#include <stdint.h>
#include <stddef.h>
#include <string.h>

__attribute((optimize("s"))) //a bit faster than -O2/-O3
int my_memcmp(const void *Vl, const void *Vr, size_t N)
{
	typedef size_t uword_tp;
	const unsigned char *l=Vl, *r=Vr;
	size_t ByteWiseCnt;
	size_t aloff = (uintptr_t)l%_Alignof(uword_tp);

	if ( aloff != (uintptr_t)r%_Alignof(uword_tp) ) {
		/*differently aligned, finish with bytewise cmp and let it fallthru to return 0
		 unless the return happens in the bytewise loop*/

		ByteWiseCnt = N;

		//the goto helps keep the output assembly small
		do_ByteWiseCmp: //run for all, at the beginning if equally misaligned, and possibly at the end
		N-=ByteWiseCnt; //<= disables the rest if run for all or at the end; adjusts N if run at the beginning 
		for (; 0!=ByteWiseCnt; ByteWiseCnt--, l++, r++)
			if (*l != *r) return *l-*r;

	}else if(aloff!=0) {
		/*do_ByteWiseCmp until aligned*/
		ByteWiseCnt= sizeof(uword_tp)-aloff < N ? sizeof(uword_tp)-aloff : N;
		goto do_ByteWiseCmp;
	}

	size_t nw = N/sizeof(uword_tp); N %= sizeof(uword_tp);
	for (; nw != 0; nw--,l+=sizeof(uword_tp),r+=sizeof(uword_tp)){
		size_t l0, r0; memcpy(&l0, l, sizeof(uword_tp)); memcpy(&r0, r, sizeof(uword_tp));
		if (l0 == r0) continue;
		else{
		   	if (1) break; //smaller assembly
			else { ByteWiseCnt = sizeof(uword_tp); goto do_ByteWiseCmp; } //larger code
		}
	}

	//if (nw!=0), then ByteWiseCnt will return within the next sizeof(size_t) bytes and
	//nw needs to be >=sizeof(size_t), otherwise it needs to be N
	if (0==(ByteWiseCnt=nw*sizeof(size_t)+N))
		return 0;

	goto do_ByteWiseCmp;
}