Collaboration with Primegrid or independent BOINC project

[KEP]

Per request by chalsall I'm going to make a suggestion in the open forum.

First up, I'm not at all a programmer, so I can't setup, nor can I host a server and I can't make any of the programs needed to make mfakt and Prime95 work with BOINC.

But in private messages I suggested chalsall and George to seek a collaboration with Primegrid. Primegrid already has several GPU running projects, so I (even though I don't know for sure) think that it should be somewhat easy for them to setup a new sub-project, with the sole purpose to help primenet and GPU-72 get all mersenne candidates factored to optimal factordepth and afterwards extensively P-1 tested. I've made no inquiry to the project managers (JimB) at Primegrid, weather they are interested or not in a collaboration with primenet, because I and chalsall think it would be fruitfull to have debate where everyone can have a saying in regards to a collaboration with Primegrid or the possibility of setting up an independent Prime95 BOINC project.

If we choose to set up our own project, we have to first of all show that we are not gone tomorrow, that way a loyal group of crunchers can be attracted to our BOINC project. Also we need some sort of batches and someone to design a cool certificate saying, how many cobblestones has been produced, how many factors has been found, how many tests has been saved and maybe the total amount of P90 GHz years that person has contributed to the project.

If we can find a way to collaborate with Primegrid, we would in very short time, experience very fast progress because many people would seek to achieve a new batch to whatever level they desire, before moving along to other projects. This eager to achieve batches, would most definently help your project get a lot of GPU work done in very short time, plus it will attract people who are unfamiliar with the mersenne search, so not only will you achieve a major boost to the mersenne search, you will also be able to factor all mersenne candidates <1G to optimal bitdepth within most of our lifetimes.

So let me and chalsall know, what you think, should we collaborate or go independent? Should we go BOINC at all? What do you think?

Regards

KEP

[chalsall]

Originally Posted by KEP

...because I and chalsall think it would be fruitfull to have debate where everyone can have a saying in regards to a collaboration with Primegrid or the possibility of setting up an independent Prime95 BOINC project.

...

So let me and chalsall know, what you think, should we collaborate or go independent? Should we go BOINC at all? What do you think?

Please be very careful, there, cowboy!

Just because I encouraged you to take this discussion public, does ***NOT*** necessarily mean I'm on your side.

I'm not necessarily not on your side, either. But please don't try to misrepresent the situation. Beyond the "out-of-the-blue" PMs today, I've never heard from you before.

[pinhodecarlos]

Do not collaborate with PrimeGrid, it is a mistake. Go independent.

[chalsall]

Originally Posted by pinhodecarlos

Do not collaborate with PrimeGrid, it is a mistake. Go independent.

Why do you say that?

Sincerely, I'm interested in knowing.

[KEP]

Originally Posted by pinhodecarlos

Do not collaborate with PrimeGrid, it is a mistake. Go independent.

I can see that it can have some sort of "conflict of interest" to collaborate with Primegrid, since Primegrid has a World Record search of their own. However given the boost that collaborating with Primegrid has given both the Prime Sierpinski Problem aswell as the SR5 conjecture, it can most certainly also have some great advantages to collaborate with Primegrid. Primegrid is a mighty force in the prime searching community and in fact they also have several GPU projects and quite a few programmers that supports their different applications, so out of plain convienience it would (at least in the short term) be a bad thing to rule out Primegrid in the first place.

However given the creditrules that Primenet is working under, I can see that the best option would be to set up your own BOINC project. This also gives the advantage of being able to determine the type and amount of work that gets processed by the BOINC clients.

My main goal of this debate was that I hoped that Primenet could catch up on the trial- and P-1 factoring. That's why I thought that a BOINC platform that runs as a supplement to Primenet, would be a great idea. However I'm no programmer and MySQL and I guess other databases doesn't wanna coorporate with me when I try setting them up, so I can be of little help on the technical part.

If the common conclusion of this thread and the debate I hope to see, concludes in the setup of a BOINC project, I would suggest following subprojects:

LMH-factoring (CPU only)
Trial factoring to high bit-depth (GPU only)
Trial factoring DC (CPU and GPU)
P-1 factoring (CPU only --> since it appears to me that the CPU is more productive)
ECM factoring (CPU and maybe GPU if possible)
LL firstpass tests (CPU only)
LL DC tests (CPU only)

This will give a total of 7 subprojects, where the initial goal should be to stop the trailing behind of the factoring, because the best part would be if the LL tests could have been at least factored to optimal bitdepth and to some extent be P-1 tested, before the final LL check. So to sum-up, my hope of having a BOINC version of Primenet, was that the candidates that gets tested was at least a bit more optimal than they appears to be at current time.

Take care

[pinhodecarlos]

Some real facts.
PrimeGrid stole the following projects without asking the members that were running them:

Sierpinski / Riesel Base 5 Problem (LLR) from mersenneforum.org
The Riesel Problem (LLR) from the original project (Lee)
Cullen Prime Search (LLR) from mersenneforum.org

Also, did you guys ask first Prime95 about this collaboration? Did you guys make a poll to Prime users?
Don't forget that finding a prime in collaboration with PrimeGrid means shared primes in top 5000 ranking.

I give more credits to someone that uses its own code rather just wrapping others code to search for Prime Numbers.

I won't say more. Just think about it.

Best regards,

Carlos

[chalsall]

Originally Posted by pinhodecarlos

Some real facts.
PrimeGrid stole the following projects without asking the members that were running them:

Sierpinski / Riesel Base 5 Problem (LLR) from mersenneforum.org
The Riesel Problem (LLR) from the original project (Lee)
Cullen Prime Search (LLR) from mersenneforum.org

Thanks for those facts, Carlos. Important to consider.

Originally Posted by pinhodecarlos

Also, did you guys ask first Prime95 about this collaboration? Did you guys make a poll to Prime users?

George referred KEP to me (via PM). I encouraged KEP to bring the discussion forward in a public manner.

This is where we are now... Discussing the proposal....

[Prime95]

I would limit a BOINC project to GPUs only doing TF where we need it (DC and LL exponents) with the goal of building up a buffer ahead of the 2 wave fronts. As far as "badges" go, I know very little of acceptable BOINC practices. If users drop out if they can't earn badges, how about badges in each 10M range of exponents?

You'll need a BOINC expert/developer, whether it is Primegrid, NFSNet, or someone new.

You'll need TheJudger's support to upgrade the mfaktc client.

You'll need a server (if the BOINC expert can't provide his own).

You'll need a method of getting work and sending results to the Primenet server. This could take some of my time, or Chris' time if the BOINC project coordinates with GPU72.

[Citrix]

Could the admins add a poll to this thread?

In general I support Primegrid, but I was annoyed when they took over Prime Cullen Prime without asking the key people involved. Currently the only project they are able to provide good support to is the proth search. I feel work has started to stagnate on the other projects. The cullen search has made very little progress over the last 1 year.

They need to pick 2-3 projects and concentrate on them instead of having many small slow projects.

I think they have too much on their plate to take over more work from GIMPS.

[chalsall]

Originally Posted by Citrix

Could the admins add a poll to this thread?

They could. I don't know how to do that, but they could.

Originally Posted by Citrix

I think they have too much on their plate to take over more work from GIMPS.

In my mind the real question is if GIMPS wants to create their own BONIC server and clients.

Very happy to help with the former. Cannot help with the latter.

[pinhodecarlos]

I don't support thieves but I don't run either Prime95. Just warning you guys.