[QUOTE=Syd;167339]I currently modify them to connect via TCP, all workers >100 connected from remote, the others are local workers.[/QUOTE]
Just out of curiosity, how do the local workers connect? Do they communicate directly with the SQL server? Or do they work off of some other protocol?

[quote=mdettweiler;167340]Just out of curiosity, how do the local workers connect? Do they communicate directly with the SQL server? Or do they work off of some other protocol?[/quote]

They communicate directly with the SQL server

[QUOTE=Syd;167342]They communicate directly with the SQL server[/QUOTE]
You know, that gives me an idea...doesn't SQL communicate over TCP anyway? Thus, wouldn't it be possible to have remote workers simply have a MySQL client talk to the server over the Internet just like they already do over your network?

[quote=mdettweiler;167344]You know, that gives me an idea...doesn't SQL communicate over TCP anyway? Thus, wouldn't it be possible to have remote workers simply have a MySQL client talk to the server over the Internet just like they already do over your network?[/quote]

Yes it can, but it dosnt work as expected. While inserting factors the tables are locked - together with the RTT over the internet it locks the tables for too long.
The second one is - the local client needs write access to the tables, when made public everyone could mess around with the tables

Maybe someone here wants to try it, here is the TCP worker (amd64) + sourcecode.

Usage:
./worker "<Worker name>" <num cpus>

Its still beta! So far it works on Linux only.

The one previously posted had a little bug, now the corrected version.

Btw, why cant I edit the old one anymore?

[quote=Syd;167454]The one previously posted had a little bug, now the corrected version.

Btw, why cant I edit the old one anymore?[/quote]

The edit window is only 60 minutes, after that you can't edit unless you're a mod.

[quote=Syd;167454]The one previously posted had a little bug, now the corrected version.

Btw, why cant I edit the old one anymore?[/quote]
I've currently got it running on one core of a Core 2 Duo (E4500)--works great! :grin: (Edit: make that two cores.)

Now if only I could get it working on Windows too, then my quad might be able to join the fun for a bit... :smile:

Hmm...I noticed some odd behavior with the worker application. As previously described, I had it running on two cores:
[code]Starting workers, using 2 CPUs
#0: 120 curves with B1: 120000, B2: 90000000 on C118
#1: 120 curves with B1: 120000, B2: 90000000 on C116
#1: 120 curves with B1: 120000, B2: 90000000 on C118
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 120 curves with B1: 120000, B2: 90000000 on C118
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197
#0: 1 curves with B1: 3000000, B2: 7000000000 on C197
#1: 1 curves with B1: 3000000, B2: 7000000000 on C197[/code]
Right after the above output it simply exited without saying anything (leaving me back at the terminal prompt as if it exited normally). This seemed to coincide exactly with the point in time when according to the worker status page, ECM very high limits had completed on that C197 and it moved on to a C114.

Now, what's confusing is that when I was first setting up the program, and it couldn't find ECM/msieve since I didn't have it properly set in my path, it did this exact same thing, that is, simply exit without saying anything. Yet, obviously in the above-described circumstance, the worker application could find ecm and msieve just fine, and continues to do so now that I started the program again--so I can't imagine that this problem could have also been caused by an inability to locate the ecm binary. :huh:

I found a bug in the worker program:
[code]#0: 1 curves with B1: 5000000, B2: 5321930062 on C120
Unknown option: -p+1[/code]It seems that the program is launching GMP-ECM with the command line argument "-p+1". As far as I know, that's an invalid argument; the correct one is "-pp1". (Or does this vary between GMP-ECM versions?)

The program immediately exited silently after the "unknown option" error (which is what I would expect based on previous observed behavior).

[quote=mdettweiler;167471]I found a bug in the worker program:
[code]#0: 1 curves with B1: 5000000, B2: 5321930062 on C120
Unknown option: -p+1[/code]It seems that the program is launching GMP-ECM with the command line argument "-p+1". As far as I know, that's an invalid argument; the correct one is "-pp1".

Or does this vary between GMP-ECM versions?[/quote]

You're right, the switch is -pp1. I'll shutdown the "server" part now and fix all bugs that came up, so the worker wont work now. Thanks for trying it!