Question: Why do we need to add eight to a loop?

[peter5232]

TinyMaix/src/arch_cpu.h

Lines 21 to 41 in 9487854

	TM_INLINE void tm_dot_prod(mtype_t* sptr, mtype_t* kptr,uint32_t size, sumtype_t* result)
	{
	sumtype_t sum=0;
	uint32_t i = 0;
	uint32_t cnt = (size>>3)<<3; //8
	for(; i+8-1 <cnt; ){
	sum += sptr[i]*kptr[i];i++;
	sum += sptr[i]*kptr[i];i++;
	sum += sptr[i]*kptr[i];i++;
	sum += sptr[i]*kptr[i];i++;
	sum += sptr[i]*kptr[i];i++;
	sum += sptr[i]*kptr[i];i++;
	sum += sptr[i]*kptr[i];i++;
	sum += sptr[i]*kptr[i];i++;
	}
	for(; i <size; i++){
	sum += sptr[i]*kptr[i];
	}
	*result = sum;
	return;
	}

Is there any skills?

[Zepan]

it is unroll to speedup

[jonnor]

Answered. Issue can be closed