Re: [Valgrind-developers] BB Granularity in Cachegrind?

[Julian S. <js...@ac...>]

Hi Josef

> recently I checked cachegrind in CVS HEAD.
> It gives out its information on BB level.
> E.g. for function foobar(), it gives cost values for
> foobar, foobar+20, foobar+50.
>
> I don't think this is intended, because the results look strange together
> with the usage of source line granularity. And it skrews up efficiency in
> the used hashes (35 BBs per source file ??).

I just checked a fix for this into CVS.  The old behaviour should have
returned now.  Maybe you can test it if you have a spare minute?

What is the status re making kcachegrind work well with the cvs head,
and therefore the soon-to-arrive 2.0 release?

See the attached message -- it looks like it would be cool if kcg worked
with the cvs head.  CVS head contains bug fixes which fix Jody's
problem, and I really don't want to make any further releases of the 1.0.X
series since the head is now so vastly superior to it.  Ideally we could
coordinate a little so that when v-2.0 finally ships (end Jan) there
is already good kcachegrind support for it.

Finally ... I think there is an outstanding issue re loading weak symbols
(didn't you mail me about that?)  But I see in the cvs head this ...

            snaffle_it
               =  ( (ELF32_ST_BIND(o_symtab[i].st_info) == STB_GLOBAL ||
                     ELF32_ST_BIND(o_symtab[i].st_info) == STB_LOCAL ||
		     ELF32_ST_BIND(o_symtab[i].st_info) == STB_WEAK)
                    &&
                    (ELF32_ST_TYPE(o_symtab[i].st_info) == STT_FUNC ||
                     (VG_(needs).data_syms 
                      && ELF32_ST_TYPE(o_symtab[i].st_info) == STT_OBJECT))
                  );

(vg_symtab2.c:1650) so that problem only exists in 1.0.X ?  Can you
clarify (I'm unsure here).

Thanks,

J

---------------------------------------------------------------------

cachegrind vs GNOME
Date: Wed Dec 11 04:16:09 2002
From: Jody Goldberg <jo...@gn...>
To: js...@ac..., nj...@ca...

valgrind-1.0.x works nicely with GNOME2 applications provided one adds
        --alignment=8

Many thanks for your excellent work.  It's reduced my dependence on
people with access to purify.  Sadly cachegrind-1.0 does not appear
to accept the alignment flag.  I tried hand tweaking things to hard
code the parm to 8 to no avail.  The 1.1.0 release appears to work
nicely, but does not seem compatible with kcachegrind or its
callgraph patch.

Any suggestions ?
    Jody