mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   No Prime Left Behind (https://www.mersenneforum.org/forumdisplay.php?f=82)
-   -   Team drive #8 k=1400-2000 n=350K-500K (https://www.mersenneforum.org/showthread.php?t=11245)

mdettweiler 2009-01-02 16:54

[quote=henryzz;156336]is prpnet tested enough yet[/quote]
For the most part, yes. I'm not sure how much load a PRPnet server can take (never tested it)--maybe we could give it a try with the n=50K candidates for k=1005-2000? It might handle those a lot better than LLRnet because it automatically deals in batches of results, rather than one at a time.

IronBits 2009-01-02 16:56

Waaaaaaaaaaaaaaaaa
/throws self on floor kicking and thrashing about

Is port 8000 doomed to be shutdown again? :cry:

gd_barnes 2009-01-02 22:56

NO NO GUYS GUYS!! Please don't do anything with n=200K-350K yet. I'm off here for only 9 hours and we're going in multiple directions here and way jumping the gun! lol

We want port 8000 to dry and get back to sieving and the 5th-6th-7th drives. The intent all along was for port 8000 to be loaded up with n=350K-352K and then be dried. AFTER the sieving drive is done by ~Jan. 20th, we can decide how port 8000 should be used. Don't worry David, I'm sure we'll use it for some effort on the k=1005-2000 drives. It'll just be about 2-3 weeks yet.

Max, it'll be very messy very quickly having too many different sieve depths for the different n-ranges when matching sieve files to results. Let's keep it organized. Here's how you will have to match up the results as per the different sieve depths in the future:
n=50K-200K sieved to P=1.8T
n=200K-350K sieved to P=~9T or whatever optimal turns out to be for the n=200K-500K range (we won't bother splitting that range off from n=350K-500K since we're sieving so quickly)
n=350K-352K sieved to P=1.4T
n=352K-500K sieved to P=~9T

That's enough different sieve depths already.

Carlos, do you have some Free-DC folks that would help substantially with a large non-top-5000 range such as n=200K-300K? The question is, how can a non-top-5000 effort be done the most quickly. Even at a low n-range, that's a huge amount of work for ~500 k's and n=100K. I'd love to have you and your friends knock it out if you think you'll have the resources to do it. You could do that at the same time that we have the team drive set up to do n=50K-200K. Like that range, it'd be best if searched by k-value.

As soon as I get the factors from Ian for P=1400G-1800G for k=1005-2000 and n=50K-500K, I'll get the 9th drive set up for that k-range and n=50K-350K. Then we can decide how we want to parse it out. Max, keep in mind that even though this is non-top-5000 work, much of it will be first-time work. We must know that the PRPnet server is working properly.

I am pulling the reference out to n=200K-210K in the first post here. It doesn't belong here.

I will pull my quad off of port 8000 now and move it over to port 400. (I haven't checked other threads; for all I know port 8000 is dry now.) Port 5000 has moved ahead of port 400 so I'm going to start splitting my cores between the two.


Gary

mdettweiler 2009-01-02 23:01

[quote=gd_barnes;156525]GUYS GUYS!! Please don't do anything with n=200K-350K yet. I'm off here for only 9 hours and we're going in multiple directions here and way jumping the gun! lol

We want port 8000 to dry and get back to sieving and the 5th-6th-7th drives. The intent all along was for port 8000 to be loaded up with n=350K-352K and then be dried. AFTER the sieving drive is done by ~Jan. 20th, we can decide how port 8000 should be used. Don't worry David, I'm sure we'll use it for some effort on the k=1005-2000 drives. It'll just be about 2-3 weeks yet.

Max, it'll be very messy very quickly having too many different sieve depths for the different n-ranges when matching sieve files to results. Let's keep it organized. Here's how you will have to match up the results as per the different sieve depths in the future:
n=50K-200K sieved to P=1.8T
n=200K-350K sieved to P=~9T or whatever optimal turns out to be for the n=200K-500K range (we won't bother splitting that range off from n=350K-500K since we're sieving so quickly)
n=350K-352K sieved to P=1.4T
n=352K-500K sieved to P=~9T

That's enough different sieve depths already.

Carlos, do you have some Free-DC folks that would help substantially with a large non-top-5000 range such as n=200K-300K? The question is, how can a non-top-5000 effort be done the most quickly. Even at a low n-range, that's a huge amount of work for ~500 k's and n=100K. I'd love to have you and your friends knock it out if you think you'll have the resources to do it. You could do that at the same time that we have the team drive set up to do n=50K-200K. Like that range, it'd be best if searched by k-value.

I will pull my quad off of port 8000 now and move it over to port 400. (I haven't checked other threads; for all I know port 8000 is dry now.) Port 5000 has moved ahead of port 400 so I'm going to start splitting my cores between the two.


Gary[/quote]
Hmm...I see what you mean. I hadn't considered having to mess with matching up the results with all those different sieve depths. :rolleyes:

Should I ask David to remove the n=200K-210K range from the server, or should we just continue on this range until it's done and load no more after that?

Max :smile:

em99010pepe 2009-01-02 23:03

[quote=gd_barnes;156525]

Carlos, do you have some Free-DC folks that would help substantially with a large non-top-5000 range such as n=200K-300K? The question is, how can a non-top-5000 effort be done the most quickly. Even at a low n-range, that's a huge amount of work for ~500 k's and n=100K. I'd love to have you and your friends knock it out if you think you'll have the resources to do it. You could do that at the same time that we have the team drive set up to do n=50K-200K. Like that range, it'd be best if searched by k-value.


Gary[/quote]

Please keep IB8000 up with the non-top-5000 ranges. Don't worry, the range will be done.

IronBits 2009-01-02 23:18

Indeed, 8 more cores coming shortly. :grin:
Just have to go through the pain of installing an XP Upgrade, so I can do a Vista 64bit upgrade. :(

Keep port 8000 busy with whatever you want, the faster the better :big grin:

Bah, make this at least 2 minutes! :cry:

[LIST=1][*][LIST=1][*]This forum requires that you wait 60 seconds between posts.[/LIST] [/LIST]

gd_barnes 2009-01-03 00:50

[quote=mdettweiler;156528]Hmm...I see what you mean. I hadn't considered having to mess with matching up the results with all those different sieve depths. :rolleyes:

Should I ask David to remove the n=200K-210K range from the server, or should we just continue on this range until it's done and load no more after that?

Max :smile:[/quote]

[quote=em99010pepe;156529]Please keep IB8000 up with the non-top-5000 ranges. Don't worry, the range will be done.[/quote]


Dang it guys. You knew the intent of this drive was to only do n=350K-352K. Why was n=200K-210K loaded into the server? Please don't jump the gun in the future. Imagine this: How is Karsten supposed to show this on his pages now for 300+ k's:

k=xxxx; n=50; n=200K-210K; n=350K-352K

All that I ask is that we think ahead a little before impulsively starting on things.

It's all a big mess as far as I'm concerned. Max, you'll have to deal with it. My thinking:

1. Stop the server.
2. Remove all n=200K-220K so that it is dry except for any straggling pairs for n=350K-352K.
3. Remove all results related to n=200K-220K.
4. Redo those k/n pairs in the future so that they are done in a proper sequence and there isn't a big hole in the future range.

Please please when doing things, try to keep as few of gaps as possible in the ranges. We're talking 100's of k's here and we can't be searching multiple n-ranges right in the middle of them with undersieved files.

I'll be able to check the threads here again in about 30 mins. but then will be off for 12 hours.


Gary

mdettweiler 2009-01-03 00:54

[quote=gd_barnes;156559]Dang it guys. You knew the intent of this drive was to only do n=350K-352K. Why was n=200K-210K loaded into the server? Please don't jump the gun in the future. Imagine this: How is Karsten supposed to show this on his pages now for 300+ k's:

k=xxxx; n=50; n=200K-210K; n=350K-352K

All that I ask is that we think ahead a little before impulsively starting on things.

It's all a big mess as far as I'm concerned. Max, you'll have to deal with it. My thinking:

1. Stop the server.
2. Remove all n=200K-220K so that it is dry except for any straggling pairs for n=350K-352K.
3. Remove all results related to n=200K-220K.
4. Redo those k/n pairs in the future so that they are done in a proper sequence and there isn't a big hole in the future range.

Please please when doing things, try to keep as few of gaps as possible in the ranges. We're talking 100's of k's here and we can't be searching multiple n-ranges right in the middle of them with undersieved files.

I'll be able to check the threads here again in about 30 mins. but then will be off for 12 hours.


Gary[/quote]
Okay, I see what you're saying. David, please remove all remaining k/n pairs from the 200K-210K range from the IB8000 server, as Gary said; as for the results files, it should be OK to leave those as they are because I should be able to easily sort out the few extra results when processing them later.

Max :smile:

gd_barnes 2009-01-03 01:06

[quote=IronBits;156537]Indeed, 8 more cores coming shortly. :grin:
Just have to go through the pain of installing an XP Upgrade, so I can do a Vista 64bit upgrade. :(

Keep port 8000 busy with whatever you want, the faster the better :big grin:

Bah, make this at least 2 minutes! :cry:

[LIST=1][*][LIST=1][*]This forum requires that you wait 60 seconds between posts.[/LIST][/LIST][/quote]


No more work in port 8000 please. Please move them to port 400. Thanks.


Gary

em99010pepe 2009-01-03 02:04

Grow up Gary. Let that range be on the server.

gd_barnes 2009-01-03 14:25

No worries. It'll be there soon enough. Keep in mind that we are removing candidates by sieving much faster than we can LLR them at this point, especially the P=1400G-1800G factors, which have not been removed from the file yet.

Ian can correct me if I'm wrong but I think he said at a previous time that he'll be done with the above sieving range in ~2 days. Then we'll be able to start n=50K-200K. By mid Jan., we'll then have a good n=350K-500K range that we can work on.

n=350K-352K was only a small 'emergency' range to work on to avoid missing any top-5000 primes. To start LLRing additional ranges would be a waste of CPU resources.


Gary


All times are UTC. The time now is 18:35.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.