[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-bugs-dist
Subject:    [valgrind] [Bug 371916] execution tree xtree concept
From:       Josef Weidendorfer <bugzilla_noreply () kde ! org>
Date:       2016-11-10 17:38:38
Message-ID: bug-371916-17878-yrmWgPddD7 () http ! bugs ! kde ! org/
[Download RAW message or body]

https://bugs.kde.org/show_bug.cgi?id=371916

--- Comment #17 from Josef Weidendorfer <josef.weidendorfer@gmx.de> ---
(In reply to Philippe Waroquiers from comment #16)
> >  About the changes in the malloc
> >   wrappers, I think it would be better if the tool can register handlers,
> >   and these handlers then call e.g. VG_(XTMemory_Full_alloc).
> I guess this could be done, but I think this would be less practical
> than the current approach:  the malloc replacement is today
> calling a tool replacement function.

Ah, I see. Then that's fine with me.

> > About the XTree API, why do you need these add/sub variants?
> > Why not just have a function to get a pointer to the value to be updated?
> > Any reduction operation may be useful, such as min/max.
> Yes, any operation can be done in the add/sub : these functions
> get a pointer to the value being updated.

I just wondered if the xtree API could work without having to pass
add/sub_data_fn handlers to make it simpler. I see that the add
handler is used for calculating the totals in the callgrind output.
What about "reduce" instead of "add"? It's a reduction operation to
be used for propagating the value up the call chains. You could do
this before the callgrind output to get smaller files.
If callgrind_annotate/kcachegrind will be able at some point
to consume call chains, it has to be another format anyway.


> As a follow-up, I intend to work on a 'syscall xtree' which e.g. might
> capture min/max (or maybe an histogram) of values.

Yes, statistics are interesting: min/max/variance/average and histograms.
I always wanted to do that in callgrind...

-- 
You are receiving this mail because:
You are watching all bug changes.=
[prev in list] [next in list] [prev in thread] [next in thread]