Hi,
currently I'm trying to optimize some code using NEON technology, and I ran into a problem bother me a lot.
int8x8_t vext_s8(int8x8_t a, int8x8_t b, const int n) //arm_neon.h
the third param of above function, it should be determined at compile time, or the gcc would complain about it.
I want a way like this as loop times may be different for different callers:
for (i=0;i<N;i++) {
c = vext_s8(a,b,i);
// do something with c...
}
But it won't work. Do I have to write like below for N=4 case?
vext_s8(a,b,0)
//do something
vext_s8(a,b,1)
//do samething
vext_s8(a,b,2)
vext_s8(a,b,3)
it's ugly, but I have no other ways, and, I don't know if this is the right place to seek help either,
Any ideas are appreciated.