|
From: Yeshurun, M. <mei...@in...> - 2005-09-03 06:27:20
|
Thanks.=20
But I think you should erase these lines from the end of section 3.5.1
in the user manual:
"Memcheck is more simplistic about floating-point loads and stores. In
particular, V bits for data read as a result of floating-point loads are
checked at the load instruction. So if your program uses the
floating-point registers to do memory-to-memory copies, you will get
complaints about uninitialised values. Fortunately, I have not yet
encountered a program which (ab)uses the floating-point registers in
this way."
Regards,
Meir
-----Original Message-----
From: val...@li...
[mailto:val...@li...] On Behalf Of Tom
Hughes
Sent: Saturday, September 03, 2005 9:14 AM
To: val...@li...
Subject: Re: [Valgrind-users] Copying undefined floating point values
In message
<942...@ha...>
"Yeshurun, Meir" <mei...@in...> wrote:
> In Valgrind 3.0, Memcheck does not complain when the value of an
> un-initialized floating point variable is copied to another floating
> point variable.
Correct. Floating point values are now tracked fully just as integer
values have always been so you will only get a complain when an
undefined value is used in a way that can effect the program behaviour.
> Is this a deliberate change? (If so, the documentation should be
> updated.)
It might not be entirely obvious, but this entry in NEWS covers it:
On the plus side, Vex and hence Memcheck tracks value flow properly
through floating point and vector registers, something the 2.X line
could not do. That means that Memcheck is much more likely to be
usably accurate on vectorised code.
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
|
|
From: Yeshurun, M. <mei...@in...> - 2005-09-03 15:58:30
|
In case it wasn't clear, I wasn't complaining, but merely pointing this out in case you hadn't noticed, for the benefit of other users. That would not have been my next question. Regards, Meir=20 -----Original Message----- From: Nicholas Nethercote [mailto:nj...@cs...]=20 Sent: Saturday, September 03, 2005 5:41 PM To: Yeshurun, Meir Cc: Tom Hughes; val...@li... Subject: RE: [Valgrind-users] Copying undefined floating point values On Sat, 3 Sep 2005, Yeshurun, Meir wrote: > But I think you should erase these lines from the end of section 3.5.1 > in the user manual: > > "Memcheck is more simplistic about floating-point loads and stores. In > particular, V bits for data read as a result of floating-point loads are > checked at the load instruction. So if your program uses the > floating-point registers to do memory-to-memory copies, you will get > complaints about uninitialised values. Fortunately, I have not yet > encountered a program which (ab)uses the floating-point registers in > this way." Those lines and related ones have been updated in the 3.0.1 docs. The=20 website is still showing the 3.0.0 docs. In answer to your next question,=20 yes it would be nice to have the latest docs on the website but it's a=20 matter of someone finding the time to do it. Nick |
|
From: Nicholas N. <nj...@cs...> - 2005-09-03 14:41:07
|
On Sat, 3 Sep 2005, Yeshurun, Meir wrote: > But I think you should erase these lines from the end of section 3.5.1 > in the user manual: > > "Memcheck is more simplistic about floating-point loads and stores. In > particular, V bits for data read as a result of floating-point loads are > checked at the load instruction. So if your program uses the > floating-point registers to do memory-to-memory copies, you will get > complaints about uninitialised values. Fortunately, I have not yet > encountered a program which (ab)uses the floating-point registers in > this way." Those lines and related ones have been updated in the 3.0.1 docs. The website is still showing the 3.0.0 docs. In answer to your next question, yes it would be nice to have the latest docs on the website but it's a matter of someone finding the time to do it. Nick |