Calling libunwind on Linux This is modified from https://gist.github.com/blakejohnson/5831456 Only "make test3" was tested (the rest may require extra things to be installed) You need an installation of Julia. The Makefile assumes this is in /home/yourusername/src/.

Makefile

all: test test2 test3

JULIADIR = $(HOME)/src/julia
JULIALIB = $(JULIADIR)/usr/lib
OPENBLASDIR = $(JULIADIR)/deps/openblas-develop
UNWINDDIR = $(JULIADIR)/deps/libunwind-1.1
# LDFLAGS = -Wl,-no_compact_unwind
LDFLAGS =
CC = gcc

test: unwind.c
	$(CC) -g -DAPPLEBLAS -framework vecLib -lblas $(LDFLAGS) unwind.c -o test

LDFLAGS += -Wl,-rpath,$(JULIALIB)

test2: unwind.c
	$(CC) -g -I$(OPENBLASDIR) -L$(JULIALIB) -lopenblas $(LDFLAGS) unwind.c -o test2

test3: unwind.c
	$(CC) -g -DFORTRANBLAS -I$(OPENBLASDIR) -I$(UNWINDDIR)/include -L$(JULIALIB) $(LDFLAGS) unwind.c -o test3 -lunwind -lopenblas

unwind.c

#include <stdlib.h>
#include <stdio.h>
#include <stddef.h>
#include <signal.h>
#include <sys/time.h>

#define UNW_LOCAL_ONLY
#include "libunwind.h"

typedef ptrdiff_t ptrint_t;

#ifdef APPLEBLAS
	#include <Accelerate/Accelerate.h>
#else
#ifdef FORTRANBLAS
	extern void daxpy_(
	    ptrdiff_t *n,
	    double *da,
	    double *dx,
	    ptrdiff_t *incx,
	    double *dy,
	    ptrdiff_t *incy
	);
#else
	#include "cblas.h"
#endif
#endif

struct itimerval timerprof;
static u_int64_t usecprof = 500;
int running;
int samples;

size_t rec_backtrace(size_t maxsize)
{
	unw_cursor_t cursor; unw_context_t uc;
	unw_word_t ip, offset;
	size_t n=0;
	char fname[64];
	
	unw_getcontext(&uc);
	unw_init_local(&cursor, &uc);
	while (unw_step(&cursor) > 0 && n < maxsize) {
		if (unw_get_reg(&cursor, UNW_REG_IP, &ip) < 0) {
			break;
		}
		fname[0] = '\0';
		unw_get_proc_name(&cursor, fname, sizeof(fname), &offset);
		printf("%p: (%s + 0x%x)\n", (void *)ip, fname, (unsigned int)offset);
	}
	printf("\n");
	return n;
}

static void sprofile_bt(int dummy) {
	if (running) {
		printf("Sample %d\n", samples);
		rec_backtrace(10);

		if (samples-- > 0) {
			timerprof.it_value.tv_usec = usecprof;
			setitimer(ITIMER_REAL, &timerprof, 0);
		}
	}
}

int start_timer(void)
{
	timerprof.it_interval.tv_sec = 0;
	timerprof.it_interval.tv_usec = 0;
	timerprof.it_value.tv_sec = 0;
	timerprof.it_value.tv_usec = usecprof;
	if (setitimer(ITIMER_REAL, &timerprof, 0) == -1)
		return -3;

	running = 1;
	samples = 10;
	signal(SIGALRM, sprofile_bt);

	return 0;
}

void stop_timer(void)
{
	running = 0;
}

int blas_op(int a, double *x, double *y) {
#ifdef FORTRANBLAS
	long n = 1000;
	double da = (double)a;
	long incx = 1, incy = 1;
	daxpy_(&n, &da, x, &incx, y, &incy);
#else
	cblas_daxpy(1000, (double)a, x, 1, y, 1);
#endif

	return y[0];
}

void fill_xy(double *x, double *y, int n) {
	int ct;
	for (ct=0; ct < n; ct++) {
		x[ct] = (double)rand() / (double)RAND_MAX;
		y[ct] = (double)rand() / (double)RAND_MAX;
	}
}

int main(void) {
	int a = 2;
	int b;
	double x[1000], y[1000];
	int ct;

	fill_xy(x, y, 1000);

	start_timer();
	for (ct = 0; ct < 10000; ct++)
		b = blas_op(a, x, y);
	stop_timer();

	return 0;
}