On Mon, 2007-08-20 at 10:38 +0200, Samuel Thibault wrote:
> Hi,
>
> Thomas Bushnell BSG, le Mon 20 Aug 2007 00:26:47 -0400, a écrit :
> > 1: I think it should be raised in gnumach itself, not just in the Debian
> > version.
>
> Well, it's the same for THREAD_MAX, TASK_MAX, etc.. ;)
Yes, of cours
Hi,
Thomas Bushnell BSG, le Mon 20 Aug 2007 00:26:47 -0400, a écrit :
> 1: I think it should be raised in gnumach itself, not just in the Debian
> version.
Well, it's the same for THREAD_MAX, TASK_MAX, etc.. ;)
> 3: Doing (2), it seems to me, should best be done by something dynamic
> based upon
At Mon, 20 Aug 2007 02:30:41 -0400,
Michael Casadevall wrote:
> > 1: I think it should be raised in gnumach itself, not just in the
> > Debian
> > version.
> >
> I agree. The value should also be dymanically changeable by the
> user if possible (I am not familiar enough with the code in question
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On Aug 20, 2007, at 12:26 AM, Thomas Bushnell BSG wrote:
On Mon, 2007-08-20 at 02:52 +0200, Samuel Thibault wrote:
Samuel Thibault, le Mon 20 Aug 2007 02:18:44 +0200, a écrit :
Samuel Thibault, le Mon 20 Aug 2007 02:06:18 +0200, a écrit :
int
On Mon, 2007-08-20 at 02:52 +0200, Samuel Thibault wrote:
> Samuel Thibault, le Mon 20 Aug 2007 02:18:44 +0200, a écrit :
> > Samuel Thibault, le Mon 20 Aug 2007 02:06:18 +0200, a écrit :
> > > int vm_object_cached_max = 200;/* may be patched*/
> > >
> > > We should probably raise
Samuel Thibault, le Mon 20 Aug 2007 02:18:44 +0200, a écrit :
> Samuel Thibault, le Mon 20 Aug 2007 02:06:18 +0200, a écrit :
> > int vm_object_cached_max = 200;/* may be patched*/
> >
> > We should probably raise it in debian at least.
>
> I've made it 2000.
That put a simple gn
Samuel Thibault, le Mon 20 Aug 2007 02:06:18 +0200, a écrit :
> int vm_object_cached_max = 200;/* may be patched*/
>
> We should probably raise it in debian at least.
I've made it 2000.
Samuel
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Troub
Samuel Thibault, le Mon 20 Aug 2007 02:06:18 +0200, a écrit :
> Hi,
>
> Samuel Thibault, le Mon 06 Aug 2007 21:10:02 +0200, a écrit :
> > Also, something odd: say I have a big file /tmp/foo
> >
> > $ cat /tmp/foo > /dev/null
> > ... "inactive" grows in vmstat, takes some time...
> > $ cat /tmp/fo
Hi,
Samuel Thibault, le Mon 06 Aug 2007 21:10:02 +0200, a écrit :
> Also, something odd: say I have a big file /tmp/foo
>
> $ cat /tmp/foo > /dev/null
> ... "inactive" grows in vmstat, takes some time...
> $ cat /tmp/foo > /dev/null
> takes almost no time since it's all cached.
> $ find /var -pri
Just a few more details: my box has 400MB memory, and inactive was
evolving between 58 and 64MB, so there was plenty of memory free for the
directories/files entries to be cached without having to eject my big
file. This is all with the >2GB support.
Samuel
--
To UNSUBSCRIBE, email to [EMAIL P
Samuel Thibault, le Mon 06 Aug 2007 21:10:02 +0200, a écrit :
> $ find /var/log -printf ""
Err, that was find /var -printf "" actually. Find /var/log doesn't
trigger the issue, probably because there are not enough
directories/files.
Samuel
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a
Also, something odd: say I have a big file /tmp/foo
$ cat /tmp/foo > /dev/null
... "inactive" grows in vmstat, takes some time...
$ cat /tmp/foo > /dev/null
takes almost no time since it's all cached.
$ find /var/log -printf ""
... "inactive" falls down to the original value
$ cat /tmp/foo > /dev/
Samuel Thibault, le Mon 06 Aug 2007 20:45:53 +0200, a écrit :
> I've ported xosview to GNU/Hurd and had a look during a gcc compilation:
> approximately only half the memory is used. I had a look at
> vm/vm_pageout.c and noticed:
>
> /* When vm_page_external_count exceeds vm_page_external_li
I've ported xosview to GNU/Hurd and had a look during a gcc compilation:
approximately only half the memory is used. I had a look at
vm/vm_pageout.c and noticed:
/* When vm_page_external_count exceeds vm_page_external_limit,
* allocations of externally paged pages stops.
*/
#ifndef
14 matches
Mail list logo
