<- Previous Log Select Different Log Next Log ->  
Searching from 2023-08-22 00:00:00 to 2023-08-22 23:59:59.999999.
Query completed in 0.50 seconds
[2023-08-22 02:45:05] <Lucifer_arma> There's an interesting problem that had never ever occurred to me
[2023-08-22 02:45:05] <armagetronbridge> 10irc:Lucifer_arma| There's an interesting problem that had never ever occurred to me
[2023-08-22 02:45:47] <Lucifer_arma> So, you have a worker that can run two jobs at a time.  You figure if you send timely updates on how many jobs are running back to the scheduler, then the scheduler won't send new jobs while there are already two running
[2023-08-22 02:45:47] <armagetronbridge> 10irc:Lucifer_arma| So, you have a worker that can run two jobs at a time.  You figure if you send timely updates on how many jobs are running back to the scheduler, then the scheduler won't send new jobs while there are already two running
[2023-08-22 02:46:08] <Lucifer_arma> The problem, of course, is that in the narrow amount of time when a job has finished, the scheduler could easily send 4-10 new jobs
[2023-08-22 02:46:09] <armagetronbridge> 10irc:Lucifer_arma| The problem, of course, is that in the narrow amount of time when a job has finished, the scheduler could easily send 4-10 new jobs
[2023-08-22 02:46:43] <Lucifer_arma> so the worker's job queue just keeps growing, even though there's already like 20 on it, because once one job finished, for a very small period of time, it's only running one job, and the scheduler sends a shitload more
[2023-08-22 02:46:43] <armagetronbridge> 10irc:Lucifer_arma| so the worker's job queue just keeps growing, even though there's already like 20 on it, because once one job finished, for a very small period of time, it's only running one job, and the scheduler sends a shitload more
[2023-08-22 02:47:34] <Lucifer_arma> obviously that means the scheduler is going to have to keep it's own list of jobs that have been assigned, and keep its own tally of how many jobs are running on each worker, *and* workers are going to have to determine for themselves how big the queue can be,
[2023-08-22 02:47:34] <armagetronbridge> 10irc:Lucifer_arma| obviously that means the scheduler is going to have to keep it's own list of jobs that have been assigned, and keep its own tally of how many jobs are running on each worker, *and* workers are going to have to determine for themselves how big the queue can be,
[2023-08-22 02:48:08] <Lucifer_arma> and instead of relying on workers to report how many running jobs they have, the workers will instead report whether or not they are accepting jobs, and the scheduler will only assign jobs to workers that are accepting them
[2023-08-22 02:48:08] <armagetronbridge> 10irc:Lucifer_arma| and instead of relying on workers to report how many running jobs they have, the workers will instead report whether or not they are accepting jobs, and the scheduler will only assign jobs to workers that are accepting them
[2023-08-22 02:48:51] <Lucifer_arma> so jobs will now have several possible states:  generated, but not assigned.  Assigned, but queued.  Running.  Finished.
[2023-08-22 02:48:52] <armagetronbridge> 10irc:Lucifer_arma| so jobs will now have several possible states:  generated, but not assigned.  Assigned, but queued.  Running.  Finished.
[2023-08-22 05:37:31] <-- Z-Man has quit (Ping timeout: 245 seconds)
[2023-08-22 05:40:14] --> Z-Man has joined the channel
[2023-08-22 05:59:05] <-- Z-Man has quit (Remote host closed the connection)
[2023-08-22 06:00:01] <-- Z-Man has quit (Remote host closed the connection)
[2023-08-22 06:16:28] --> Z-Man has joined the channel
[2023-08-22 06:17:17] --> Z-Man has joined the channel
[2023-08-22 10:26:22] --> monr0e has joined the channel
[2023-08-22 10:37:32] <armagetron-bridge> 08discord:delinquent| Isn't this already a thing? I'm fairly certain pending states are a thing, even on x86. I'm pretty sure that was what PowerPC's claim to fame was, making that a common feature. I remember when early 00's x86 machines *didn't* have that feature - MSN was a fucking nightmare, especially if you got tabs in like, 05.
[2023-08-22 10:37:32] <armagetronbridge> 08discord:delinquent| Isn't this already a thing? I'm fairly certain pending states are a thing, even on x86. I'm pretty sure that was what PowerPC's claim to fame was, making that a common feature. I remember when early 00's x86 machines *didn't* have that feature - MSN was a fucking nightmare, especially if you got tabs in like, 05.
[2023-08-22 18:03:55] <-- monr0e has quit (Ping timeout: 250 seconds)
[2023-08-22 23:06:31] <Lucifer_arma> this isn't anything to do with that.  This is a project I'm working on for myself, a little parallel processing thing
[2023-08-22 23:06:31] <armagetronbridge> 10irc:Lucifer_arma| this isn't anything to do with that.  This is a project I'm working on for myself, a little parallel processing thing
[2023-08-22 23:27:50] <-- Armanelgtron has quit (Ping timeout: 246 seconds)
[2023-08-22 23:28:07] --> Armanelgtron has joined the channel
[2023-08-22 23:28:08] -!- iridium.libera.chat set mode #armagetron +nt
[2023-08-22 23:28:08] -!- Channel #armagetron created on 2021-05-20 17:23:14 UTC

View entire month
DISCLAIMER: These logs of public chat may contain some content which may not be appropriate for all audiences. Use at your own risk.
Logs from 2006-2009 pulled from wrtlprnft
Format changes at: 2015-08-25, 2017-02-20, and 2020-03-23. Times (2015 and later) should be Eastern.


 
 
 ArmaNelgTron.tk
 © NelgTron 2014-2024. Made for . [About this site] [Credits]