|
From: Zhou, J. \(Jack\) <zh...@al...> - 2007-08-20 02:42:07
|
Hi, I am using Valgrind in a automatic testing tool. It's really powerful. Just one question about spacetime: Why is the output of massif tool designed in form of spacetime? How is the spacetime suppose to be used to evaluate a program's performance? All I care about is the space that the program occupy. Is there any possibility that I can get the memory(space) or relative memory rather than spacetime. Is there any paper or material that I can refer to? Thanks in advance. Regards, Jack ----------------------------------------------------------------- IMSE Team, R&D, Lucent Tech., QD, China 86-532-88615642 http://ihgpweb.ih.lucent.com/~zhouj/ |
|
From: Nicholas N. <nj...@cs...> - 2007-08-20 07:39:58
|
On Mon, 20 Aug 2007, Zhou, Jie (Jack) wrote:
> I am using Valgrind in a automatic testing tool. It's really powerful.
>
> Just one question about spacetime:
>
> Why is the output of massif tool designed in form of spacetime? How is
> the spacetime suppose to be used to evaluate a program's performance?
> All I care about is the space that the program occupy. Is there any
> possibility that I can get the memory(space) or relative memory rather
> than spacetime.
> Is there any paper or material that I can refer to?
Spacetime measurements came out of memory heap profilers for Haskell. See
this paper for details:
@InProceedings{Roj:ldvu1996,
author = {Niklas R{\"o}jemo and Colin Runciman},
title = {Lag, drag, void and use -- heap profiling and
space-efficient compilation revisited},
booktitle = {Proceedings of the ACM SIGPLAN International Conference
on Functional Programming (ICFP '96)},
pages = {34--41},
address = {Philadelphia, Pennsylvania, USA},
month = jun,
year = 1996,
}
I implemented Massif to use the same approach. However, it doesn't work as
well for programs in other languages -- Haskell's memory usage is quite
unusual due to its laziness.
For this reason, I'm partway through a rewrite of Massif. You can check out
the branch holding it like this:
svn co svn://svn.valgrind.org/valgrind/branches/MASSIF2
It's still experimental, but should give you an idea of what I'm aiming for.
I haven't had time to work on it recently, but hope to spend some time on it
soon.
Nick
|
|
From: Zhou, J. \(Jack\) <zh...@al...> - 2007-08-21 07:53:59
|
Hi,Nick,
Thanks for your response.
I skimmed the paper you referred in the mail. Is the spacetime from =
massif the sum of lag/use/drag/void spacetime or one or some of them?=20
Regards,
Jack
-----Original Message-----
From: Nicholas Nethercote [mailto:nj...@cs...]=20
Sent: 2007=C4=EA8=D4=C220=C8=D5 15:40
To: Zhou, Jie (Jack)
Cc: val...@li...
Subject: Re: [Valgrind-users] Does anybody know about spacetime
On Mon, 20 Aug 2007, Zhou, Jie (Jack) wrote:
> I am using Valgrind in a automatic testing tool. It's really powerful.
>
> Just one question about spacetime:
>
> Why is the output of massif tool designed in form of spacetime? How is
> the spacetime suppose to be used to evaluate a program's performance?
> All I care about is the space that the program occupy. Is there any
> possibility that I can get the memory(space) or relative memory rather
> than spacetime.
> Is there any paper or material that I can refer to?
Spacetime measurements came out of memory heap profilers for Haskell. =
See=20
this paper for details:
@InProceedings{Roj:ldvu1996,
author =3D {Niklas R{\"o}jemo and Colin Runciman},
title =3D {Lag, drag, void and use -- heap profiling and
space-efficient compilation revisited},
booktitle =3D {Proceedings of the ACM SIGPLAN International =
Conference
on Functional Programming (ICFP '96)},
pages =3D {34--41},
address =3D {Philadelphia, Pennsylvania, USA},
month =3D jun,
year =3D 1996,
}
I implemented Massif to use the same approach. However, it doesn't work =
as=20
well for programs in other languages -- Haskell's memory usage is quite=20
unusual due to its laziness.
For this reason, I'm partway through a rewrite of Massif. You can check =
out=20
the branch holding it like this:
svn co svn://svn.valgrind.org/valgrind/branches/MASSIF2
It's still experimental, but should give you an idea of what I'm aiming =
for.=20
I haven't had time to work on it recently, but hope to spend some time =
on it=20
soon.
Nick
|
|
From: Nicholas N. <nj...@cs...> - 2007-08-21 22:52:33
|
On Tue, 21 Aug 2007, Zhou, Jie (Jack) wrote: > Thanks for your response. > I skimmed the paper you referred in the mail. Is the spacetime from massif > the sum of lag/use/drag/void spacetime or one or some of them? It's the sum, because Massif doesn't pay attention to the use of the allocated blocks, just their lifetimes. Nick |
|
From: Zhou, J. \(Jack\) <zh...@al...> - 2007-08-22 01:14:05
|
Nick, Thanks very much! Jack -----Original Message----- From: Nicholas Nethercote [mailto:nj...@cs...]=20 Sent: 2007=C4=EA8=D4=C222=C8=D5 06:51 To: Zhou, Jie (Jack) Cc: val...@li... Subject: RE: [Valgrind-users] Does anybody know about spacetime On Tue, 21 Aug 2007, Zhou, Jie (Jack) wrote: > Thanks for your response. > I skimmed the paper you referred in the mail. Is the spacetime from = massif=20 > the sum of lag/use/drag/void spacetime or one or some of them? It's the sum, because Massif doesn't pay attention to the use of the=20 allocated blocks, just their lifetimes. Nick |