prime95 and Chrome
 

I'm running prime95 with Priority 1. For just about all processes and programs that I run, the priority system seems to be working, prime95 drops down and the other process takes over.

However, Chrome does not work that way. For the first few weeks of running prime95 on my new computer everything worked fine, including Chrome. However, at some point (and I don't know what changed if anything), Chrome stopped taking priority over prime95. It takes a long time to change tabs, type anything, open Chrome, let alone actually perform anything. But the instant that I stop all my worker threads, Chrome catches back up to itself and everything I clicked or tried to do snaps into place.

Is this a known issue with a known fix? Any suggestions? I've been working like this for about three weeks now and I'm getting tired of pulling up prime95 and stopping it every time I want to go online. Also, my girlfriend tends to forget to start prime95 back up which defeats the purpose of having it running.

Any help or suggestions would be greatly appreciated! Thanks!

[url]http://mersenneforum.org/showthread.php?t=19502[/url]

I'm glad to see someone else is having this problem too. Sorry you're having the problems, though. I know they can be frustrating. Other people use this PC but until now they wouldn't even know Prime is running. Now, I'm considering just giving up and running just three workers so that chrome can respond properly.

Yeah, I saw that thread right after I posted mine. Did loading an older version of Chrome solve your issue at all?

There is not only Chrome, but other programs too, including Acrobat Reader (newer versions after v9, that is years ago already), just to say one of the most known tool. All this programs have in common the fact that they have parts of them (threads) running at priority 1, same as P95, so they won't push P95 out of the way.

For all these, there is a simple solution, discussed already many times, "pause one thread while running xxxx program". See undoc.txt :smile:

[edit: the bad guy in me says: use firefox! hehe]

[QUOTE=LaurV;380052]There is not only Chrome, but other programs too, including Acrobat Reader (newer versions after v9, that is years ago already), just to say one of the most known tool. All this programs have in common the fact that they have parts of them (threads) running at priority 1, same as P95, so they won't push P95 out of the way.[/QUOTE]I noticed once, when installing Acrobat Reader X, that the install program had that problem.:bangheadonwall:

I looked into changing the priorities. In Windows 7's task manager, I can right click processes to see their priority and change it. There are multiple Chrome.exe's (it has always been this way), some of which run at "normal" priority and some of which run at "below normal" priority. What confuses me is prime95.exe runs at "normal" priority. There is also a "low" which is not used.

Setting everything to normal did nothing. Same issues. Is the 1-10 numbering system a Windows 95 relic?

As to Prime95 priority, George has explained that it operates at a different level than that accessed through Task Manager > Processes. I don't remember the terminology for these different mechanisms, but P95 is supposed to run so that it always yields to other processes, regardless of the Normal setting seen in Task Manager. Unfortunately, a number of Adobe applications, including Photoshop, get stalled by P95 running on all cores. I used
[CODE]PauseWhileRunning=photoshop[3],bridge[3][/CODE] in Prime.txt for quite some time when the CPU was completely occupied. The numbers in square brackets indicate how many cores to stop.

Currently, I have this line commented as I'm only using 6 of 8 cores for P95.

What you can set in task manager as priority, from lowest to highest, translated to the "numbers" scale of P95 into something like 3, 5, 6, 7, 9, 11 (take that with a grain of salt, it can be like 3, 5, 6, 7, 8, 10, or a bit different, according with how old your OS is, and with if memory won't fail me, you can google for it).

You are not able to access the lowest/highest "numbers" from task manager. Those "low, below normal" etc, are limited to the priority [U]class[/U] you are in. The full priority is a larger number. Also, whatever you set there (in task manager), affects only the main process/thread, and not the "subprocesses/subthreads" which were [U]already spawned[/U]. However, the security model that the cpu is using, says that a process [U]can not[/U] spawn a thread with a higher priority than itself. So, think about like if you are not admin, then you can not execute a program in admin mode, but if you are admin, you can execute program in both admin and user mode. Same with priorities, you can not start a process with priority 3 if you have priority 2.

Therefore, if you set a process to "below normal", all processes spawned by it [U]after that moment[/U] will run "below normal" (which again, is like a screen-saver priority, somewhere at 5 or 7, on P95's scale).

Real life example:

I use [B]Aliqueit [/B]to factor aliquots sequences. I usually start it with /LOW in the batch, but sometime I forget, and start it normally by typing the command. Aliqueit will start [B]yafu [/B]to factor large composites, and during the ECM phase, yafu will start [U]8 copies[/U] of [B]gmp/ecm[/B] for EACH curve (times 8) he is doing, i.e. start [B]ecm[/B], do a curve, [B]ecm [/B]exits and sends resut to [B]yafu[/B], if the number of curves is not enough, another [B]ecm [/B]process is spawn, and so on. Now the computer runs like shit and AutoCad/etc lags a lot, and I suddenly remember that I forgot about priorities. I go to task manager and set the priority of [B]aliqueit [/B]to "low".

What's happening then? Well, the computer still runs like shit, until the current index is factored and [B]yafu [/B]exits, because [B]yafu[/B] was started [U]before[/U], so it still runs "normal pri". I must go to [B]yafu[/B], and set [U]its[/U] priority to low too, so the [U]current[/U] instance of [B]yafu[/B] will run "low pri". But then, AutoCad still runs like shit for few seconds until the eight copies of [B]ecm [/B]end their job and exit. (I could set their priority to low too, but is too much effort, they will finish in few seconds anyhow, and re-spawn). Then, Autocad runs well. New [B]ecm [/B]spawned by [B]yafu [/B]will have low priority, and when the C120 is factored and [B]yafu [/B]exits, [B]aliqueit [/B]will go to the next index and launch [B]yafu [/B]again, also low priority, because now aliqueit has low priority, and he can't spawn "normal pri" processes. Then everything works as from the book in AutoCad. I hope this example fixes some things. And I used AutoCad only as example, any resource-demanding tool will fit.

TL;dr version:
for details about priorities and classes see [URL="http://msdn.microsoft.com/en-us/library/windows/desktop/ms686277%28v=vs.85%29.aspx"]here[/URL] and [URL="http://msdn.microsoft.com/en-us/library/windows/desktop/ms685100%28v=vs.85%29.aspx"]here[/URL].

A recent update of chrome caused this problem for me too.I put the following line in prime.txt and it seems to have solved the issue.
[CODE]PauseWhileRunning=chrome.exe[1][/CODE]

It was annoying losing a day's worth of crunching every time I forgot to turn Prime95 back on.

Is there a way to control which worker is stopped?

[QUOTE=garo;380127]A recent update of chrome caused this problem for me too.[/QUOTE]

I have some reservations about the silent upgrade techniques of Chrome.
Imagine if one were just about to place a bid in the last moments of an auction when an upgrade commences.
Instead of having an entire worker thread suspended when there is the potential of blocking contention at lowest priority, would it be preferable for a throttle style of operation (a portion of time active, with brief periods of inaction)?
I tend to leave lots of Chrome tabs active all the time.

Throttling=xx(%)

Put this in prime.txt, as in 'throttling=90'

The problem some see in this solution is that it works by pausing workers for the specified amount. In the example above, workers pause for 10% of the time. This causes constant temperature fluctuations in the CPU, which increases stress on its structure.

[QUOTE=kladner;380147]Throttling=xx(%)

Put this in prime.txt, as in 'throttling=90'
[/QUOTE]

The line that I've sometimes used was spelled like:
[CODE]Throttle=33[/CODE]

I read this thread after a several minutes widespread process blockage following a reboot today, but I don't really know the cause, and it came good.
I don't see any of the vast number of chrome threads with priority below 6.
Maybe some malware checking agent interplay was blocking for me?
I was thinking more like 99% duty. I don't really expect a thread should rely on any significant, perhaps rare, time slice at lowest priority.
I wouldn't think it reasonable that a browser window might block itself due to something it might put in play at lowest priority.

OUCH! You are correct about the syntax. Pardon my blooper. :gah:

[QUOTE=snme2pm1;380154]Maybe some malware checking agent interplay was blocking for me?[/QUOTE]
I am Firefox to my teeth, so again, take this with a grain of salt: it may not be chrome itself, but again some Adobe tools (see the discussion about Acrobat above). For example, some page uses Flash to play its animation, and if Chrome is waiting for some feedback from flash, and flash runs in idle pri.. You got the picture. This would justify why when you start Chrome everything is fine, but as long as you work and access "some" pages, with flash on them, etc... Not necessary flash, I use it as an example only. Do your own research before trusting me here :smile:

On the other aspects, I agree with you about Chrome updating policies. You can set it from somewhere to ask you before doing updates, but its mechanism is still very unsafe. In theory, Google can do whatever they like in your computer, or this is what my hacker [STRIKE]feelings[/STRIKE] friends are telling me. That is why I use Firefox, nothing is as good as an open browser... [URL="https://www.mozilla.org/en-US/firefox/desktop/fast/"]The speed[/URL] is secondary :smile:

I'd rather just shut the entire core down. It's working perfectly. It picks worker number four.