Difference between revisions of "User talk:AaronR"

From Robowiki
Jump to navigation Jump to search
(Impressive)
(Whoosh!)
 
(28 intermediate revisions by 4 users not shown)
Line 19: Line 19:
  
 
: Wow, that was fast! =D « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 09:39, 15 August 2009 (UTC)
 
: Wow, that was fast! =D « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 09:39, 15 August 2009 (UTC)
 +
 +
:: You are faster on this (but you have orange bar to tell you). But by the way, I stole your signature =) &raquo; <span style="font-size:0.9em;color:darkgreen;">[[User:Nat|Nat]] | [[User_talk:Nat|Talk]]</span> &raquo; 09:46, 15 August 2009 (UTC)
 +
 +
::: That's OK, your arrows point the other way! <insert yet another smiley here> « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 09:51, 15 August 2009 (UTC)
 +
 +
: Welcome back! I've been thinking of you recently while I migrate content. =) Btw, you might be curious to know that Horizon barely missed the cut when [[Anti-Surfer Challenge/Pre-Chat|selecting bots]] for the new [[Anti-Surfer Challenge]]. --[[User:Voidious|Voidious]] 15:19, 15 August 2009 (UTC)
 +
 +
:: Thanks! The new version certainly wouldn't cut it for that challenge, it's way too slow. I think I'll be asking around for ideas on how to speed it up without losing battle performance. « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 19:57, 15 August 2009 (UTC)
 +
 +
::: I'm not sure if you already use this trick, but it can help a lot. You say you precise predict two waves, and I assume you sum the dangers for each one. Well, after you predict the first wave for a movement option, you could check if the danger from that alone is already greater than the lowest danger you've found so far in your other movement options; if it is, no need to predict the second wave at all. If you weight the waves by distance or bullet time, this will happen a lot. Going further, you can always first calculate the danger of the movement option that was safest the previous time. (Props to [[User:Krabb|Krabb]] for originally showing me this.) --[[User:Voidious|Voidious]] 20:11, 15 August 2009 (UTC)
 +
 +
:::: Ah! That will probably help a lot. I'll implement that now and see how much faster it is. (By the way, currently I'm just weighting the waves by a fixed factor--i.e. the wave that will hit second counts half as much as the wave that will hit first. I haven't tweaked that constant at all since it was introduced.) « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 20:16, 15 August 2009 (UTC)
 +
 +
::::: Well, that got 10 rounds of Horizon 1.1 ditto down from 65 seconds to 43 seconds. It's still way slower than I'd like though. Making it surf only one wave brings it down to 32, which is still slow, so I get the feeling that I need to optimize elsewhere, too. « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 21:34, 15 August 2009 (UTC)
 +
 +
::::: For comparison, the previous version, 1.03, does a 10 round mirror match in 10-12 seconds. The only thing I changed that would make any speed difference is the surfing danger function. « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 21:41, 15 August 2009 (UTC)
 +
 +
::::: The other big speed optimization for DC surfing is to cache the nearest neighbors (aka scan cluster, aka similar situations). Do you do that already? You certainly don't need to recalculate them each tick. I calculate and cache the neighbors for each wave if they are not cached, then clear the caches when I'm hit, I detect an onBulletHitBullet, or the first wave I'm surfing changes. --[[User:Voidious|Voidious]] 21:45, 15 August 2009 (UTC)
 +
 +
::::: I'm sure that would speed things up... but that's not the problem. Like I said, the only difference between 1.03 and 1.1 is the WS danger function. By the time I start precise prediction, I've already calculated a GF array and I'm just passing that around. Thanks for the suggestion, though--I'll implement that, too. « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 21:58, 15 August 2009 (UTC)
 +
 +
== Leave of absence ==
 +
 +
While everyone's talking about resurrected Robocoders, don't forget that I'm still around! I just haven't had any time lately because of school. I have this great<sup>[citation needed]</sup> idea for a new<sup>[citation needed]</sup> strategy for melee movement, but I refuse to work on any new bots until I figure out what I broke in the last version of [[Horizon]]! And since I can't find anything obvious in the diffs, you probably won't hear much from me for a while... « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 18:24, 1 October 2009 (UTC)
 +
 +
I sure haven't forgotten... [[Horizon]] 1.1.1b is currently one of [[Dookious]]' 5 losses. :-P --[[User:Voidious|Voidious]] 18:34, 1 October 2009 (UTC)
 +
 +
: Look at the [http://darkcanuck.net/rumble/BattleDetails?game=roborumble&name=voidious.Dookious%201.573c&vs=ar.horizon.Horizon%201.1.1b battle details]. Horizon got creamed, but Dooki crashed on one of the battles and the server recorded the zero score. I know, I'm disappointed too. « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 18:55, 1 October 2009 (UTC)
 +
 +
:: Doh! Sorry if I got your hopes up. =) Thanks for lookin' out. --[[User:Voidious|Voidious]] 23:50, 1 October 2009 (UTC)
 +
 +
I think you are a part of resurrection too.  Btw, we don't have Cite extension installed on this wiki =P --[[User:Nat|<span style="color:#099;">Nat</span>]] [[User talk:Nat|<span style="color:#0a5;">Pavasant</span>]] 02:45, 2 October 2009 (UTC)
 +
 +
== Welcome and Welcomeback template ==
 +
 +
Just for a note, with your current fancy signature (like I used to have), you can't just use <code><nowiki>{{subst:Welcome|~~~~|subst=subst:}}</nowiki></code>, you need to use <code><nowiki>{{subst:Welcome|[[User:AaronR|]] ~~~~~|subst=subst:}}</nowiki></code> instead. I used to use that too when I use my old signature. --[[User:Nat|<span style="color:#099;">Nat</span>]] [[User talk:Nat|<span style="color:#0a5;">Pavasant</span>]] 03:53, 3 October 2009 (UTC)
 +
 +
== Bad results ==
 +
 +
Hey - I see some bad results for Diamond from your rumble client: [http://darkcanuck.net/rumble/BattleDetails?game=roborumble&name=voidious.Diamond%201.45&vs=bigpete.Stewie%201.0], [http://darkcanuck.net/rumble/BattleDetails?game=roborumble&name=voidious.Diamond%201.45&vs=pulsar.PulsarNano%200.2.4], [http://darkcanuck.net/rumble/BattleDetails?game=roborumble&name=voidious.Diamond%201.45&vs=ms.Ares%200.19], [http://darkcanuck.net/rumble/BattleDetails?game=roborumble&name=voidious.Diamond%201.45&vs=rsk1.RSK1%204.0]. Could you take a look at your client config, or if Diamond really is erroring out on your system, could you tell me what exception you're seeing? Thanks dude. --[[User:Voidious|Voidious]] 01:35, 12 October 2009 (UTC)
 +
 +
: Well, I have two clients running on different computers, and both stopped running Diamond battles a while ago and are just doing Horizon 1.1.1x battles now. In other words, I don't have the output anymore. All the Horizon battles look like they're running fine, though. I'll add the bot to my exclude list; if you see any bad battles for any other bots coming from my system, tell me and I'll ask Darkcanuck to delete all the recent uploads from my client. « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 01:50, 12 October 2009 (UTC)
 +
 +
:: Grr... I added Diamond to the exclude list, but I didn't restart one of my clients and it uploaded a bunch more bad battles! I didn't see the battle output, either, only the "Uploading CrummyBot 0.1 vs. voidious.Diamond 1.45" lines. It's definitely specific to Diamond on that client, though. I'll look into it. « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 02:12, 12 October 2009 (UTC)
 +
 +
:: The exclude works for me as a solution. But if you do have a chance sometime to try and duplicate it, I'd definitely like to know of any problems with Diamond that I should fix. By the way, you may need to delete Diamond from the robots directory, too. I've noticed that my client will run excluded bots sometimes, but will never download them. --[[User:Voidious|Voidious]] 02:34, 12 October 2009 (UTC)
 +
 +
::: Looks like it's a case of a bad robot cache extraction. Just to be safe, I'm going to delete the Diamond jar download and rebuild the whole cache on that client. What should we do about the already-uploaded bad results? « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 03:02, 12 October 2009 (UTC)
 +
 +
::: Ah, that's a relief! You may need to delete it from roborumble/temp also (it's safe to delete everything in there). I'm not too concerned with the bad results - I was able to see that it was a slight improvement after all pairings completed, and I'm sure it won't be too long until my next version. =) Thanks for checking it out. --[[User:Voidious|Voidious]] 03:08, 12 October 2009 (UTC)
 +
 +
:::: Well, in one of my more dramatic examples of "completely underestimating the problem", it turns out that this was actually a file system issue! When I went to clear out the robot cache, I couldn't delete one of Diamond's directories. After Googling the extremely unusual error message, it appeared that the best solution was simply to reboot my computer (thank you very much, Microsoft). When I restarted, Windows ''started running CHKDSK on its own''! As I'm sure everyone here knows, there are few computer-related things as scary as a system running CHKDSK or your local equivalent. Anyway, it fixed the file system and it looks like everything's going to be fine now... and I can get back to running RoboRumble ''on my Linux machine''... « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 04:17, 12 October 2009 (UTC)
 +
 +
::::: Rebooting is one of the easiest solution. I found out that when system is unstable (and with my usage of the computer, it get unstable really fast), hard-turning-off (push the power button for ten seconds) is the best way to clean it, without the need of CHKDSK. --[[User:Nat|<span style="color:#099;">Nat</span>]] [[User talk:Nat|<span style="color:#0a5;">Pavasant</span>]] 04:35, 12 October 2009 (UTC)
 +
 +
:::::: That won't fix a broken file system though, which is what I had in this case. If anything, it could make it worse--especially with Windows (most Linux file systems are better at [http://en.wikipedia.org/wiki/ACID ACID]), hard resetting a system that happens to be in the middle of a file system operation could corrupt data. « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 04:50, 12 October 2009 (UTC)
 +
 +
::::::: Maybe the next version of the 'rumble superpack' should had a VM image edition, to be immune to operating system blunders... :) --[[User:Rednaxela|Rednaxela]] 05:17, 12 October 2009 (UTC)
 +
 +
::::::: It fixed mine. Rednaxela, which OS should be in the VM? Ubuntu Live CD? --[[User:Nat|<span style="color:#099;">Nat</span>]] [[User talk:Nat|<span style="color:#0a5;">Pavasant</span>]] 05:24, 12 October 2009 (UTC)
 +
 +
:::::::: I'm doubtful I would actually do it, but if I were to, certainly not be based on the ubuntu live CD. It'd be something far more minimalistic: command line only, bare minimum requirements to run java, plus java, plus robocode and bots. --[[User:Rednaxela|Rednaxela]] 05:32, 12 October 2009 (UTC)
 +
 +
::::::: I'd suggest Gentoo Extradimensional Edition. It can fix itself AND solve the halting problem! The only thing is, you have to compile everything from source... « [[User:AaronR|AaronR]] « [[User talk:AaronR|Talk]] « 05:34, 12 October 2009 (UTC)

Latest revision as of 06:34, 12 October 2009

Wow, does that page layout actually work in your browser without overlaps? That requires a 1100 px (Safari) or 1180 px (FireFox) wide browser for me, which I wouldn't be using no matter what resolution I'm in. :-P I think 800 px is a normal min width to design for. --Voidious 02:38, 12 November 2007 (UTC)

My resolution is 1440x900, so I can get my browser down to 4/5 of my screen width without any overlap whatsoever (plus, I always keep my browser window maximized and just use tabs). Still, I suppose it is kind of ridiculous... --AaronR 02:43, 12 November 2007 (UTC)

Thanx AaronR for pointing out what I should have done with those Oscillator pages in the first place. It is just that the 'old' wiki is the only one I know and I am not used to all the possibilities of the new one yet. -- GrubbmGait

You're welcome. =) --AaronR 01:42, 25 November 2007 (UTC)

I added a template page for userboxes, so you can use the syntax you're used to from Wikipedia. See it at Template:Userbox. I haven't added documentation to the template page yet, that would require another template. -- Synapse 04:00, 6 September 2008 (UTC)

Credits for the CacheCleaner for Robocode

I am currently integrating your CacheCleaner for cleaning the robots folder for RoboRumble into Robocode. I should like to add your real name instead of your username (AaronR) to the credits and source header. Do you want me to put your full name into the credits or "just" your username? If you want to use your full name, what is your name then? =) --Fnl 20:55, 3 December 2007 (UTC)

Well, since I contribute to the Wikipedia page on Robocode under my full name, I don't imagine it would take much trouble to figure it out, would it? =) --AaronR 00:50, 4 December 2007 (UTC)
Thank you Rotenberg! That was one of the places I hadn't looked yet. =) Thank you for all of your contributions, especially with this new Wiki and also the Robocode Wikipedia page =) --Fnl 14:28, 4 December 2007 (UTC)
Much thanks from me, too! Your help has been invaluable in getting the new wiki going, both with your contributions and your knowledge of MediaWiki / wiki practice. --Voidious 15:02, 4 December 2007 (UTC)
Anytime. =) The new wiki is going to be "ready" soon, so let's keep at it! --AaronR 16:46, 4 December 2007 (UTC)

Nice to see you are back! Hope you didn't just visit-and-gone-again like ABC =) » Nat | Talk » 09:38, 15 August 2009 (UTC)

Wow, that was fast! =D « AaronR « Talk « 09:39, 15 August 2009 (UTC)
You are faster on this (but you have orange bar to tell you). But by the way, I stole your signature =) » Nat | Talk » 09:46, 15 August 2009 (UTC)
That's OK, your arrows point the other way! <insert yet another smiley here> « AaronR « Talk « 09:51, 15 August 2009 (UTC)
Welcome back! I've been thinking of you recently while I migrate content. =) Btw, you might be curious to know that Horizon barely missed the cut when selecting bots for the new Anti-Surfer Challenge. --Voidious 15:19, 15 August 2009 (UTC)
Thanks! The new version certainly wouldn't cut it for that challenge, it's way too slow. I think I'll be asking around for ideas on how to speed it up without losing battle performance. « AaronR « Talk « 19:57, 15 August 2009 (UTC)
I'm not sure if you already use this trick, but it can help a lot. You say you precise predict two waves, and I assume you sum the dangers for each one. Well, after you predict the first wave for a movement option, you could check if the danger from that alone is already greater than the lowest danger you've found so far in your other movement options; if it is, no need to predict the second wave at all. If you weight the waves by distance or bullet time, this will happen a lot. Going further, you can always first calculate the danger of the movement option that was safest the previous time. (Props to Krabb for originally showing me this.) --Voidious 20:11, 15 August 2009 (UTC)
Ah! That will probably help a lot. I'll implement that now and see how much faster it is. (By the way, currently I'm just weighting the waves by a fixed factor--i.e. the wave that will hit second counts half as much as the wave that will hit first. I haven't tweaked that constant at all since it was introduced.) « AaronR « Talk « 20:16, 15 August 2009 (UTC)
Well, that got 10 rounds of Horizon 1.1 ditto down from 65 seconds to 43 seconds. It's still way slower than I'd like though. Making it surf only one wave brings it down to 32, which is still slow, so I get the feeling that I need to optimize elsewhere, too. « AaronR « Talk « 21:34, 15 August 2009 (UTC)
For comparison, the previous version, 1.03, does a 10 round mirror match in 10-12 seconds. The only thing I changed that would make any speed difference is the surfing danger function. « AaronR « Talk « 21:41, 15 August 2009 (UTC)
The other big speed optimization for DC surfing is to cache the nearest neighbors (aka scan cluster, aka similar situations). Do you do that already? You certainly don't need to recalculate them each tick. I calculate and cache the neighbors for each wave if they are not cached, then clear the caches when I'm hit, I detect an onBulletHitBullet, or the first wave I'm surfing changes. --Voidious 21:45, 15 August 2009 (UTC)
I'm sure that would speed things up... but that's not the problem. Like I said, the only difference between 1.03 and 1.1 is the WS danger function. By the time I start precise prediction, I've already calculated a GF array and I'm just passing that around. Thanks for the suggestion, though--I'll implement that, too. « AaronR « Talk « 21:58, 15 August 2009 (UTC)

Leave of absence

While everyone's talking about resurrected Robocoders, don't forget that I'm still around! I just haven't had any time lately because of school. I have this great[citation needed] idea for a new[citation needed] strategy for melee movement, but I refuse to work on any new bots until I figure out what I broke in the last version of Horizon! And since I can't find anything obvious in the diffs, you probably won't hear much from me for a while... « AaronR « Talk « 18:24, 1 October 2009 (UTC)

I sure haven't forgotten... Horizon 1.1.1b is currently one of Dookious' 5 losses. :-P --Voidious 18:34, 1 October 2009 (UTC)

Look at the battle details. Horizon got creamed, but Dooki crashed on one of the battles and the server recorded the zero score. I know, I'm disappointed too. « AaronR « Talk « 18:55, 1 October 2009 (UTC)
Doh! Sorry if I got your hopes up. =) Thanks for lookin' out. --Voidious 23:50, 1 October 2009 (UTC)

I think you are a part of resurrection too. Btw, we don't have Cite extension installed on this wiki =P --Nat Pavasant 02:45, 2 October 2009 (UTC)

Welcome and Welcomeback template

Just for a note, with your current fancy signature (like I used to have), you can't just use {{subst:Welcome|~~~~|subst=subst:}}, you need to use {{subst:Welcome|[[User:AaronR|]] ~~~~~|subst=subst:}} instead. I used to use that too when I use my old signature. --Nat Pavasant 03:53, 3 October 2009 (UTC)

Bad results

Hey - I see some bad results for Diamond from your rumble client: [1], [2], [3], [4]. Could you take a look at your client config, or if Diamond really is erroring out on your system, could you tell me what exception you're seeing? Thanks dude. --Voidious 01:35, 12 October 2009 (UTC)

Well, I have two clients running on different computers, and both stopped running Diamond battles a while ago and are just doing Horizon 1.1.1x battles now. In other words, I don't have the output anymore. All the Horizon battles look like they're running fine, though. I'll add the bot to my exclude list; if you see any bad battles for any other bots coming from my system, tell me and I'll ask Darkcanuck to delete all the recent uploads from my client. « AaronR « Talk « 01:50, 12 October 2009 (UTC)
Grr... I added Diamond to the exclude list, but I didn't restart one of my clients and it uploaded a bunch more bad battles! I didn't see the battle output, either, only the "Uploading CrummyBot 0.1 vs. voidious.Diamond 1.45" lines. It's definitely specific to Diamond on that client, though. I'll look into it. « AaronR « Talk « 02:12, 12 October 2009 (UTC)
The exclude works for me as a solution. But if you do have a chance sometime to try and duplicate it, I'd definitely like to know of any problems with Diamond that I should fix. By the way, you may need to delete Diamond from the robots directory, too. I've noticed that my client will run excluded bots sometimes, but will never download them. --Voidious 02:34, 12 October 2009 (UTC)
Looks like it's a case of a bad robot cache extraction. Just to be safe, I'm going to delete the Diamond jar download and rebuild the whole cache on that client. What should we do about the already-uploaded bad results? « AaronR « Talk « 03:02, 12 October 2009 (UTC)
Ah, that's a relief! You may need to delete it from roborumble/temp also (it's safe to delete everything in there). I'm not too concerned with the bad results - I was able to see that it was a slight improvement after all pairings completed, and I'm sure it won't be too long until my next version. =) Thanks for checking it out. --Voidious 03:08, 12 October 2009 (UTC)
Well, in one of my more dramatic examples of "completely underestimating the problem", it turns out that this was actually a file system issue! When I went to clear out the robot cache, I couldn't delete one of Diamond's directories. After Googling the extremely unusual error message, it appeared that the best solution was simply to reboot my computer (thank you very much, Microsoft). When I restarted, Windows started running CHKDSK on its own! As I'm sure everyone here knows, there are few computer-related things as scary as a system running CHKDSK or your local equivalent. Anyway, it fixed the file system and it looks like everything's going to be fine now... and I can get back to running RoboRumble on my Linux machine... « AaronR « Talk « 04:17, 12 October 2009 (UTC)
Rebooting is one of the easiest solution. I found out that when system is unstable (and with my usage of the computer, it get unstable really fast), hard-turning-off (push the power button for ten seconds) is the best way to clean it, without the need of CHKDSK. --Nat Pavasant 04:35, 12 October 2009 (UTC)
That won't fix a broken file system though, which is what I had in this case. If anything, it could make it worse--especially with Windows (most Linux file systems are better at ACID), hard resetting a system that happens to be in the middle of a file system operation could corrupt data. « AaronR « Talk « 04:50, 12 October 2009 (UTC)
Maybe the next version of the 'rumble superpack' should had a VM image edition, to be immune to operating system blunders... :) --Rednaxela 05:17, 12 October 2009 (UTC)
It fixed mine. Rednaxela, which OS should be in the VM? Ubuntu Live CD? --Nat Pavasant 05:24, 12 October 2009 (UTC)
I'm doubtful I would actually do it, but if I were to, certainly not be based on the ubuntu live CD. It'd be something far more minimalistic: command line only, bare minimum requirements to run java, plus java, plus robocode and bots. --Rednaxela 05:32, 12 October 2009 (UTC)
I'd suggest Gentoo Extradimensional Edition. It can fix itself AND solve the halting problem! The only thing is, you have to compile everything from source... « AaronR « Talk « 05:34, 12 October 2009 (UTC)