On Monday 22 November 2004 20:37, Rick Pali wrote:
> A few days ago we were suggesting improvements to Prime95 and one of them
> was some sort of notification that the software couldn't get a new
> assignment or contact the server. This was a timely suggestion as I just
> discovered one of my copies of Prime95 (v23.8.1) was having trouble
> contacting the server and lost almost four days of processing time over it.
>
> The strange thing about these occurrences, and I've seen them before, is
> simply killing the process and restarting the program fixes everything
> nicely. I'd love for Prime95 to alert me if it hasn't been processing for,
> say an hour.

Three minutes ought to be (more than) enough (allow a couple of tries to 
recover the server IP address from a nameserver plus no response to the TCP 
SYN sent to start the comms channel).

 However it seems to be obvious that this is a Prime95/mprime software 
problem as it occurs on both linux and windoze versions. Personally I'd have 
thought that the solution is to create a new thread each time communication 
with the server is required - the new thread just does the comms and 
terminates when finished, normal processing can continue as soon as the 
thread has been created provided worktodo.ini isn't empty. The point here is 
that it doesn't matter (much) if the comms thread fails or hangs up, the 
worst that will happen is a small amount of system resources are "lost", even 
then if memory is needed modified data pages from an idle/hung thread will 
get forced to swap file.

Some of my linux systems run with automatic server comms disabled; I run a 
cron job which looks for prime.spl and executes "mprime -c" if and only if it 
exists. This seems to be bulletproof and also solves the "problem" of 
arranging the server comms to occur at a specific time of day.

Regards
Brian Beesley
_______________________________________________
Prime mailing list
[EMAIL PROTECTED]
http://hogranch.com/mailman/listinfo/prime

Reply via email to