Abolish STRANGE_MALLOC?

[Nicholas Clark <nick@ccl4.org>]

At the moment we have code conditionally compiled on STRANGE_MALLOC that is
simple and assume that malloc has non-pathological behaviour, and realloc()
is the way to go. Or we have the default which is complex, and prefers
malloc() followed by free(). For example in av.c

#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
		Renew(AvALLOC(av),newmax+1, SV*);
#else
		bytes = (newmax + 1) * sizeof(SV*);
#define MALLOC_OVERHEAD 16
		itmp = MALLOC_OVERHEAD;
		while ((MEM_SIZE)(itmp - MALLOC_OVERHEAD) < bytes)
		    itmp += itmp;
		itmp -= MALLOC_OVERHEAD;
		itmp /= sizeof(SV*);
		assert(itmp > newmax);
		newmax = itmp - 1;
		assert(newmax >= AvMAX(av));
		Newx(ary, newmax+1, SV*);
		Copy(AvALLOC(av), ary, AvMAX(av)+1, SV*);
		if (AvMAX(av) > 64)
		    offer_nice_chunk(AvALLOC(av), (AvMAX(av)+1) * sizeof(SV*));
		else
		    Safefree(AvALLOC(av));
		AvALLOC(av) = ary;
#endif


ie pretty much everywhere we compiling that code in the #else.

Should we just get rid of it all, and assume that malloc()/realloc()/free()
work these days? I tried with perlbench on x86 FreeBSD and Linux.
A, D, E are normal perl. B, C, F were also compiled with -DSTRANGE_MALLOC

                           A       B       C       D       E       F
                        ----    ----    ----    ----    ----    ----
AVERAGE (FreeBSD)        100     100     100     100     100     100
AVERAGE (Linux)          100      99     100     101     101      99

So it seems to make little difference here.

Is there a good reason to keep it, and the code complexity?

Nicholas Clark

---

Follow-Ups from:

Andy Dougherty <doughera@lafayette.edu>

---

[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index][Thread Index][Top&Search][Original]

---