blackdown linux wont work?

[safemode]

I'm testing out some different java runtimes and blackdown linux is hanging after startup once it grabs the circuit to work on.


here is some verbose output. -verbose flag.

[Loaded sun.nio.cs.StreamEncoder$ConverterSE from /usr/lib/mozilla/plugins/java2/lib/rt.jar]
[Loaded sun.net.www.protocol.http.AuthenticationInfo from /usr/lib/mozilla/plugins/java2/lib/rt.jar]
[Loaded sun.net.www.protocol.http.PathMap from /usr/lib/mozilla/plugins/java2/lib/rt.jar]
[Loaded java.util.AbstractList$ListItr from /usr/lib/mozilla/plugins/java2/lib/rt.jar]


no errors are thrown, it's not using any cpu. it just stops doing stuff.

[safemode]

My mistake. The server is not responding AGAIN. should have expected that. hrm. eventually we should be able to go a week without the server not responding.

[Bok]

Yeah,

I had it running under blackdown just fine on RH9, Suse 8.2 for AMD64 and Gentoo. It seemed slower under linux though for some reason :(

Bok

[safemode]

fastest linux jre i've found was 1.3.1 from sun. I'm testing blackdown's 1.4.2 jre with some of the non-standard options regrading heapsize and such. But i can't really test it when the server is down for one reason or another every other day.

[michaelgarvie]

Sorry java web server ran out of memory so new clients couldn't download code. Old clients kept running ok since evolution server didn't fall. Will start with -Xmx option this time.

[michaelgarvie]

I'd like to remind you of the HUGE speed improvement to be gained when using the -server option to enable the server VM.

[Bok]

oh I always use -server under linux and windows :)

Bok

[safemode]

gui's still aren't working here. has that null exception.

[safemode]

This doesn't make sense. I have my client set to connect every 5 minutes. It was on for the past few hours and when i stop it, and restart it, it starts all the way back on the generation it was on before the server was restarted, all of the work done in the last few hours seems to have been ignored. Is my client not actually connecting to the server or what?

[michaelgarvie]

Nullpointer when running GUI means you probably don't have your DISPLAY variable set up properly.

The server did receive all these genotypes you were working on and some of the probably migrated to other islands. However this doesn't mean that it logged them to hard disk. This happens every N generations where N is currently set to 200 I think.

[safemode]

oh ok, i was under the assumption that nothing was saved locally to harddisk and that it had to retrieve any info from the server of my current position and everything.

I'd have thought so too, but the DISPLAY variable is not the problem. On my win98 box it does the same thing. No gui.

On a different note somewhat, i've gotten my generation rate up from 1 generation every 3-4 minutes to 1 generation a minute. I'll need to do some more testing with and without the option but i'm using -XX:+AggressiveHeap and i'm currently about 3 times faster than my P4 1.7Ghz (using my tbird 1.13Ghz machine). More testing is needed of course.

[safemode]

further testing has shown that it's really impossible to benchmark the evolution client, certain generations are easier to test than others. The only way to benchmark the evolution client's algorithms is to test it againts the same genetic specimen. This is currently not possible.


The only way to really benchmark things would be for there to be a special benchmark client that works against a complex but fixed circuit genome.


The argument i mentioned in my previous post has made a significant improvement, but how much i can't say. Mostly the benchmarking client would be helpful in determining which java arguments work best for different machines that have different amounts of memory. This is pretty low priority though. Looks like soon we'll be creating a better than human circuit for the current project. 22 gates, will we be moving off of rail circuits after this one and hit a plain assumption A at 22 gates ?

[michaelgarvie]

Server writes to hard disk, not client.

Very interesting about aggressive heap thing. Will try to setup a server with a fixed size evaluation you can test with. If not you can test using the same island (eg login as safemode2 with the fast one and test this for 10 minutes then disconnect and login as safemode2 with the other one and test for 10 minutes). These should be similar.

No GUI on windows? wonder if this is happening to everyone..

These next 10 days will be a bit hectic. I'll be writing up a couple of papers for the next online testing conference! I'll put a few of the top contributing islands and teams in the acknowledgments of course! and will be refering the circuits by the names people gave them!

[safemode]

no gui in windows, same null pointer exception error.

[michaelgarvie]

fixed