_oil_function_impl_composite_over_argb_sse2, composite_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_sse2_2, composite_over_argb() {
sl@0: 		return &_oil_function_impl_composite_over_argb_sse2_2, composite_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_sse2_3, composite_over_argb() {
sl@0: 		return &_oil_function_impl_composite_over_argb_sse2_3, composite_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_over_argb_const_src_mmx, composite_over_argb_const_src() {
sl@0: 		return &_oil_function_impl_composite_over_argb_const_src_mmx, composite_over_argb_const_src;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_add_argb_mmx, composite_add_argb() {
sl@0: 		return &_oil_function_impl_composite_add_argb_mmx, composite_add_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_add_argb_const_src_mmx, composite_add_argb_const_src() {
sl@0: 		return &_oil_function_impl_composite_add_argb_const_src_mmx, composite_add_argb_const_src;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_in_over_argb_mmx, composite_in_over_argb() {
sl@0: 		return &_oil_function_impl_composite_in_over_argb_mmx, composite_in_over_argb;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_in_over_argb_const_src_mmx, composite_in_over_argb_const_src() {
sl@0: 		return &_oil_function_impl_composite_in_over_argb_const_src_mmx, composite_in_over_argb_const_src;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #ifdef	__SYMBIAN32__
sl@0:  
sl@0: OilFunctionImpl* __oil_function_impl_composite_in_over_argb_const_mask_mmx, composite_in_over_argb_const_mask() {
sl@0: 		return &_oil_function_impl_composite_in_over_argb_const_mask_mmx, composite_in_over_argb_const_mask;
sl@0: }
sl@0: #endif
sl@0: