Sharing memory and processor for LA possible?
 

6,343+ is 57% sieved. The sieving is currently somewhat slowed down, probably due to processor heating above the optimal temperature. I am checking that tomorrow. The sieving is currently at the rate of 0.6 million range of special-q per day, which had been previously supposed to be over 1 million.

Although it is trivial to run the sieving over multiple processors, by assigning a different sieving range to each processor. Is it possible to run the Linear Algebra over several computers, sharing the processor and memory which resides on different computers, connected by using a LAN cable?
Today, all the remaining Cunningham numbers are much bigger in size, and for doing the Linear Algebra quickly, we have to enhance it by using multiple processors and shared memory. To save cost, we have to share the resources instead of building a new system specially for Linear Algebra.
I have my Core 2 Quad which is currently sieving for 6,343+ and plan upon buying another Core 2 Duo with upto 1 GB of RAM. Is there any software for sharing the resources residing on different computers for using them together? How about writing that code in msieve for sharing the processor and memory - only little more of effort it should be! I can't build another computer specially for Linear Algebra or send to anyone through post. Sharing the resources should be a nice experiment and implementation.
How was the Linear Algebra for M1039 (Block Weidemann) implemented and run distributed - I would like that software, if not possible for implementing that within the msieve software.

[SIZE=1]Mr. ALEX, SANDER and others. Please. I think that it is time to stop ignoring about me. Time to change your attitude. I am not able to get motivated. November 2006 was the period - when I withdrew from taking SSRI tablets that were initially prescribed by psychiatrist. I have properly diagnosed what had happened during that period. See what would happen upon withdrawal of such tablets over [URL="http://en.wikipedia.org/wiki/SSRI"]here[/URL]. It is quite [I]spontaneous[/I] for causing attention deficit hyperactivity disorder, obsessive compulsive disorder or even schizophrenia, aggression, anxiety, disturbances in sleep, thought and concentration. Time of withdrawal is the period when such symptoms should be peak, than any other time in my life. You should not take it too seriously, since these SSRI discontinuation symptoms should disappear in due course of time. Hope that you understand the situation and start answering (reading) my valid points.

I would wish to be like Mr. Tom Womack, who is leading to factor large Cunningham numbers, with the assistance of others within the factoring forum. It all lies based upon your co-operation.[/SIZE]

If you are not satisfied with my points, read more about my situation over [URL="http://en.wikipedia.org/wiki/SSRI_discontinuation_syndrome"]here[/URL].

I don't see my choice.

[QUOTE=Raman;176961] How about writing that code in msieve for sharing the processor and memory - only little more of effort it should be![/QUOTE]
This is not true at all.

If it was up to me I would use MPI combined with threads, but whoever wanted to use the MPI version must install the MPI libraries and runtime. Nothing about MPI is trivial.

The software from M1039 was custom-written, I have no details on it. You can contact the paper authors to see if they will share it with you.

I'm SO confused by this thread.

[URL="http://en.wiktionary.org/wiki/TMI"]TMI[/URL], dude!

[QUOTE=jasonp;176992]
If it was up to me I would use MPI combined with threads, but whoever wanted to use the MPI version must install the MPI libraries and runtime. Nothing about MPI is trivial.
[/QUOTE]
I'd use it on our Beowulf! But it'd probably have a user base of 1-2. :smile:

For me, requesting for help with sieving by people all over the world, should be virtually useless, because each person will have his own vast amount of resources, and can do his own big Cunningham Number (or other project - GIMPS, Aliquot, Seventeen or Bust, etc.)

If I were to ask for help with sieving - I would prefer to do so with my friends, my university computers, etc. Lots of resources should be available locally itself. At this time, I am in vacation and have no other resources. Once my college re-opens, I should have access to more resources back.

Distributing the Linear Algebra should be important. Since, it is very trivial to distribute the sieving over multiple computers, by assigning a different sieve range to each processor, sharing the LA over several computers should be done so carefully. It is crucial since all the remaining Cunningham numbers are large enough, so it should be run on a high speed system - otherwise it would take a long time right to complete up, that way. I have no idea absolutely on how to run the Linear Algebra in parallel, so I have to get the assistance of someone, either locally or over the forum.

My B.Tech course is over, and I got all India rank of 129 in GATE 2009. I should be joining IIT for M.Tech degree by this month of August, 2009.

Actually, it does not seem to be a defect in the processor - due to overheating, which I thought about previously, but slowness is because of the poor parameters that I am using - I suspect for that fact.

On one of the cores, I am sieving the rational side with alim = 30 million and rlim = 50 million. And I sieve the special-q range from 35 to 40 million.

On the average, upon close watch - each special-q yields only about 10 relations! Very rarely - more than 20. Although the time taken for sieving each special-q is roughly the same as before - about 12 seconds per processor? Though the number of special-q sieved per sieve region spent up is more frequent, catching in a sievable special-q every 20 or so, I guess that they are the values for which the value of special-q is prime, and that the Legendre symbol (N/q) = 1, right?

That is being the fact that: 6,343+ is about, around 58% sieved, so far, thus.