User talk:PEZ

From Robowiki
Revision as of 09:39, 11 November 2011 by Rednaxela (talk | contribs) (→‎Memory Issue with CassiusClay 2rho.02ml?: lazy array initialization)
Jump to navigation Jump to search

Facebook told me that this is your birthday. Happy Birthday! --Nat Pavasant 00:23, 1 January 2010 (UTC)

Ah, saw this now. Thanks! --PEZ 15:16, 5 April 2010 (UTC)

Spring cleaning

I love the battlefield sketches. =) Any idea what year that was? Seems like those should be in a Robocode Museum... --Voidious 16:03, 5 April 2010 (UTC)

Interesting stuff. I've had some similar experiences, finding various sketches when cleaning. Also glad to see you haven't forgotten about Robocode entirely ;) --Rednaxela 16:13, 5 April 2010 (UTC)

Well, all my sketches went to bin immediately after I converted them to text files. If I clean now, I probably find my kindergarten works =) [well, there is some in my all-purpose notebook(s)] It seems that all ideas in that is the basic off melee now. --Nat Pavasant 16:25, 5 April 2010 (UTC)

How long ago for those Melee doodles? I found them on the backside of a shopping list containing x-mas food and such. I seem to recall trying to keep my daughter from wrecking the store. And that I was just beginning to learn Robocode ... It might be eight years. I found lots of more sketches. Might decorate my page with some of them in the future. Funny how, after all these years, many of the drawings still make sense to me. --PEZ 07:08, 6 April 2010 (UTC)

If it was when you are just beginning to learn Robocode, it is exactly eight year =) [quote: January 6 2003 - I've been Robocoding for a year now.] --Nat Pavasant 10:24, 6 April 2010 (UTC)

Good find! =) I think I started Robocoding in late 2002 (autumn). That would make it a year after Robocode was released? Funny, I remember feeling that I was a real latecomer and thinking about those before me as "The Ancients". --PEZ 13:38, 6 April 2010 (UTC)

Well, if I have to divide the Robocode timeline, I would do it like:

  • Sep. 2000 - Feb. 2001 : Stone Age (Prehistory) first developing
  • Feb. 2001 - Jul. 11, 2002 : Bronze Age (Prehistory) first release until 1.06
  • Jul. 11, 2002 - Jan. 20, 2003 : Iron Age (Prehistory) 1.06 until Wiki started
  • Jan. 20, 2003 - Feb. 28, 2005 : Antiquity (History) until 1.07
  • Feb. 28, 2005 - Jun. 14, 2006 : Dark Age where there is no maintainer =(
  • Jun. 14, 2006 - ? : Renaissance
  •  ? - ? : Modern
  •  ? - present : Contemporary where there is a lot robocoder reunite and significant invention (for example, Diamond, Porria, Glacier, etc.)

Sadly, the Robocode's history around '06-'09 are not well recorded, so I'm not sure. But I believe that there is a time when there are only small robocoders who are still active on the wiki around the beginning of '09, and that is about the time when the old wiki was becoming obsolete, yet all links point there which confused me at that time. --Nat Pavasant 16:40, 6 April 2010 (UTC)

I'm sure everyone sees it their own way. For instance, your "Dark Age" was a very exciting time for me. And tons of stuff happened in 2006-2009, even if it's not well summarized on the wiki. For me, I kind of remember it like this:

  • Nov. 2002 - Jan. 2003 : Confused Age - Robocode existed, but it was very confusing
  • Jan. 2003 - Oct. 2005 : Non-existent Age - Robocode ceased to exist
  • Oct. 2005 - Present : The Voidious Era!

=) --Voidious 17:23, 6 April 2010 (UTC)

LOL! I'm just trying to match Robocode's history with world history =) --Nat Pavasant 06:39, 7 April 2010 (UTC)

Re: "Escape angle is everything?"

To reply to your picture upload comment: Nah, gun accuracy also matters. Plus, if one is optimizing for score rather than just survival, then factors like energy levels, enemy hitrate, and expected enemy firepower matter significantly. In this area, I think Midboss goes further than anything before it, doing a detailed statistical simulation of how the battle could play out at different firepowers, considering the various score bonuses etc. ;) --Rednaxela 14:39, 12 April 2010 (UTC)

It was Escape Angle in the context of Bullet Power selection. =) Really cool that there are rocket surgery going on in the bullet power area! Still, have you tried releasing a version of Midboss which always fires power 2? --PEZ 04:35, 13 April 2010 (UTC)

Well, early versions of Midboss used RougeDC's bullet power selection, which was also somewhat adaptive to hitrate but not in nearly as depth. I had one point measured RougeDC's bullet power selection to be improved over 'traditional' always 2, and I'm pretty sure Midboss's bullet power selection also improved from the old versions that were like RougeDC. You're right though, it would be good to do a fresh comparison. Perhaps what I should do... is 1) re-release the current version of Midboss (because it got some bad results in there), and 2) Release a version that copies Diamond's bullet power selection, for a known-good-reference. I'm pretty sure that Midboss makes some interesting gains in edge cases when enemy hitrate is extremely high or low, or when energy is getting low, but would be nice to give a more clear demonstration of that. Funny that you refer to it as "rocket surgery", because really, the detailed statistical simulation code is really tricky to debug. In part because it contains both a 'simulation' component that simulates the ways future hits/misses could play out up to a certain depth, then an 'estimator' part that is active when the depth limit is reached, and checking both to make sure both behave 1) correctly, and 2) approximately the same, can get tricky. --Rednaxela 05:00, 13 April 2010 (UTC)

Sounds fun though! --PEZ 05:40, 13 April 2010 (UTC)

Time for CassiusClay 2sigma! =) --Voidious 14:27, 13 April 2010 (UTC)

Discussion weirdness

I don't get the discussion UI. Is it buggy? I can't expand the thread. To see it I have to log out. --PEZ 14:42, 2 November 2011 (UTC)


The discussions just need to be styled to look more like what we are used to seeing. — Chase-san 15:23, 2 November 2011 (UTC)

div.lqt-thread-replies, div.lqt_thread div.lqt-thread-replies div.lqt-post-wrapper {
    background-image: none !important;
}

div.lqt-post-wrapper {
    padding-left: 0em !important;
    padding-top: 0em !important;
    padding-bottom: 0em !important;
}

div.lqt-post-sep, div.lqt-replies-finish-corner, .lqt-command-link, .lqt-add-reply {
    display: none !important;
}

.lqt-thread-signature {
    display: list-item !important;
    margin-left: 0em !important;
    padding-left: 0em !important;
    list-style-type: disc;
    list-style-image: url('http://file.csdgn.org/dashdash2.gif');
    list-style-position: inside;
}

.lqt-show-replies {
    margin-left: 0 !important;
}

.lqt-thread-user-signature {
    margin-left: -5px !important;
}

.lqt-thread-toolbar-timestamp {
    color: #000 !important;
}

div.lqt-thread-replies {
    padding-left: 2em !important;
}

Here is my CSS to make the liquid threads look more like what we are used to (in browser edit format), use the more menu to reply to a message. I removed many otherwise busy (to me) elements.

Chase-san 16:04, 2 November 2011 (UTC)

I placed the css on my user page for it and it sure makes the threads look prettier and less cluttered. However, it still doesn't allow me to see the replies for any thread once I have read them. (Clicking "Show replies" gives me a spinner and then nothing changes.) Maybe it's a webkit thing? Because it doesn't work with Safari or Chrome, nor from my iPhone or iPad. Haven't tried other browsers yet. It's really frustrating. Remember that I was the one that favored the conversations based knowledge repo over at the old wiki and had to be dragged into supporting the remake for a more encyclopedic version. (I can't apologize enough for being so stubborn about it, but there it is.) Anyway, if someone knows of a css tweak for fixing the threads I can be a ginny pig. Then we can put it in the site wide css so that others experiencing the same problems can benefit from it -- PEZ 19:13, 9 November 2011 (UTC)

Hmm, so I just saw that problem on this page on Linux/Chrome. It went away after I logged in, and keeps working when I log out/in again now. That's pretty sucky! --Voidious 19:20, 9 November 2011 (UTC)

I could live with it if it was just this page. But for me it's general for all threads all over the site. -- PEZ 19:22, 9 November 2011 (UTC)

Huh, that bug sounds very weird... I've never seen that bug on any browser I've used, regardless of if logged in or out. Hmm... anyone have objections to modifying things to get rid of the "Show replies" thing entirely so it always shows replies? Even without the bug you're seeing, I personally think that it would be best to always show all anyway. --Rednaxela 19:39, 9 November 2011 (UTC)

Sounds good to me. Good call! --Voidious 19:41, 9 November 2011 (UTC)
Yeah, that would be much more like regular page content and would fit my bill perfectly. Though I'm not sure it'll make the bug go away. -- PEZ 20:46, 9 November 2011 (UTC)

I just checked with Chrome's development console and it doesn't seem to be a browser issue. Clicking "Show X Replies" does not load the data into the page at all. -- PEZ 18:40, 10 November 2011 (UTC)

I just changed the settings so it won't show the "Show X Replies" links until depth exceeds 2000 or reply count exceeds 25000. I doubt those limits will be hit. That solve things for you? --Rednaxela 18:59, 10 November 2011 (UTC)

I still have that "Show X Replies" actually. -- PEZ 19:55, 10 November 2011 (UTC)

Er, on what page? I'm not seeing "Show X Replies" anywhere, including places I used to see it before. Perhaps you're seeing a cached version of a page? --Rednaxela 23:17, 10 November 2011 (UTC)

All over. =) On this page too for instance. "Show 9 Replies". And I do see your question about it so it shouldn't be a caching issue. (Not that I have checked how those threads are injected on the page.) This is one of the strangest bugs I have heard of, and I suspect you agree? I'm almost happy (shame on me) that Voidious also experienced it. -- PEZ 07:04, 11 November 2011 (UTC)

Er.. wow?!? Because I was 99% sure I essentially completely disabled the code that wrote the "Show 9 Replies" messages... and I'm no longer seeing them on any page when I did before so... uhhhh.... --Rednaxela 08:35, 11 November 2011 (UTC)

Wiki idea

I'd kill for a wiki extension that lets you like/upvote any piece of wiki content. I often see you guys write excellent stuff and there's just not a place to say it. It would open up for top lists of most upvoted content and stuff too. Great for content discovery. -- PEZ 20:49, 9 November 2011 (UTC)

Discuss here

Yes, I'm old school. Just through your discussions in this section instead of that strange threaded part of the page. =) --PEZ 14:42, 2 November 2011 (UTC)

Memory Issue with CassiusClay 2rho.02ml?

I've been getting "java.lang.OutOfMemoryError: Java heap space" in my Roborumble client when running some battles with CassiusClay 2rho.02ml. There might be a memory management issue or bug in that version of Cassius Clay. It crashes my client, which is one reason why it's taken almost an entire day now to run the battles for it. -- Skotty 03:48, 11 November 2011 (UTC)

Oh, my client also crashes with error message. I haven't put much significance to it since my client crashes very often (I have some faulty RAM capsule in my machine I think). Is there a way I can check how much memory my bot consumes? I'm now wondering if 02ml exposes a memory leak in CC or if it just has too many buffers now. (02ml adds a third virtual gun with identical segmentation to the other two guns.) But the buffers are allocated statically and shouldn't really increase CC:s need for memory over the course of the battle... Are you others OK with me keeping 2rho.02ml in the rumbe until it has run its missing 100 or so pairings? The indication so far is that I gain 0.5% APS by adding that third gun and I would like to know if it's worthwhile keeping it. I should inform you that I'm not by my rr@h machine so I can't restart the client and participate in running those battles (as I suspect my machine has stopped running them now.) Maybe you people who have access to your clients can temporarily increase the heap size of your clients? I will of course pull 02ml if you people wish so. -- PEZ 06:57, 11 November 2011 (UTC)

  • just for your infomation, it grabs alot (around 400Mb) when the battle begins, and stays constant till the end of the 35 rounds. (using Windows Task Manager). I restart my client every 25 battles with 'java -Xmx512M ...' --GrubbmGait 08:12, 11 November 2011 (UTC)
Perhaps it would make sense to use an approach I did in my Midboss anti-aliased VCS gun. That is... lazily allocating VCS buffers as-needed. I find that in most battles, a majority of buffers in the segmentation may not get data in a given battle. It dropped my memory usage by an order of magnitude to only initialize the bin arrays when needed. --Rednaxela 08:39, 11 November 2011 (UTC)