Combined sieving discussion

[ltd]

Originally Posted by NeoGen

And a question... what does it mean "15000, Merge with SOB when done till here!!! "?
After that value do we have to change to SOB for sieving or all will continue the same? (with proth sieve)

I will try to give you a complete description. As allways with PSP it is all your choice what you want to do.

To make the choices a little bit clearer i will give a short summary about the sieving so far.

PSP startet with a sieve range max n=2Mil and that was extended to run with an upper bound of n=20Mil.
This dat file was sieved upto p=85T.
This year we decided to extend the search to an upper bound of n=50M.
It was done cause there is the chance to sieve together with SOB and they had already changed to an upper bound of n=50M and combined sieving only makes sence if the ranges are identical.
To have a chance for some combined sieving together with SOB we have to catch up with there work or the additional load to them would be to big.

So now we come to the choices:
1. sieve with the normal dat file ranges lower then 85T. Advantages: Higher number of factors found. Easiest to do. Recovers factors missed before due to software/human errors
Disadvantage: It is second pass sieving so most factors found are between 20Mil and 50Mil.
2. Choose ranges out of the area marked as "Available as firstpass test NOT combined"
Advantage: Here you can still use the normal dat file and as it is first pass sieving you will find factors from the complete set.
Disadvantages: There are not that many ranges left in this group. ( Or the p is much to big to make sence at the moment)
3. Choose one of the combined ranges.
Advantage: You help PSP and SOB at the same time.
Disadvantage: You need a special SOB.dat with 21K.(I can put them on the server on request)
The client is about 15% slower then on a not combined range.
There is lot more manual work to be done as you need to reserve the range on two different lists and also you have to report the factors at two different locations.

Hope that makes it a little bit clearer. If not please ask again until i learn to write answers that can be understand.

Lars

[Mystwalker]

Originally Posted by ltd

You need a special SOB.dat with 21K.(I can put them on the server on request)

Request!

[ltd]

Originally Posted by Mystwalker

Request!

File can be found here:

http://www.ldausch.de/test/sievecomb.zip

[NeoGen]

Thanks Lars!
I'm starting to get more of the "big picture" now that I read your explanation.

The third option seems to be the most profitable, as it helps two projects at the same time, but a couple of questions still popped in my mind.

Joining that yahoo group is mandatory if we want to do combined sieving, right? (I click the link and always get the message "You are not a member of the group SierpinskiSieve.")

I understand that the sieving part of this project helps find factors that later ease the work on the PRP'ing part. But which of the three options helps more to the PRP'ing? (which one finds more factors, or better factors... or something like that?)

[NeoGen]

Also... I'm having difficulty getting anything from ldausch.de (stats... combined sieving zip file...)
Anything wrong there in the backend or is it me?

[ltd]

You are right in principle is the thrid option the most profitable.
The second and third option are the best for the short term as they have a bigger chance to find factors that remove a prp test short before it wuld be handed out.

But it is mandatory that at least some work is done on the first option to make the dat file smaller. ( The best way to get a smaller file is a prime but it looks like nobody wants to find one. )

It is your choice. I will use my machines to balance out between the different choices. So if all of you stay with catch up i will do the combined ranges and vice versa.

For the Reservation of the SOB ranges it is necessarry to join the yahoo group. But i think there is an alternative to make things easier for us all.
That approach would only have one disadvantage. It does not work if you also want to join the stats at SOB.

What would you think if i reserve a bigger chunk ( lets say 20T) in the name of the PSP project and we handle all the reservation here alone. I would enter all the PSP results into our database as normal and you get your points also as normal.

I would then pack the results for SOB and send them over to joe. He can then enter them into the database of SOB. (No not on my name)
If someone wants to make a reservation of his own he can still do it but must then handle the reservations on yahoo and also the return of the results to SOB on his own.

What doe you all think about that approach to make life easier.

Lars

[Mystwalker]

Being only mildly stats-addicted, I would prefer such procedure.

[vfrey]

Originally Posted by ltd

What doe you all think about that approach to make life easier.

I think this would be the best, cause I will not create a Yahoo-account to make reservations on the SOB-list.

[ltd]

Originally Posted by NeoGen

Also... I'm having difficulty getting anything from ldausch.de (stats... combined sieving zip file...)
Anything wrong there in the backend or is it me?

At least the server is reachable. ( I am updating the stats at the moment.)
I can not check how good/bad the connectivity is from the outside world cause the server is located at the same provider where i have my DSL connection.
But at least i have no reports of an outage of the server.

Lars

[NeoGen]

Now it's working all right for me too. I can see the stats, download the sob files and all that. It must have been a temporary thing.

And my vote will also go for you getting us some ranges to crunch here. I don't feel like joining SOB and/or Yahoo.

[ltd]

I have send joe a mail to make the reservation of a chunk of 20T.
I will give you an information when it is available.

Lars