uint8_t ADMImage::substract(ADMImage *src1,ADMImage *src2)
{
#ifdef ADM_CPU_X86
uint32_t r1,r2;
if(CpuCaps::hasMMX())
{
return tinySubstractMMX(YPLANE(this),YPLANE(src1),YPLANE(src2),src1->_width*src1->_height);
}
#endif
return tinySubstract(YPLANE(this),YPLANE(src1),YPLANE(src2),src1->_width*src1->_height);
}
uint8_t ADMImage::substract(ADMImage *src1,ADMImage *src2)
{
#if 1 && defined( ARCH_X86 ) || defined (ARCH_X86_64)
uint32_t r1,r2;
if(CpuCaps::hasMMX())
{
return tinySubstractMMX(YPLANE(this),YPLANE(src1),YPLANE(src2),src1->_width*src1->_height);
}
#endif
return tinySubstract(YPLANE(this),YPLANE(src1),YPLANE(src2),src1->_width*src1->_height);
}
static uint8_t tinySubstractMMX(uint8_t *dst, uint8_t *src1, uint8_t *src2,uint32_t l)
{
int delta;
uint32_t ww,hh;
uint8_t *s1,*s2,*d1;
int ll,rr;
ll=l>>2;
rr=l&3;
s1=src1;
s2=src2;
d1=dst;
#ifdef GCC_2_95_X
__asm__(
"pxor %mm7,%mm7"
::
);
#else
__asm__(
"pxor %%mm7,%%mm7"
::
);
#endif
for(int x=0;x<ll;x++)
{
__asm__(
"movd (%0),%%mm0 \n"
"movd (%1),%%mm1 \n"
"punpcklbw %%mm7,%%mm0 \n"
"punpcklbw %%mm7,%%mm1 \n"
"paddw %%mm0,%%mm0 \n"
"psubusw %%mm1,%%mm0 \n" // mm1=sum
"packuswb %%mm0, %%mm0\n"
"movd %%mm0,(%2) \n"
:: "r"(s1),"r"(s2),"r"(d1)
);
s1+=4;
s2+=4;
d1+=4;
}
__asm__(
"emms\n"
::
);
if(rr) tinySubstractMMX(d1, s1, s2,rr);
return 1;
}
static uint8_t tinySubstractMMX(uint8_t *dst, uint8_t *src1, uint8_t *src2,uint32_t l)
{
int delta;
uint32_t ww,hh;
uint8_t *s1,*s2,*d1;
int ll,rr;
ll=l>>2;
rr=l&3;
s1=src1;
s2=src2;
d1=dst;
ADM_CLEAR_MM7();
for(int x=0; x<ll; x++)
{
__asm__ volatile(
"movd (%0),%%mm0 \n"
"movd (%1),%%mm1 \n"
"punpcklbw %%mm7,%%mm0 \n"
"punpcklbw %%mm7,%%mm1 \n"
"paddw %%mm0,%%mm0 \n"
"psubusw %%mm1,%%mm0 \n" // mm1=sum
"packuswb %%mm0, %%mm0\n"
"movd %%mm0,(%2) \n"
:: "r"(s1),"r"(s2),"r"(d1)
:"memory"
);
s1+=4;
s2+=4;
d1+=4;
}
ADM_EMMS();
if(rr) tinySubstractMMX(d1, s1, s2,rr);
return 1;
}