Sean Hefty wrote:
  The resulting executables will still run without Valgrind, just a
  little bit more slowly than they otherwise would, but otherwise
  unchanged.  When not running on valgrind, each client request
  consumes very few (eg. 7) instructions, so the resulting performance
  loss is negligible unless you plan to execute client requests
  millions of times per second.  Nevertheless, if that is still a
  problem, you can compile with the NVALGRIND symbol defined (gcc
  -DNVALGRIND) so that client requests are not even compiled in.  */

It just seems that we're using two checks where we really want one.  We check to
see if memcheck.h exists, and if it does, we include it.  But if the user
doesn't want valgrind, then we disable using it through NVALGRIND.

How about something more like:

if with-valgrind is set then {
        if memcheck is found then
                INCLUDE_MEMCHECK_H = 1
                if with_valgrind then set CPPFLAGS
        else
                error
}

The error message in your second patch is a little misleading otherwise, since
it could occur even though valgrind support was not requested.

- Sean

Basically, you are right.
I wrote this piece of code based on the libibverbs to be consistent with it, maybe Roland can contribute to this thread too ...

I think that the way it is written now, it allows to the library to be prepared to valgrind anyway and use its header file when possible (to allow the user later on use valgrind) unless the user explicitly don't want to use valgrind (to avoid the minor overhead of marking the buffer as valid) or if the user explicitly specified that he wants valgrind, but the requested header file doesn't exist.

I think that we should keep this as the way i suggested because libibverbs work this way for quite a while and there wasn't any problem
with this implementation and i would like to be consistent with it.

what do you think?

thanks
Dotan

_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to