/*
 * Tested with:
 * gcc version 4.8.2 (Ubuntu 4.8.2-19ubuntu1)
 *
 * With and without OMP, this works just fine with no special options:
 * $ gcc test2.c -o test2 && ./test2 alpha bravo charlie
 * 0x7fffb15cfee0 ./test2
 * 0x7fffb15cfee0 alpha
 * 0x7fffb15cfee0 bravo
 * 0x7fffb15cfee0 charlie
 *
 * Using "-mavx2 -DUSE_VTYPE" works fine too.
 *
 */

#include <stdio.h>
#include <string.h>
#ifdef _OPENMP
#include <omp.h>
#endif
#if USE_VTYPE && __AVX2__
#include <immintrin.h>
#endif

/*
 * I had to place this in a separate function to avoid having the unaligned
 * branch optimized away (even using -O0), since gcc is totally convinced it
 * simply can't be unaligned!
 */
void out(char *pt)
{
	if ((size_t)pt & 31)
		printf("%s unaligned!\n", pt);
	else
		puts(pt);
}

/*
 * Exact same stuff as in test.c, but put in a separate function. I'm not
 * sure it actually ends up inlined, but anyhow it still works fine.
 */
static inline void func(int argc, char **argv, int i)
{
#if USE_VTYPE && __AVX2__
		__m256i _pt[1024 / sizeof(__m256i)];
		char *pt = (void*)_pt;
#else
		__attribute__ ((aligned(32))) char pt[1024];
#endif

		sprintf(pt, "%p ", pt);
		strcat(pt, argv[i]);
		out(pt);
}

int main(int argc, char **argv)
{
	int i;

#ifdef _OPENMP
#pragma omp parallel for
#endif
	for (i = 0; i < argc; i++) {
		func(argc, argv, i);
	}
	return 0;
}