void Msharpen::blur_plane(ADMImage *src, ADMImage *blur, int plane)
{
/*
uint64_t mask1 = 0x00001C711C711C71LL;
uint64_t mask2 = 0x1C711C711C710000LL;
uint64_t mask3 = 0x0000200000002000LL;
uint64_t mask4 = 0x0000000000ff0000LL;
const unsigned char *srcp = src->GetReadPtr(plane);
const unsigned char *srcp_saved = srcp;
unsigned char *blurp_saved = blurp;
int src_pitch = src->GetPitch(plane);
int blur_pitch = blur->GetPitch(plane);
int h = src->GetHeight(plane);
int w = src->GetRowSize(plane);
*/
const unsigned char *srcp,*srcpn,*srcpp;
const unsigned char *srcp_saved ;
unsigned char *wk,*wk_saved;
unsigned char *blurp_saved,*blurp ;
int src_pitch;
int blur_pitch;
int work_pitch;
int h;
int w ;
int wh ,ww,hh;
blurp_saved=blurp=blur->GetWritePtr((ADM_PLANE)plane);
srcp_saved=srcp=src->GetReadPtr((ADM_PLANE)plane);
wk_saved=wk=work->GetWritePtr((ADM_PLANE)plane);
ww=src->GetWidth((ADM_PLANE)plane);
hh=src->GetHeight((ADM_PLANE)plane);
src_pitch=src->GetPitch((ADM_PLANE)plane);
blur_pitch=blur->GetPitch((ADM_PLANE)plane);
work_pitch=work->GetPitch((ADM_PLANE)plane);
w=ww;
h=hh;
wk+=work_pitch;
srcpp=srcp;
srcp+=src_pitch,
srcpn=srcp+src_pitch;
int val;
// Vertical only for now
#ifdef ADM_CPU_X86
if(CpuCaps::hasMMX())
{
int off;
#ifdef GCC_2_95_X
__asm__(
ADM_ASM_ALIGN16
"pxor %mm7,%mm7\n"
: : );
#else
__asm__(
ADM_ASM_ALIGN16
"pxor %%mm7,%%mm7\n"
: : );
#endif
int wmod8=w>>3;
for (int y=1; y<h-1 ;y++)
{
off=0;
for (int x =0;x< wmod8; x++)
{
__asm__(
ADM_ASM_ALIGN16
"movq (%0),%%mm0\n"
"movq %%mm0,%%mm6\n"
"punpckhbw %%mm7,%%mm0\n" // High part extended to 16 bits
"punpcklbw %%mm7,%%mm6\n" // low part ditto
"movq (%1),%%mm1\n"
"movq %%mm1,%%mm5\n"
"punpckhbw %%mm7,%%mm1\n"
"punpcklbw %%mm7,%%mm5\n"
"movq (%2),%%mm2\n"
"movq %%mm2,%%mm4\n"
"punpckhbw %%mm7,%%mm2\n"
"punpcklbw %%mm7,%%mm4\n"
"paddw %%mm1,%%mm0\n"
"paddw %%mm5,%%mm6\n"
"paddw %%mm1,%%mm2\n"
"paddw %%mm5,%%mm4\n"
"paddw %%mm0,%%mm2\n"
"paddw %%mm6,%%mm4\n"
"psrlw $2, %%mm4\n"
"psrlw $2, %%mm2\n"
"packuswb %%mm2,%%mm4\n"
"movq %%mm4,(%3)\n" //
: : "r" (srcpn+off),
"r" (srcp+off), "r" (srcpp+off), "r" (wk+off)
);
off+=8;
}
// mod 8 fix
for(int x=wmod8*8;x<w;x++)
{
val=2*srcp[x]+srcpn[x]+srcpp[x];
wk[x]=(val)>>2;
}
srcp+=src_pitch;
srcpp+=src_pitch;
srcpn+=src_pitch;
wk+=work_pitch;
}
__asm__("emms\n");
}
else
#endif
{
for (int y=1; y<h-1 ;y++)
