Ltd's Things to do List (help if you can)

[ltd]

List of things planned for the future.

1. sieve interface to primegrid. This will enable automatic factor transmission
from the BOINC server to PSP.
2. LLR interface between primegrid and PSP. This will prepare us to hand out llr work to the BOINC world and receive residues. (no date sheduled yet)
3. llrnet server for doublecheck work. To give everybody with slower systems an oportunity to help i plan to start a double check effort for all results
returned by llrnet so far.
4. modifications on the database backend and the scripts running the system.
Their are some things i want to improve. Their are no known errors in the scripts at the moment but their is some room for performance gains which would allow point 5
5. Create some more stats to show the progress and the contribution to the project.
6. update the html pages.
7. build i nice web interface for www.psp-project.de as it looks horrible.
(That point will be my biggest problem as I am not a designer. If anybody has a suggestion please send it to me)
8. all the time. Housekeeping and checking for primes.

[Citrix]

Could I help out with setting up the double checking llrnet. We can use manual reservation to remove the numbers done with PRP3 and PRP. Then all we have left are the llr numbers that can go onto the llrnet server.

[mdettweiler]

Originally Posted by ltd

7. build i nice web interface for www.psp-project.de as it looks horrible.
(That point will be my biggest problem as I am not a designer. If anybody has a suggestion please send it to me)

Really? I wouldn't say it's horrible--at least it's clean, fast, and self-explanatory. Of course there's still room for improvement, but when it does get redone, please try to keep those virtues.

A quick little improvement that could be made to the existing stats pages, though: On the SR5 part of the server, could you please add links to the home page for the sieve import, and its requisite stats page? That would be great.

[VJS]

Originally Posted by Citrix

Could I help out with setting up the double checking llrnet. We can use manual reservation to remove the numbers done with PRP3 and PRP. Then all we have left are the llr numbers that can go onto the llrnet server.

I'm already working on the prp's, when Lars gets ack from holiday perhaps we may start a double check que. We will have tosee what kind of results I get to see if it's worth double checking.

At this point it looks like it's not worth double checking project wide. Perhaps a few users may need some double checking here and there but thats about it.

[ltd]

First sorry for answering so late.

@Citrix: The problem with the llrnet server is not the server itself but the link with the database. Some of my tables(and transfer scripts) are not designed to work directly with several lrnet servers so I will have to make some changes in the background tasks. I think it is no option to run the DC check server on a file basis as the sieve effort still removes lots of potential DC tests. It would be a waste of resource not to remove them from the queue as soon as possible.

If the first runs show no suspect data it is also possible that i will let the server run dry from time. (I would anounce that in advance)

@Anonymous:

Thanks for the reminder about the sr5 page. I had forgoten about that one. As these are very important links I have already updated the page.
When there are additional stats i will at least builöd a new page for them so that the original one does not get to overloaded.

@VJS:

The other way round when i am back from work and have holiday we will start with some small scale tests if there are problems with the low llrnet results. But i also do not expect enough errors to start a project wide DC effort.

Two more things about the future of DC testing.
THE PSP llrnet server will stay seperated. The firstpass server will never switched to hand out DC tasks. It will always be your decission if you want to do DC tests.

Looking at the stats each DC tests gives the same amount of points as the first pass test has given. So if there is real interest/need for DC testing it will be possible to create additional stats that combine points from both stats runs.

[mdettweiler]

Originally Posted by ltd

@Anonymous:

Thanks for the reminder about the sr5 page. I had forgoten about that one. As these are very important links I have already updated the page.
When there are additional stats i will at least builöd a new page for them so that the original one does not get to overloaded.

Thanks!

Originally Posted by ltd

@VJS:

The other way round when i am back from work and have holiday we will start with some small scale tests if there are problems with the low llrnet results. But i also do not expect enough errors to start a project wide DC effort.

Two more things about the future of DC testing.
THE PSP llrnet server will stay seperated. The firstpass server will never switched to hand out DC tasks. It will always be your decission if you want to do DC tests.

Looking at the stats each DC tests gives the same amount of points as the first pass test has given. So if there is real interest/need for DC testing it will be possible to create additional stats that combine points from both stats runs.

Maybe when PrimeGrid gets started with PSP LLR, they could do doublecheck, sort of like how Riesel Sieve's BOINC side of the project mostly does doublechecking, while the non-BOINC side does first-pass?

[VJS]

Anonymous,

DC is really not a concern at the moment.

Major concerns for Bionic are Sieve first (Which is being done), then LLR (which is in the works).

[ltd]

As VJS said DC is nothing to worry about at the moment.

I only think it will be a nice feature for people with smaller machine or people that want to test the client first before they commit their machines to WU with large runtimes.
The first DC WU would run for about 1-2hours on a modern CPU.

We are doing already small sample DC Tests on different sizes and the errorrates are low enough that i do not expect any surprise.
So the exttra tests would only remove a very small remaining risks that a bad PC was overlooked so far.

[mdettweiler]

I hate to pick another nit about the SR5 stats pages, but I noticed that the home page's title (the stuff that shows up in the title bar of the browser) says "Prime Sierpinski Basic Stats" instead of "Sierpinski/Riesel Base 5 Basic Stats" or whatever should go there.

[Lennart]

Originally Posted by ltd

List of things planned for the future.

1. sieve interface to primegrid. This will enable automatic factor transmission
from the BOINC server to PSP.
2. LLR interface between primegrid and PSP. This will prepare us to hand out llr work to the BOINC world and receive residues. (no date sheduled yet)
3. llrnet server for doublecheck work. To give everybody with slower systems an oportunity to help i plan to start a double check effort for all results
returned by llrnet so far.
4. modifications on the database backend and the scripts running the system.
Their are some things i want to improve. Their are no known errors in the scripts at the moment but their is some room for performance gains which would allow point 5
5. Create some more stats to show the progress and the contribution to the project.
6. update the html pages.
7. build i nice web interface for www.psp-project.de as it looks horrible.
(That point will be my biggest problem as I am not a designer. If anybody has a suggestion please send it to me)
8. all the time. Housekeeping and checking for primes.

Somwhere betwen 4 an 6 you need to put in some sleep

Lennart