GPU to 72 status... - Page 179

kladner · 2013-02-12, 22:25

Quote:

Originally Posted by kracker

MaxHighMemWorkers, 2, 3, or 4 what is the difference?

Sorry. I should have explained my settings more thoroughly. To recap:

In prime.txt:
"PauseWhileRunning=photoshop[3],bridge[3]"
I have a hex core Phenom II, running P-1 on all six. This line causes three workers to pause when either Photoshop or Adobe Bridge starts up, and resume when they quit.

"LowMemWhileRunning=photoshop,bridge"
This makes the remaining workers switch to Stage 1 work if any of them is running Stage 2. When P'shop and/or Bridge shut down, S2 work will resume, though not necessarily on the same workers as before.

In local.txt:
"MaxHighMemWorkers=4"
This limits the number of workers which can be in Stage 2 at the same time. This allows more memory to be allocated to each instance in S2. Stage 1 completes more quickly than Stage 2. If the number of HighMemWorkers is half or less of the total workers, you will build up a backlog of assignments with S1 completed which are waiting for an opening on the S2 front. At some point then, you have to do something to let that backlog get processed. On the other hand, if the number of HighMemWorkers is greater than half of the total there is a good chance that the S2's will keep up with the S1's.

kracker · 2013-02-12, 23:53

Quote:

Originally Posted by kladner

Sorry. I should have explained my settings more thoroughly. To recap:

In prime.txt:
"PauseWhileRunning=photoshop[3],bridge[3]"
I have a hex core Phenom II, running P-1 on all six. This line causes three workers to pause when either Photoshop or Adobe Bridge starts up, and resume when they quit.

"LowMemWhileRunning=photoshop,bridge"
This makes the remaining workers switch to Stage 1 work if any of them is running Stage 2. When P'shop and/or Bridge shut down, S2 work will resume, though not necessarily on the same workers as before.

In local.txt:
"MaxHighMemWorkers=4"
This limits the number of workers which can be in Stage 2 at the same time. This allows more memory to be allocated to each instance in S2. Stage 1 completes more quickly than Stage 2. If the number of HighMemWorkers is half or less of the total workers, you will build up a backlog of assignments with S1 completed which are waiting for an opening on the S2 front. At some point then, you have to do something to let that backlog get processed. On the other hand, if the number of HighMemWorkers is greater than half of the total there is a good chance that the S2's will keep up with the S1's.

Ahh, I get it now, tnx!

chalsall · 2013-02-13, 21:03

Just a status update...

At the suggestion of Jerry, I have added a new field to the LLTF and DCTF forms: GHz Days of work.

How this works is if a number is entered here (>0), the "Number of assignments" field will be ignored, and instead the requested number of GHz Days of work will be assigned (up to the nominal limit of assignments allowed per request; 100 for most people). Please note that at least one assignment will be issued, even if it's more than the GHzDays field specifies.

This is largely intended for spider use, but if anyone would like to see this implemented on the other forms (e.g. LL P-1) please let me know.

Also, Aramis... You will be pleased to know that (at least here) programmers listen to their users. Scott has come up with an elegant solution which will hopefully avoid any confusion for MISFIT users.

Lastly, I'm continuing to monitor the assignment load on Primenet. Yesterday the assignment count went back up to about 1,000 a day. If needed, I'll lower the Pledge level back down to 72 to ensure we're ready to release at that level. Also, the "What Makes Sense" option has changed to mean "Lowest TF Level" from "Lowest Exponent". The thinking is to ensure that the candidates at 72 bits are not reassigned to go to 73, so they will remain available to release back to Primenet if needed.

GPU72.com out.

James Heinrich · 2013-02-13, 21:18

I took a break from my usual LMH-TF work and LLTF'd 100 exponents 71-72, found 3 factors. If the demand remains high I can chew through another 100.

chalsall · 2013-02-13, 21:22

Quote:

Originally Posted by James Heinrich

I took a break from my usual LMH-TF work and LLTF'd 100 exponents 71-72, found 3 factors. If the demand remains high I can chew through another 100.

Please.

James Heinrich · 2013-02-13, 21:29

I've grabbed 200 then.

chalsall · 2013-02-13, 21:31

Quote:

Originally Posted by James Heinrich

I've grabbed 200 then.

Thanks!

kracker · 2013-02-14, 00:06

What is more useful(et needed)? 70 to 71, 71 to 72, etc?
(right at the moment)

EDIT: w00t looks like one of my laptops found a P-1 factor a hour or so ago.

chalsall · 2013-02-14, 00:09

Quote:

Originally Posted by kracker

What is more useful(et needed)? 70 to 71, 71 to 72, etc?
(right at the moment)

Anything going to 72.

kracker · 2013-02-14, 00:15

Quote:

Originally Posted by chalsall

Anything going to 72.

Thanks.

ixfd64 · 2013-02-14, 00:17

It might be a good idea to lower the default pledge level back to 72 bits right now.

2013-02-13, 21:03	#1961
chalsall If I May "Chris Halsall" Sep 2002 Barbados 10011000110110₂ Posts	Just a status update... At the suggestion of Jerry, I have added a new field to the LLTF and DCTF forms: GHz Days of work. How this works is if a number is entered here (>0), the "Number of assignments" field will be ignored, and instead the requested number of GHz Days of work will be assigned (up to the nominal limit of assignments allowed per request; 100 for most people). Please note that at least one assignment will be issued, even if it's more than the GHzDays field specifies. This is largely intended for spider use, but if anyone would like to see this implemented on the other forms (e.g. LL P-1) please let me know. Also, Aramis... You will be pleased to know that (at least here) programmers listen to their users. Scott has come up with an elegant solution which will hopefully avoid any confusion for MISFIT users. Lastly, I'm continuing to monitor the assignment load on Primenet. Yesterday the assignment count went back up to about 1,000 a day. If needed, I'll lower the Pledge level back down to 72 to ensure we're ready to release at that level. Also, the "What Makes Sense" option has changed to mean "Lowest TF Level" from "Lowest Exponent". The thinking is to ensure that the candidates at 72 bits are not reassigned to go to 73, so they will remain available to release back to Primenet if needed. GPU72.com out. Last fiddled with by chalsall on 2013-02-13 at 21:16 Reason: s/nominal limit allowed/nominal limit of assignments allowed/

2013-02-14, 00:06	#1966
kracker "Mr. Meeseeks" Jan 2012 California, USA 2³·271 Posts	What is more useful(et needed)? 70 to 71, 71 to 72, etc? (right at the moment) EDIT: w00t looks like one of my laptops found a P-1 factor a hour or so ago. Last fiddled with by kracker on 2013-02-14 at 00:08

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Status	Primeinator	Operation Billion Digits	5	2011-12-06 02:35
62 bit status	1997rj7	Lone Mersenne Hunters	27	2008-09-29 13:52
OBD Status	Uncwilly	Operation Billion Digits	22	2005-10-25 14:05
1-2M LLR status	paulunderwood	3*2^n-1 Search	2	2005-03-13 17:03
Status of 26.0M - 26.5M	1997rj7	Lone Mersenne Hunters	25	2004-06-18 16:46

2013-02-13, 21:18	#1962
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 23×149 Posts	I took a break from my usual LMH-TF work and LLTF'd 100 exponents 71-72, found 3 factors. If the demand remains high I can chew through another 100.

2013-02-13, 21:29	#1964
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 23×149 Posts	I've grabbed 200 then.

2013-02-14, 00:17	#1969
ixfd64 Bemusing Prompter "Danny" Dec 2002 California 2×11×109 Posts	It might be a good idea to lower the default pledge level back to 72 bits right now.