On Sunday 13 January 2008, Nicholas Nethercote wrote:
> On Fri, 11 Jan 2008, Josef Weidendorfer wrote:
> 
> > An interesting thing for cachegrind/callgrind would be to get global figures
> > on cache events related to data names. This would involve calling
> > VG_(get_dataname_and_offset) for every read/write access, which is a lot of
> > overhead.
> >
> > A solution would be that a tool can build up its own optimized data 
> > structure
> > at the time the debug info of an ELF object is read in. One would need a 
> > hook
> > for this...
> 
> A while back I actually started writing a separate tool that did exactly 
> this.  I didn't get very far with it, though.

What was the problem?
Perhaps 2 years (VG2) ago I started with this too, by adding an hook into the
debug reader, and at least got to the point to print out the list of global
data variables with read/write counts in the end.

However, the idea was to get this data at least on a function-by-function basis,
and make it work with dynamic allocations, too.
That was the point where I had no further time for it.

I never checked the slowdown, though.

Josef

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
Valgrind-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/valgrind-developers

Reply via email to