Pages: 1 2 [3] 4 5 6 7 :: one page |
|
Author |
Thread Statistics | Show CCP posts - 26 post(s) |
Luke S
Zeta Corp.
|
Posted - 2010.08.23 18:05:00 -
[61]
Stupid question: Does module lag happen to all types of mods? Such as mining lasers? ---
|
DmitryEKT
Atlas Alliance
|
Posted - 2010.08.23 18:25:00 -
[62]
Originally by: Kesper North Can you explain why you are using Maelstroms as a laser platform?
because, even after the projectile buff, a maelstrom with tachyons does more dps than a maelstrom with 1400mm's.
also, he said he was shooting for hours, so he probably didnt want something ammo-dependant. and matari ships are cool. so there ya go. i'd have taken abaddons myself...
|
Helios Black
Wraith.Wing Wildly Inappropriate.
|
Posted - 2010.08.23 18:30:00 -
[63]
Gotta say that I love the amount of blogs dedicated to informing us what's going on with fixing lag. Nice to hear about the technical aspects of things and know there's actually people out there who care about the issue and are working on it.
Thanks CCP!
Originally by: Helicity Boson Maybe you should go play Uno with your mom so you can be "second winner" again.
|
Dubyou Bush
|
Posted - 2010.08.23 18:32:00 -
[64]
STFU ***** and fix lag quit talkin
|
Helios Black
Wraith.Wing Wildly Inappropriate.
|
Posted - 2010.08.23 18:36:00 -
[65]
Originally by: Dubyou Bush STFU ***** and fix lag quit talkin
**** off.
Originally by: Helicity Boson Maybe you should go play Uno with your mom so you can be "second winner" again.
|
Master Akira
Child Head Injury and Laceration Doctors
|
Posted - 2010.08.23 18:38:00 -
[66]
Originally by: Dubyou Bush STFU ***** and fix lag quit talkin
You are an awesome poaster
|
Hykke
Free Imperial Vikings
|
Posted - 2010.08.23 18:39:00 -
[67]
Great series of dev blogs recently, this one in particular impressed me.
Being a developer myself I can appreciate the efforts you guys are doing to hunt down those illusive bugs and performance issues causing lag.
My guess is that sometime in the far far past there was a bug caused by the effect system not agreeing that a module should be cycled. The developer who got this bug report, then looked over the code and decided that this bug was so rare that he would just write a quick workaround. After this workaround was implemented the bug disappeared, and the problem was "solved" ... only to reappear 1.5 million times later :-)
Experience has taught me that a quick workaround is often the wrong solution to a problem
Great to see that you are not trying to implement a quick fix, but are trying to get to the root of the issue.
|
Ki Rathos
Minmatar Urban Mining Corp Swords Of Athena
|
Posted - 2010.08.23 18:47:00 -
[68]
I too was curious if there was a way to offload the effects processing to another node. If it is sounds like that might do the trick.
Think it would be cool too if you guys could develop some kind of tool that allows the thin client to say run on my comp and take requests from your central server, maybe have it connect into sisi.
Would leave it on while I am working, you guys would then have a veritable army of PC's ready to put tons of automated pew pew ships onto a node. Maybe once you guys add in some diagnostic tools. A system like that would allow you to mass test whenever you needed from actual client machines, sort of the end to end way to check on things.
|
Vaerah Vahrokha
Minmatar Vahrokh Consulting
|
Posted - 2010.08.23 19:25:00 -
[69]
The developer facial expression... almost repaid for all of the months worth of effort spent in asking for lag fixes. - Auditing & consulting
When looking for investors, please read http://tinyurl.com/n5ys4h + http://tinyurl.com/lrg4oz
|
|
CCP Shadow
C C P C C P Alliance
|
Posted - 2010.08.23 19:43:00 -
[70]
Off-topic posts removed. Stay classy.
|
|
|
Grez
Empire Assault Corp Dead Terrorists
|
Posted - 2010.08.23 20:03:00 -
[71]
Originally by: CCP Shadow Off-topic posts removed. Stay classy.
*grabs a glass of whiskey*
Will do, Sparky!
Thanks for the awesome dev blog! ---
|
Trebor Daehdoow
|
Posted - 2010.08.23 20:12:00 -
[72]
Originally by: Vuk Lau I will have liberty to say that I almost pooped all over Socratesz when I heard the famous "EVE is now, from a technical standpoint, in a better state than it has ever been." so my bet is that was pure example of nonsense
NDA or NDA, this is TMI IMHO.
Originally by: CCP Warlock But let's just say, that there's a little samizdat activity in CCP at the moment to provide an appropriately inappropriate t-shirt in order to properly memorialise that particular entry into the great list of "They can't hit an elephant from ther...." remarks.
Remember to print extras for the CSM. I take an XL.
Seriously now, excellent blog, it gives a good view of how the debugging process really works. And kudos to Vuk Lau for bringing the video to the meeting.
Confessions of a Noob Starship Politician Spending Hours blogging the Minutes
|
Selnix
Gallente North Eastern Swat Pandemic Legion
|
Posted - 2010.08.23 20:23:00 -
[73]
Originally by: CCP Veritas In other words, I believe Dogma is doing stupid things, and I intend to beat the stupid out of it before considering giving it rocket boots.
Given the current state of Rockets, only a stupid Dogma would try to use such boots.
Thx for a nice blog tho.
|
Amy Garzan
Gallente The Warp Rats Intrepid Crossing
|
Posted - 2010.08.23 20:39:00 -
[74]
Yay! Keep up the good work and techie blogs. I love reading them. -------------------------------------------------- 101010 The Answer to Life, The Universe, and Everything |
Mashie Saldana
BFG Tech
|
Posted - 2010.08.23 20:52:00 -
[75]
Originally by: Kesper North Can you explain why you are using Maelstroms as a laser platform?
I think it was mentioned elsewhere that it was civilian lasers.
18 months |
Glyken Touchon
Gallente Independent Alchemists
|
Posted - 2010.08.23 21:13:00 -
[76]
Originally by: Mashie Saldana
Originally by: Kesper North Can you explain why you are using Maelstroms as a laser platform?
I think it was mentioned elsewhere that it was civilian lasers.
Probably because it has 8 turrets & they didn't want to risk hurting the titan
|
Rage Spear
Minmatar Republic Military School
|
Posted - 2010.08.23 21:51:00 -
[77]
Best blog for a long, long time.
Thanks for writing this and good luck with finding some way to let us cycle our guns AND do everything else instead of just one or the other.
|
wizard87
|
Posted - 2010.08.23 21:57:00 -
[78]
Good blog, great pictures.
You're a good communicator CCP Veritas - you're analogies seem to make sense for the less technically inclined.
Hope you get a good appraisal as you appear to be one of the good eggs at CCP.
|
Brigitte Helm
Minmatar Native Freshfood
|
Posted - 2010.08.23 22:34:00 -
[79]
Thank you for a well thought out dev blog that explains not only the problem but how you find what the true problem is. Plus great photo
One minor correction the actual quote is "They couldn't hit an elephant at this dist..." General John Sedgwick's final words Hug a Carebear, Kill a pirate, squish a Rat, and tickle a dev.
Make Eve fun.... |
Xianthar
Vanishing Point. The Initiative.
|
Posted - 2010.08.23 22:43:00 -
[80]
surprised your using cooperative scheduling given the very dynamic workload, i figure you'd used preemptive scheduling or tasklets with some sort of tasklet priority assignment.
|
|
|
CCP Atropos
|
Posted - 2010.08.23 22:43:00 -
[81]
Originally by: Hawk TT Edited by: Hawk TT on 23/08/2010 16:55:18 Great Blog! Nice screenshots
For those interested in the Python GIL (Global Interpetter Lock) drawbacks and in Python cooperative-multitasking: David Beazley's - Understanding Python GIL - PyCon 2010
This is David Beazley's presentation @ PyCon 2010...who the f**k is David Beazley?!?! - OK, Google him The presentation is fun, it's informative, it shows how Python behaves on multiple-CPU cores (BAAAAD!) etc. The presentaiton also covers the issues when threads release control prematurely And there is some hope - a new GIL is being developed, unfortunately for Python 3.2+, but still it could be ported to Stackless 2.7...More or less, it seems that GIL will survive for the time being and will not be "killed" any time soon...
Hopefully CCP is working on something on their own...GIL in Python = No Benefits form Multicore Processing
I watched this since I was unable to attend PyCon this year and I have to agree; a very informative and not too technical view of life with the GIL.
Software Engineer Core Engineering |
|
Zendoren
Aktaeon Industries
|
Posted - 2010.08.24 00:57:00 -
[82]
That presentation goes to show you that, more often then not, laziness is the cause of most problems in life.
Sort answer is, the python Dev's are lazy and do not want to make a proper interpreter for its code, and CCP does not want to get into the programing language inventing business!
Looks like CCP is stuck between a rock and a hard place on the threading front!
So, how is that code refactoring project coming? Was the 3rd time the charm?
|
Jinli mei
Test Alliance Please Ignore
|
Posted - 2010.08.24 02:31:00 -
[83]
Is CCP doing anything fancy and magical like replacing the GIL for Python with something less suck?
|
OmegaTwig
THORN Syndicate Controlled Chaos
|
Posted - 2010.08.24 02:47:00 -
[84]
Originally by: Luke S Stupid question: Does module lag happen to all types of mods? Such as mining lasers?
If you can get everyone in Jita into one belt with enough Veldspar so that everyone can be mining at once, with drones and everything then yes i believe this will happen with ANYTHING that cycles and produces an effect...
BTW: Love the Devblog, +1 for pics... Gonna have to check more regularly for them because they are coming out more frequently!
FakeEdit: Found the RSS Feed, YAY!
|
Blazde
4S Corporation Morsus Mihi
|
Posted - 2010.08.24 03:31:00 -
[85]
Originally by: CCP Veritas I've been busy profiling it under some basic scenarios and there's some low hanging fruit to be plucked in there. There's significant algorithmic optimizations to be made as well, and those follow you around regardless of what language you're in
This single line answers all my frustrations with the feedback we've had so far about lag because any mention of basic low-level code optimisations has been glaringly absent before now. Please, please crack on with this. I spend a lot of time optimising code and it's really not uncommon to get 10-20 even 100 times speedup in code that hasn't been agressively optimised before, even if it was written fairly well in the first place. I have something I call the '10x rule': if giving your code ten times more time to run gives some significant benefit (removes a bottleneck, gives results previously unreachable, puts it acceptably inside a performance limit, cures lag!, etc) then optimise the hell out of it before even considering any costly high-level redesigns. In Python (or any other high level language) this applies more than ever because the language makes it very easy for a coder to do very costly things without realising it, if they don't fully understand what's happening underneath the hood.
In this case it sounds like the Dogma code is *the* fleet-fight bottleneck (makes perfect sense, activating modules/applying effects in very large numbers is pretty much the definition of a fleet fight), and it hasn't been agressively optimised in the past - resisting temption to ask why - so this should yield great results. (Hell just send me the Dogma sources and I'll send em back with all no-brainer optimisations done )
Regarding the Dogma yielding bug, was this introduced around October 2008? Ie. the exact time lag seemed to get much better. This was when fleet fight lag changed very dramatically from: - Not being able to lock - Warping/basic movement commands taking forever to initiate. You could 'queue up' warp commands, and often spent entire battles warping to the place you wanted to be at an hour earlier - Modules taking forever to active
to: - Warping/moving being extremely responsive - Locking often being pretty good - Ghost wrecks - Modules activating but not cycling, and getting stuck
There were theories at the time that devs had purposely changed the priorities of various commands, and it was seen as a very good thing because being in the right place and being able to lock stuff but not kill it fast is much less frustrating than being in the wrong place all the time being killed by the single ceptor that was lucky enough to lock you.
Some datapoints from my logs: 2008 Jan to June An attempt is underway to activate the module - 116 occurences An attempt is underway to deactivate the module - 2 occurences 2009 Jan to June (I did play quite a bit during this period, lag was just that good) activate - 0 deactivate - 3 2010 Jan to June activate - 7 deactivate - 245 (mainly occurs when deactivating mods by clicking their mini-icon next to target)
Regarding stuck modules I still think the problem is clientside as well. The server may stall sending the expected message back but it is the client that stops you interacting with the mod when it's blinky-red. If you could interact with it you could send new deactivation requests (which do somehow work), as shown by the mini-icon workaround. It's broken in normal non-laggy gameplay too because you can't do stuff you want to with a deactivating module like: - Right click on it to get it's show-info (eg. check your optimal/falloff) - Check or change it's autorepeat/autoreload status - Pre-heat it
Good engineers don't just fix a problem once afterall so I think you should fix this clientside too.
Anyway good luck and thanks for the blog, I might just have a new favourite dev (favourite new dev at the very least). _
Northern Coalition - Best friends forever <3 |
Liol Wongster
|
Posted - 2010.08.24 03:31:00 -
[86]
Originally by: Ki Rathos .........
Think it would be cool too if you guys could develop some kind of tool that allows the thin client to say run on my comp and take requests from your central server, maybe have it connect into sisi.
Would leave it on while I am working, you guys would then have a veritable army of PC's ready to put tons of automated pew pew ships onto a node. Maybe once you guys add in some diagnostic tools. A system like that would allow you to mass test whenever you needed from actual client machines, sort of the end to end way to check on things.
+1 for this. I would happily let this run for the hours I'm not able to be online and active myself. The mass tests all happen at a time that is no good for me to log in with my chars anyway. Anyway I caould help, I'd do happily. |
Vaal Erit
Science and Trade Institute
|
Posted - 2010.08.24 03:34:00 -
[87]
It looks like the lag hunting team and the thin client are making good progress. From the sound of this dev blog we might be getting a decent improvement in lag situations with the next expansion. Go go CCP! - It's not "Play through a pre-set story, become stronger, do endgame". Gameplay is open ended, and you make your own story. Unless you're too afraid of 'pvp grief' to do anything relevant |
Thor Brawn
|
Posted - 2010.08.24 03:36:00 -
[88]
Good work!!! Keep going and this game will become hotter than a Sandy beach with Victoria Secret models...
Well almost
|
Whatever Dood
|
Posted - 2010.08.24 03:38:00 -
[89]
Edited by: Whatever Dood on 24/08/2010 03:41:42 Edited by: Whatever Dood on 24/08/2010 03:41:11
Originally by: Jinli mei Is CCP doing anything fancy and magical like replacing the GIL for Python with something less suck?
I'd guess they plan on doing something like running a separate instance of the Python interpreter on each thread, each with completely separate GIL's, and then manage their thread interactions themselves.
I swear my IQ dropped 20 points watching that presentation. I have never seen anything so totally ******ed in my life. "This is a problem that's been solved in operating systems" - yeah, like two-three DECADES ago. "Don't use this talk to justify not using threads" - maybe to justify not using Python?
I had to watch the presentation here, original link doesn't play all the way through for me:
http://www.mefeedia.com/watch/28851035
|
Blazde
4S Corporation Morsus Mihi
|
Posted - 2010.08.24 03:46:00 -
[90]
Originally by: Axemaster Actually, why not use a gpu to process this? They seem tailor-made for these kinds of massive parallel computing problems.
GPU computing is not like having dozens of multi-purpose CPUs at you disposal. It's more like having dozens of scientific calculators each with say 128 bytes of RAM and attached to a shared hard drive. You can load some pretty complex programs onto them but you need to spend time pushing buttons to give them data at the start and to read it back from their screens at the end. They can't access your EVE database, or do any other useful I/O. Unless you can find enough highly parallel math based work for them to do then you spend all you time interacting with the calculators and never get any useful work done yourself. You in this case being a high performance general purpose computer yourself so you're pretty damn good at doing the same things you're spending time telling the calculators to do. Infact you're much, much better, there's just a lot of them.
It's surprisingly difficult to even come up with problems that benefit from GPU computing. _
Northern Coalition - Best friends forever <3 |
|
|
|
|
Pages: 1 2 [3] 4 5 6 7 :: one page |
First page | Previous page | Next page | Last page |