Palemoon becoming more sluggish as the day progesses? Topic is solved

[PalleP]

Tried x86 trybuild. No noticeable difference.

I really wonder how you after a few hours can conclude this. I expect to test a week, due to the weird nature of this bug.

[Rickkins]

Off-topic:

I don't think it's reporting RAM usage. page file = swap.
To be fair I double-checked myself and I'm using a good chunk of it too; gather it's used up by system processes rather than applications.

Always good to learn that it's not just me....

[el es]

I apologize for my late entry - I just saw this thread.
To work with the cycle collector, you are using: dom.cycle_collector.incremental

From Moonchild's Pale Moon Tweak Guide (PMTG) Last edited 01 Apr 2015
Under Performance/memory tweaks. (slightly paraphrased)

javascript.options.mem.gc_incremental default: true
When set to true, this enables "incremental garbage collection" in Pale Moon.

When set to true, you want to also set
javascript.options.mem.gc_incremental_slice_ms to 20

This is for improved performance of the garbage collector.
(Firefox default is 10, which is a little on the short side).

The PM default in 28.3.0 is still 10.
I think this might have an effect on what we're seeing here.
- - - - - - - - - - -
For help troubleshooting, you can try Process Explorer by Mark Russinovich -
he's the developer of Windows Sysinternals.
Like Task Manager, Process Explorer can show a process that is maxing out the CPU (or memory). It can also show which thread (and the stack) that is causing the problem.
This information is not available with a debugger.

[Moonchild]

The PM default in 28.3.0 is still 10.
I think this might have an effect on what we're seeing here.

Unlikely, but thanks for pointing it out.

[GogrillaMincefriend]

I don't know if it's relevant but I made a post on the performance issue in t'other thread (and another with some evidence of repeated writes to the page file) where I was seeing PM (was v.28.2.2 at the time I was testing that - I also leave PM running for weeks on end) where PM was exhibiting god-awful performance because it seemed to be swapping at every opportunity (private working set was continually "shrinking" but commit charge remained the same).

Well weeks ago I was futzing about trying to get some traces done in windbg and managed to accidentally kill the browser, and since then I've not had the same bad performance so I've been unable to perform any more analysis on the issue myself... if it is the same issue others were experiencing it certainly seems highly intermittent. I've not made any tweaks to the GC yet.

But I'd be interested to see if anyone experiencing the issue in this thread can fire up perfmon/procmon and do the same tests and see if they see the same sort of thing.

[Moonchild]

Part of me wonders if this might have something to do with session storage.
The people who suffer from this, are you all using restored sessions on startup, one way or another? If so, does the issue go away if you clean your session out completely (i.e.: throw away your stored session and re-open pages from history/bookmarks?)

[Rickkins]

Part of me wonders if this might have something to do with session storage.
The people who suffer from this, are you all using restored sessions on startup, one way or another? If so, does the issue go away if you clean your session out completely (i.e.: throw away your stored session and re-open pages from history/bookmarks?)

I always set the session restore to remember and load previous session.
However, if this would solve the problem I certainly could live without it....
Testing now....

[PalleP]

The people who suffer from this, are you all using restored sessions on startup, one way or another? If so, does the issue go away if you clean your session out completely (i.e.: throw away your stored session and re-open pages from history/bookmarks?)

I don't use restored sessions when starting PM, but use a blank page.
But when I restart a hanging PM it will open with the restored sessions, and everything will run smoothly and fast for the next hours.
So your thoughts don't seem to be the problem here, while the changes you made in the new nightly seems to cure the problem.
It is too early to conclude anything, but I have not experienced any hanging or erratic scrolling since I installed it.

[Moonchild]

I don't use restored sessions when starting PM, but use a blank page.

Okay, thanks. That's one thing to strike off then.

while the changes you made in the new nightly seems to cure the problem.
It is too early to conclude anything, but I have not experienced any hanging or erratic scrolling since I installed it.

Ah, that's great news! It's not a "nightly" or regular unstable build, but a special dev branch to try and address this problem for those experiencing it.

[Rickkins]

I realize that a single day is by no means definitive, but since booting this morning with session restore disabled, I have NOT had to restart PM at all today. Definitely progress...

[Cassette]

I've been putting the test build through my usual paces for the past three days and so far so good. I can't say definitively that it's "fixed" because I don't know what triggers the slow downs and it often takes many days for it to happen. If I have any issues, I will post an update, otherwise you can assume I continued to have no issues.

[Moonchild]

Thanks for testing -- it seems the changes I made are an improvement, so I've merged it into master and it will ride the unstable to the next release!

[PalleP]

-- it seems the changes I made are an improvement

They most certainly are, no issues here. Maybe it is even a cure, time will show.
Are these changes expected to improve the Gmail issue?

[Rickkins]

Testing the dev version now as well, with session restore still switched off, except for crashes...

[PalleP]

Well, today I have been using the dev build for a week and I have not experienced any issues at all.
So I am very likely to say that the changes cures the problem, since the official build would certainly have failed several times during the week.
It will be interesting what Cassette and the others with the exact same problem have to say.
And what else have I noticed with cpu and ram usage ect.... Nothing.
And what about the Gmail syndrome? Today I logged into Gmail and PM froze completely after a minute or so. After forced termination of PM, Gmail opened as stored session and I logged out. I looked in about:memory, and the Gmail ghost window was present. So the 2 issues have probably nothing to do with each other.
If any info is wanted to dig deeper into what has actually cured the problem, just ask.

[Rickkins]

Testing dev version with session restore switched off(except for crash recovery) since Sunday. No rebooting the browser in that time...

[PalleP]

Sorry for my post yesterday.
The dev build does not cure the problem
It seems to happen less often, but on the other hand, this time my PM froze completely. This has only happened before if I had used Gmail.
I was only browsing Ebay and opened different listings in seperate tabs to compare prices.
Suddently when I opened a few new tabs, PM froze completely. There were 14 tabs in all.
PM was using almost a whole cpu core, so I opened "Process monitor" to see what PM was doing, using all this power. Absolute nothing according to "Process Monitor". Only profiling.
I have attached a native "Process Monitor" logfile.
I waited 10 minutes to see if PM would unfreeze by itself. That didn't happen so i force terminated it.
I then reopened PM, and all 14 Ebay listings were restored without problems, and after some seconds to load the pages PM went to idle as expected (the 14 pages are actually opened now, cpu usage is almost zero).
I need advice what to try next time it happens to provide useful information.
Here is a screenshot, showing how cpu greedy PM can be.

[badnick]

All those who have this problem: what Windows settings you have made for Performance and Virtual memory?

[Rickkins]

I know that there are some boards that automatically show the user's system specs, providing they have entered them in of course. I dunno if there's any way to add them to this particular board software, but I think it could be useful for easy comparisons to help us maybe find some common answers whatever...

[badnick]

I know that there are some boards that automatically show the user's system specs, providing they have entered them in of course. I dunno if there's any way to add them to this particular board software, but I think it could be useful for easy comparisons to help us maybe find some common answers whatever...

It's about these setting. I remember I had a related issue some 3-4 years ago with Firefox.For virtual memory I used to have my own settings and my last try for solving the issue was to check "automatically manage paging file size for all drivers" then the problem disappeared. People who have this problem worth trying if they set virtual memory in other way than in the picture.