Re: How do I pursue a gnome-shell memory leak ?

A lot of this has to do with the fact that texture memory is counted as mapped in our process, even though it's really on the dedicated GPU memory.

So, every pixel of every window is counted by being in our process (and also in the X server), because it's part of our address space, but they're all sharing the same actual memory on the GPU.

It depends on the driver, of course, but Intel and NVIDIA has this issue.

If you can try running gnome-shell with valgrind or massif, that will give you more accurate allocation counts, since it will count all the times that some code runs malloc(). The profiling results from massif would be useful regardless, but I think you'll find that we only allocate around 300M or so.

On Thu, Dec 5, 2013 at 7:54 AM, Petko Ditchev <pditchev gmail com> wrote:
Yes , I guess it's related . I dismissed that bug for being too old to be related , but the comments are recent and similar to what I have . Thanks for the pointer.

On 12/05/2013 02:50 PM, Germán Póo-Caamaño wrote:
On Thu, 2013-12-05 at 14:32 +0200, Petko Ditchev wrote:
I see that the process starts eating memory in no particular way .
Sometimes it stays at ~120MB , and some times it starts rising and can
reach 1.5GB . I alt+F2 : 'r' , to restart the shell and put out the fire
, but the issue has to get fixed at some point , and I don't know what
additional info I can give in a bug report . I'm on Manjaro , updated
(gnome-shell .
Maybe related to

desktop-devel-list mailing list
desktop-devel-list gnome org


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]