Reinventing wheels?

From RoboWiki
Fragment of a discussion from Talk:Multithreading
Jump to: navigation, search

It's more likely a sign of the state of Java (or the author's knowledge of it) at the time the page was drafted. It looks like the original is over 10 years old.

Besides Toad, do any bots actually make use of multithreading? I've always felt the community considered it to be a "dirty" tactic.

Enamel 32 (talk)03:42, 1 March 2019

I'd be in favour of disabling multi-threading in the rumble, or somehow only allowing a single thread to run at the same time. Otherwise there could be some very dirty tactics like starving the opponent of CPU or memory bandwidth during their turns, which is against the spirit of trying to keep the bots' processing isolated except through the robocode engine.

Skilgannon (talk)21:52, 1 March 2019

I agree that computational tasks should not be allowed to run multithreaded in rumble or even in general robocode, or at least we should not allow threads to be run when other bots are running.

However as long as you do file I/O for data logging purposes, multithreading is essential as it’s not fair for your main threads to run blocking I/O only to get a lot skipped turns.

Anyway, a better solution is to provide this in robocode, so fairness is guaranteed.

Xor (talk)01:27, 2 March 2019

I usually put the engine in debug mode if I am going to be logging a lot of data. Also, robocode does increase your processing time allowed if you do any file IO.

Skilgannon (talk)16:04, 2 March 2019

But debug mode disables security iirc? Then this mode is unsafe for running with arbitrary opponent. Also it seems that skipped turns has no special handling if you do IO. It just increases the max skipped turns allowed before you get disabled

Java already has nonblocking IO support, may be we should support this in robocode as well. Currently doing so requires disabling security.

Xor (talk)01:47, 3 March 2019
Personal tools