[Opera-Linux] OK, what can I do about Opera's never-ending thirst for memory?!

[Daniel Eckl]

Hi!

Perhaps we should look in another direction as well.

If I can remember correctly, Kenny once stated that he is running out
of file descriptors or such (?) and after looking for this, he saw
that all cache files are constantly opened and in read access. How
about something in the opera caching mech and the file cache is
constantly copied in memory?

Best,
Daniel

2009/6/19 Eirik Byrkjeflot Anonsen <eirik at opera.com>:
> Kenneth Crudup <kenny at panix.com> writes:
>
>> On Thu, 18 Jun 2009, Daniel Eckl wrote:
>>
>>> I think I am puzzled by all these values... I just read 16,5MB and
>>> 8,4MB? Where do I fail in interpreting the writings?
>>
>> I'm thinking these are the sizes of all the pixmaps the X-Server says
>> the process has. I was hoping these would grow quite large, but that's
>> not the case.
>
> No, I'm pretty sure that is the size of all the pixmaps the X server
> process has allocated on behalf of opera (i.e. the memory is allocated
> in the X server process).
>
>>
>> BTW a few hours later, and I had to kill the NVidia-running Opera last
>> night because it grew to a VM of 6GB and a res of 3GB and everything
>> else I had running caused the system to thrash in swap (5.5GB of Opera
>> was on the swap disk).
>>
>> However, the XVnc-running instance of Opera (a direct clone of that one
>> running on the NVidia-based machine) is still hovering around the same
>> 1045MB/750MB VM/Res figures as yesterday.
>>
>> It's *gotta* be either (something related to?) compiz (whichi isn't
>> running on the XVnc box), the NVidia driver, or one of the libraries
>> linked into Opera on behalf of the NVidia driver.
>
> Definitely interesting.
>
> Compiz and the server-side nvidia driver "should" not make much of a
> difference, since none of that code is run in the opera process.
> However, it will affect which messages are passed to opera (in
> particular the order and the number of them), which could theoretically
> trigger some bad behaviour in opera.  Possible, but it seems unlikely to
> me.
>
> On the other hand, the nvidia driver does replace the OpenGL libraries
> that are linked into the applications themselves.  But I don't think
> opera is using OpenGL direct rendering contexts (or any other OpenGL for
> that matter), at which point I would expect that code to be inactive.
>
> Though I guess it is possible that Qt is using OpenGL behind the
> scenes...  I think the nvidia driver has an environment variable you can
> set to force it to never use OpenGL direct rendering.  I wouldn't expect
> it to make any difference to opera, but I would expect it to keep the
> (application-side) nvidia code inactive.
>
> eirik
> --
> Opera-Linux: https://list.opera.com/mailman/listinfo/opera-linux
> More lists: https://list.opera.com/mailman/listinfo/
> Unsubscribe: mailto:opera-linux-request at opera.com?subject=unsubscribe
>
>