|
From: Yeshurun, M. <mei...@in...> - 2005-09-02 05:18:10
|
Tom, you are my savior once again. Simply lowering Valgrind's base all
the way down to the 32GB mark (making Valgrind's part ridiculously
large...) in order to squeeze the client into the first 16GB made
Valgrind run about 5 - 10 times faster on our programs.
Thanks again to all of you.
Meir
-----Original Message-----
From: val...@li...
[mailto:val...@li...] On Behalf Of Tom
Hughes
Sent: Tuesday, August 23, 2005 9:22 AM
To: val...@li...
Subject: Re: [Valgrind-users] PIE slowness
In message
<942...@ha...>
"Yeshurun, Meir" <mei...@in...> wrote:
> I would like to understand why Valgrind is considerably slower when
> built as a PIE.
We don't actually know for sure, and we haven't bothered looking at it
in detail as we plan to get rid of PIE mode completely.
There will be a small slow down because position independent code is
slightly less efficient, but that shoule be very small.
A bigger issue is that memcheck is inefficient at dealing with memory
above the 16Gb mark, but even restricting the client address space
to 16Gb doesn't completely remove the speed penalty.
Tom
--=20
Tom Hughes (to...@co...)
http://www.compton.nu/
-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle
Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing &
QA
Security * Process Improvement & Measurement *
http://www.sqe.com/bsce5sf
_______________________________________________
Valgrind-users mailing list
Val...@li...
https://lists.sourceforge.net/lists/listinfo/valgrind-users
|