mersenneforum.org  

Go Back   mersenneforum.org > New To GIMPS? Start Here! > Information & Answers

Reply
 
Thread Tools
Old 2011-11-28, 05:42   #1
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts
Default What exactly does the "Drop checked CPUs" button do?

Does it erase everything but results from the database? Would you be able to recover the computer if it communicated with PrimeNet after dropping it?

And while we're at it, what exactly do the reliability and confidence numbers represent? Like is reliability a percentage? What about confidence, since it's not on a 1 or 100 scale as far as I can tell?

And also why does restarting Stage 2 P-1 take so long?

Last fiddled with by Dubslow on 2011-11-28 at 05:48
Dubslow is offline   Reply With Quote
Old 2011-11-28, 10:11   #2
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22·3·641 Posts
Default

Quote:
Originally Posted by Dubslow View Post

And also why does restarting Stage 2 P-1 take so long?
I think I partly, but not wholly, understand. (On my system, it's a couple of minutes.) However, be warned that I've not done a detailed investigation, and the following are just what I've speculated so far without bothering to get actual times and numbers for support.

Because all the stage 2 workareas must be reallocated and reinitialized again, the operating system has to swap out all the least-recently-used pages, used by other tasks, whose frames are going to be reassigned to the workarea spaces. This amounts to quite a few megabytes of disk writing. (E.g., I usually allocate 1200 MB for stage 2 workareas.) If your hard disk is not the fastest, the time needed for this may be quite noticeable (it's not all done in one efficient chunk write). So, that'd be part of it.

Also, if, like me, your disks are just partitions on one big hard disk, and you've designated one of those as your dedicated swap file, keep in mind that each page-out might require a seek from a track in, say, your system residence disk's partition to a swap partition's track that may be quite a few cylinders away.

I'd think the time needed for the computations of values to be stored in the workareas would be on the order of the time for LL iterations. If a computed value to be stored requires several squarings or other multiplications, that's several iteration-times. If there are quite a few such values, there could be, I suppose, on the order of a few or several hundred iteration-times required.

It also could be that all this stage 2 startup activity may require significantly more paging than simply the page-out of LRU page frames to be reassigned.

I myself am not yet satisfied that those factors sufficiently account for the couple-of-minute delay I've observed. So far I've been willing to chalk it up to my slow-seek slow-write-and-readback disk's sloth.

Last fiddled with by cheesehead on 2011-11-28 at 10:25
cheesehead is offline   Reply With Quote
Old 2011-11-28, 10:37   #3
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

52×7×53 Posts
Default

Quote:
Originally Posted by Dubslow View Post
Does it erase everything but results from the database? Would you be able to recover the computer if it communicated with PrimeNet after dropping it?
You do not lose the credit, if that was the question. All former credit stays. The computer is gone. You can merge computers together (check them, use merge button, if you are sure the entries come from the same physical machine), or whatever. The computer that you deleted (droped) is restored next time when it tries to contact primenet. Like also adding new computers. As I upgrade/change parts very often, sometime primenet detects my computer like something else, and I end up in having more computers in that list than I really have. I go deleting all of them, when I don't remember the names.

For the rest of your questions I have no idea.

Last fiddled with by LaurV on 2011-11-28 at 10:44
LaurV is offline   Reply With Quote
Old 2011-11-28, 10:42   #4
Brian-E
 
Brian-E's Avatar
 
"Brian"
Jul 2007
The Netherlands

7·467 Posts
Default

Quote:
Originally Posted by Dubslow View Post
And while we're at it, what exactly do the reliability and confidence numbers represent? Like is reliability a percentage? What about confidence, since it's not on a 1 or 100 scale as far as I can tell?
They represent how trustworthy your LL tests (including double checks) have been so far. Reliability would indeed be a percentage except that it's a fraction of 1 instead of 100, so 98% is represented as 0.98. Confidence is the number of LL tests on which your reliability score is based. More information in this thread.
Brian-E is offline   Reply With Quote
Old 2011-11-28, 16:45   #5
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts
Default

Thanks Brian-E, that makes much more sense now.
Quote:
Originally Posted by cheesehead View Post
I think I partly, but not wholly, understand. (On my system, it's a couple of minutes.) However, be warned that I've not done a detailed investigation, and the following are just what I've speculated so far without bothering to get actual times and numbers for support.

Because all the stage 2 workareas must be reallocated and reinitialized again, the operating system has to swap out all the least-recently-used pages, used by other tasks, whose frames are going to be reassigned to the workarea spaces. This amounts to quite a few megabytes of disk writing. (E.g., I usually allocate 1200 MB for stage 2 workareas.) If your hard disk is not the fastest, the time needed for this may be quite noticeable (it's not all done in one efficient chunk write). So, that'd be part of it.

Also, if, like me, your disks are just partitions on one big hard disk, and you've designated one of those as your dedicated swap file, keep in mind that each page-out might require a seek from a track in, say, your system residence disk's partition to a swap partition's track that may be quite a few cylinders away.

I'd think the time needed for the computations of values to be stored in the workareas would be on the order of the time for LL iterations. If a computed value to be stored requires several squarings or other multiplications, that's several iteration-times. If there are quite a few such values, there could be, I suppose, on the order of a few or several hundred iteration-times required.

It also could be that all this stage 2 startup activity may require significantly more paging than simply the page-out of LRU page frames to be reassigned.

I myself am not yet satisfied that those factors sufficiently account for the couple-of-minute delay I've observed. So far I've been willing to chalk it up to my slow-seek slow-write-and-readback disk's sloth.
I'm in Windows, and I'm not sure why you're referring to HDD's, because isn't it all in memory? (I have a 7200 RPM partitioned HDD, with Windows on one partition, another for Linux (broken ATM) and one system reserved and swap respectively.)
Dubslow is offline   Reply With Quote
Old 2011-11-28, 22:16   #6
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts
Default

Another question: If you use the Pfactor worktype with a non-Mersenne number (or not prime exponent Mersenne) how will it calculate bounds to use? Will it at all, or just throw it out?
Dubslow is offline   Reply With Quote
Old 2011-11-28, 23:18   #7
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22×3×641 Posts
Default

Quote:
Originally Posted by Dubslow View Post

I'm in Windows, and I'm not sure why you're referring to HDD's, because isn't it all in memory?
"... isn't it all in memory?" Isn't what "it" all in memory?

Do you understand the basic concepts of virtual memory and paging?

The swap files and page files are on disk (which may be a partition of a HDD), not in RAM. "Swap" means to transfer from RAM to disk (or other device), or vice versa. Similarly, the verb "page out" means to write the content of a page frame from RAM to page file on disk, and "page in" means to read the content of a page frame from page file on disk and put it in RAM.

Does that answer your question? (I'm guessing it doesn't.) I'll be glad to give or link to more explanation if you'll explain what you mean by "it all in memory".

Last fiddled with by cheesehead on 2011-11-28 at 23:36
cheesehead is offline   Reply With Quote
Old 2011-11-28, 23:54   #8
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3×29×83 Posts
Default

Whatever Prime95 puts into the 8-11GB I give it for stage two.

Quote:
out all the least-recently-used pages, used by other tasks, whose frames are going to be reassigned to the workarea spaces. This amounts to quite a few megabytes of disk writing. (E.g., I usually allocate 1200 MB for stage 2 workareas.)
That (especially the last sentence) gave me the impression that 'work areas' is the stuff in RAM that Prime95 needs; if there's plenty of it, why can't it all stay in RAM? Better yet: What is 'it', meaning what does P95 put into stage 2 RAM and what would need to go on the HDD in virtual memory?
Dubslow is offline   Reply With Quote
Old 2011-11-29, 00:22   #9
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

769210 Posts
Default

Quote:
Originally Posted by Dubslow View Post
Whatever Prime95 puts into the 8-11GB I give it for stage two.

That (especially the last sentence) gave me the impression that 'work areas' is the stuff in RAM that Prime95 needs; if there's plenty of it, why can't it all stay in RAM? Better yet: What is 'it', meaning what does P95 put into stage 2 RAM and what would need to go on the HDD in virtual memory?
If prime95 were the only program running on your computer, there would be no conflict between its memory use and the memory used by other tasks. However, the operating system uses memory and other tasks (browser, word processor, etc.) use memory, so when prime95 makes its memory request, it's not doing so in front of an unblemished expanse of memory that has never been used for something else.

When prime95 makes its memory allocation requests, the OS will probably satisfy that request by assigning page frames that already have content placed there by other tasks (or itself) earlier. Unless such other tasks have exited or have deallocated their own memory areas, then at least some of those page frame contents are still assigned to those other tasks. Before the OS lets prime95 use them, it has to save the contents of all those previously-used page frames by writing (paging out or swapping out) their content to disk.

That's what I was referring to when I wrote, "... the operating system has to swap out all the least-recently-used pages, used by other tasks, whose frames are going to be reassigned to the workarea spaces."

It's not that what prime95 puts in the workareas is necessarily being paged or swapped out (but see below), it's that whatever has been put there before prime95 is allowed to use that area needs to be written to disk, for future use by the other tasks.

Since prime95 runs at lowest priority, it's even quite possible that one of those other tasks will interrupt prime95's minutes-long stage 2 initialization and cause one or more of prime95's pages to be paged or swapped out, resulting in still more total paging/swapping that has to occur before stage 2 initialization is complete.
cheesehead is offline   Reply With Quote
Old 2011-11-29, 00:43   #10
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts
Default

Huh. It takes 8 minutes on mine -- maybe because it was using 11GB. OTOH, no more than 2 were being used when I started P95, and 11GB leaves 1 G to spare, so I can't see more than 1.5GB needing to be swapped out, and even that I find to be rather doubtful. I still don't think this can account for the whole 8 minutes, but thank you for clearing up much of the process. :) Maybe it isn't the swapping out but rather P95 figuring out what data needs to go into the memory allocated, after it's been cleared for use? That would make more sense that 1.5GB of paging as far as why my initialization takes that much longer than yours (and I have a 2600k so it's definitely not CPU-clogged).
Dubslow is offline   Reply With Quote
Old 2011-11-29, 01:04   #11
bcp19
 
bcp19's Avatar
 
Oct 2011

2A716 Posts
Default

Quote:
Originally Posted by Dubslow View Post
Huh. It takes 8 minutes on mine -- maybe because it was using 11GB. OTOH, no more than 2 were being used when I started P95, and 11GB leaves 1 G to spare, so I can't see more than 1.5GB needing to be swapped out, and even that I find to be rather doubtful. I still don't think this can account for the whole 8 minutes, but thank you for clearing up much of the process. :) Maybe it isn't the swapping out but rather P95 figuring out what data needs to go into the memory allocated, after it's been cleared for use? That would make more sense that 1.5GB of paging as far as why my initialization takes that much longer than yours (and I have a 2600k so it's definitely not CPU-clogged).
Using this example:

Code:
[Nov 26 08:07] Using 1490MB of memory.  Processing 480 relative primes (0 of 480 already processed).
[Nov 26 08:08] M7098941 stage 2 is 2.24% complete. Time: 70.904 sec.
[Nov 26 08:09] M7098941 stage 2 is 13.23% complete. Time: 71.388 sec.
You notice it takes the same amout of time to do the first 2.24% as the next 10.99%? It seems to me that the initialization of the P-1 loads all these primes into memory, then starts working on them, which is why the init takes so long. 8 min for 11G =22.9M per second, which is around the rate I get when copying from HDD to External via USB.
bcp19 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
"Merge CPUs" removed them completly? KradziejSoku PrimeNet 6 2020-10-16 11:04
"Merging" CPUs daxmick Information & Answers 7 2018-01-04 06:46
Prime95 minor bug, "TortureThreads" not checked for maximum threads Explorer09 Software 1 2014-08-19 04:13
How can I bypass this "Like" button? Stargate38 Lounge 11 2014-06-22 19:43
Would Minimizing "iterations between results file" may reveal "is not prime" earlier? nitai1999 Software 7 2004-08-26 18:12

All times are UTC. The time now is 10:00.

Sun Mar 7 10:00:57 UTC 2021 up 94 days, 6:12, 0 users, load averages: 1.97, 2.06, 2.13

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.