Re: Can gcc-4.2 be used to measure coverage on SMP platforms?

[Holger Blasum <hbl at sysgo dot com>]

On Sat, Aug 23, 2008 at 11:01:29PM -0700, nimble dude wrote:
> > Searching gcc manuals for further info on gcov did not yield any
> > information. The documentation on SMP and gcov support is rather
> > sparse.
> > - Does gcc-4.2 support atomic_increment of the gcov related information?
> > - If so, is there another option that needs to be specified along with "gcc"
> 2nd try...
> Would be grateful if somebody could confirm one way or the other.

Concerning gcc version 3:
"Especially our implementation does not support threading
and is not able to cope very well with constructors, destructors 
and dynamically loaded objects yet." J Hubicka 2005, Profile driven 
optimisations in GCC, in: Proceedings of the GCC Developers' Summit, 
June 21-24, 2005, Ottawa, Canada, pp. 107-124, 
http://www.gccsummit.org/2005/2005-GCC-Summit-Proceedings.pdf

But that was in 2005, for gcc-4.x you may want to double-check with the 
lcov list https://lists.sourceforge.net/lists/listinfo/ltp-coverage that
also has gcov users. (Note: if you do not want exact numbers 
but just the conservative information whether something is covered at 
all or not you might make use of the non-SMP code instrumentation 
(please carefully observe the generated code on your platform) - one 
of (possibly more) preparations (eg avoid counter overflow) would 
be to modify the instrumentation in find_spanning_tree.)

> > - Does gcc-4.3.1 support atomic_increment of gcov counters
> > natively(i.e, without further flags)?

Is there any indication that atomic_increment of gcov counters 
is supported non-natively, that is with further flags?

-- 
Holger Blasum
SYSGO AG Office Mainz
Web: http://www.sysgo.com