Welcome to MilkyWay@home

I think you made your point !!

Message boards : Number crunching : I think you made your point !!
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · Next

AuthorMessage
Brian Silvers

Send message
Joined: 21 Aug 08
Posts: 625
Credit: 558,425
RAC: 0
Message 5536 - Posted: 17 Oct 2008, 22:41:42 UTC - in response to Message 5532.  


the 1st real indication from you that you disagree with the tactics that Milksop has taken.


Apparently you have not read my other posts to them. I have been critical of them, but yet I know that they are operating within their rights. Whether it is the best course of action to take given the ease at which people are riled up over things like this, that is a different subject, which is:

Just because you can do something, doesn't always mean you should.
ID: 5536 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 5537 - Posted: 17 Oct 2008, 22:48:43 UTC - in response to Message 5536.  

Just because you can do something, doesn't always mean you should.


However, given the circumstances it did draw their attention and things are happening, I think.

As regards to people getting riled because someone has, legitimately, modified the code to carry out crunching much much more efficiently. My reaction is to say - get over yourself, please!
ID: 5537 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Milksop at try

Send message
Joined: 1 Oct 08
Posts: 106
Credit: 24,162,445
RAC: 0
Message 5538 - Posted: 17 Oct 2008, 22:54:06 UTC - in response to Message 5536.  

Just because you can do something, doesn't always mean you should.

That is right and the very reason we are not doing it with our personal accounts and for our team.

But I guess I don't have to explain another time why we think someone just has to show the extent of the waste. You have read it often enough ;)
Just repeating an argument does not make it better.

PS:
The last sentence is a general remark and applies of course also to my own posts (more specifically this one), just to cause no unnecessary irritation.
ID: 5538 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Milksop at try

Send message
Joined: 1 Oct 08
Posts: 106
Credit: 24,162,445
RAC: 0
Message 5539 - Posted: 17 Oct 2008, 22:57:01 UTC - in response to Message 5537.  

However, given the circumstances it did draw their attention and things are happening, I think.

This is what we intended. Thanks for the support!

ID: 5539 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Brian Silvers

Send message
Joined: 21 Aug 08
Posts: 625
Credit: 558,425
RAC: 0
Message 5540 - Posted: 17 Oct 2008, 22:57:15 UTC - in response to Message 5535.  

What I have not done is to simply give the project the source code. I explained the reason for that to you already more than a week ago.


I know, but like I said, holding things back like that can make people question your motives. If you don't care, then that's one thing. If you do care, then perhaps you should consider softening your approach. I know that what you're doing wouldn't be technically against the letter of the GPL (if the code was GPL'd), but it would be against the spirit of it as well as against the spirit of OSS/OSI.

I said already several times we are not hiding something.


...and as I just mentioned, by not contributing your improvements back to the community, without conditions, you do violate the spirit of the GPL, if there was a GPL, even if you aren't breaking the letter of it (since you aren't distributing).

I know you know what I'm talking about, and you should know that I know what I'm talking about too... What you're doing is very borderline behavior, and this is likely why Crunch3r warned you (that this sort of thing would happen).
ID: 5540 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Milksop at try

Send message
Joined: 1 Oct 08
Posts: 106
Credit: 24,162,445
RAC: 0
Message 5541 - Posted: 17 Oct 2008, 23:54:10 UTC - in response to Message 5540.  
Last modified: 18 Oct 2008, 0:07:54 UTC

The code does not have a license at all. You can do whatever you want with it. The Boinc part (coming from Berkeley) is under the LGPL license. That means you are allowed to create closed source projects with it (as done at several other projects).

I said already several times we are not hiding something.

...and as I just mentioned, by not contributing your improvements back to the community, without conditions, you do violate the spirit of the GPL, if there was a GPL, even if you aren't breaking the letter of it (since you aren't distributing).

I know you know what I'm talking about, and you should know that I know what I'm talking about too...

Fine, we exchanged arguments and we both still stand on our positions. That happens.
But you know I gave a description of the changes leading to that substantial speed improvement. If someone asks you the way to the next gas station, do you get in the car and drives him to the station? Most probably you will tell him: "Just drive one mile in that direction, at the crossway there turn right, the next left and you are there".

What you're doing is very borderline behavior, and this is likely why Crunch3r warned you (that this sort of thing would happen).

He told us that people will react angrily, accuse us of cheating and resort to personal attacks (edit: that applies of course not to you). But we thought and still think that some signal has to be send. By not using our personal accounts we think we are on the right side of the border ;)
ID: 5541 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Al

Send message
Joined: 9 Aug 08
Posts: 10
Credit: 46,370
RAC: 0
Message 5543 - Posted: 18 Oct 2008, 1:16:40 UTC - in response to Message 5531.  
Last modified: 18 Oct 2008, 1:26:36 UTC

You just bit me with your app, Milksop. You did a WU in under 1.5 hours when it took me more than 11 hours for the same one... I don't expect I'll get credit for it. Looks like I should abort when it gets to the due date.

Al


I'm not trying to side with Milksop with what I'm about to say, but...

My AMD system is capable of going through a 260 credit task in around 18,000 - 19,000 seconds with the same application you're using. You reported the task 23,352 seconds past deadline, thus my AMD system could've caused the same situation to happen for you.

The larger issue for you is resource allocation and/or system uptime and/or cache size...


Well, DSL is coming soon.. That will let me send results as soon as they are ready, but it won't affect any processing speed. I have the same cache size as you, but your speeds are more than double mine. I have no idea why that is and probably can't do anything about it. Is this a AMD vs Intel comparison?
ID: 5543 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Misfit
Avatar

Send message
Joined: 27 Aug 07
Posts: 915
Credit: 1,503,319
RAC: 0
Message 5544 - Posted: 18 Oct 2008, 2:42:30 UTC

ID: 5544 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Brian Silvers

Send message
Joined: 21 Aug 08
Posts: 625
Credit: 558,425
RAC: 0
Message 5548 - Posted: 18 Oct 2008, 20:02:31 UTC - in response to Message 5543.  


Well, DSL is coming soon.. That will let me send results as soon as they are ready, but it won't affect any processing speed.


So I take it you're using dialup that may not be available at all times? You're right, that won't change processing speed, but it would change how quickly you could report a completed task.


I have the same cache size as you, but your speeds are more than double mine. I have no idea why that is and probably can't do anything about it. Is this a AMD vs Intel comparison?


I thought about clarifying what I was talking about when I mentioned "cache" because of the potential confusion, but I didn't... What I'm talking about is the workunit cache, in number of days, that you have set in your BOINC preferences, not the amount of cache memory that your processor has. My thought is that if you are requesting 3 days, you might consider dropping back to 2 days. If you are already at 2 days, then you might consider dropping to 1 and using the "Maintain enough for an additional" number of days...to perhaps 0.5 days.
ID: 5548 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Al

Send message
Joined: 9 Aug 08
Posts: 10
Credit: 46,370
RAC: 0
Message 5549 - Posted: 18 Oct 2008, 21:11:31 UTC - in response to Message 5548.  
Last modified: 18 Oct 2008, 21:13:20 UTC


Well, DSL is coming soon.. That will let me send results as soon as they are ready, but it won't affect any processing speed.


So I take it you're using dialup that may not be available at all times? You're right, that won't change processing speed, but it would change how quickly you could report a completed task.


It's on the house phone and I'm not always here either. I think I would have been late even if I has sent it after finishing - never quite sure if I need to account for another 4 hours... But Boinc never gives me the right times. It reported 10:57 for my last task when the log shows it's been running for 17:13. But the trickiest part is how many 260 credit tasks I get - that seems to be increasing lately.


I have the same cache size as you, but your speeds are more than double mine. I have no idea why that is and probably can't do anything about it. Is this a AMD vs Intel comparison?


I thought about clarifying what I was talking about when I mentioned "cache" because of the potential confusion, but I didn't... What I'm talking about is the workunit cache, in number of days, that you have set in your BOINC preferences, not the amount of cache memory that your processor has. My thought is that if you are requesting 3 days, you might consider dropping back to 2 days. If you are already at 2 days, then you might consider dropping to 1 and using the "Maintain enough for an additional" number of days...to perhaps 0.5 days.


I toyed with that yesterday. I had 2 days set in Boinc, but 3 at Milkyway, which must have overridden it. There seem to be fewer tasks now.

So how are you processing so fast? Are you overclocked or something?
ID: 5549 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Al

Send message
Joined: 9 Aug 08
Posts: 10
Credit: 46,370
RAC: 0
Message 5551 - Posted: 19 Oct 2008, 3:24:53 UTC - in response to Message 5544.  

Jeez, looks like the big dawg scared everyone off!
ID: 5551 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Brian Silvers

Send message
Joined: 21 Aug 08
Posts: 625
Credit: 558,425
RAC: 0
Message 5552 - Posted: 19 Oct 2008, 3:50:49 UTC - in response to Message 5549.  

But Boinc never gives me the right times. It reported 10:57 for my last task when the log shows it's been running for 17:13.


You mean the estimated time to completion, or the CPU time?


So how are you processing so fast? Are you overclocked or something?


My system is overclocked from 2.2 GHz (normal) to 2.75 GHz, but an Athlon64 is a better processor than a Celeron... Also, I suspect the cache amount listed for your Celeron is wrong and should probably be 256K... The cache amount reporting in BOINC is a bit buggy... Doesn't affect performance, just not detected correctly all the time.
ID: 5552 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Misfit
Avatar

Send message
Joined: 27 Aug 07
Posts: 915
Credit: 1,503,319
RAC: 0
Message 5553 - Posted: 19 Oct 2008, 5:37:29 UTC - in response to Message 5551.  

Jeez, looks like the big dawg scared everyone off!

Me too!
me@rescam.org
ID: 5553 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile zed

Send message
Joined: 17 Oct 07
Posts: 23
Credit: 26,399,335
RAC: 0
Message 5555 - Posted: 19 Oct 2008, 9:36:48 UTC - in response to Message 5553.  

Not scared off. Just pointless in continuing a conversation with fundamentalist. And besides the current topic has nothing to do with the thread.
ID: 5555 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Al

Send message
Joined: 9 Aug 08
Posts: 10
Credit: 46,370
RAC: 0
Message 5556 - Posted: 19 Oct 2008, 15:26:38 UTC - in response to Message 5552.  
Last modified: 19 Oct 2008, 15:59:19 UTC

But Boinc never gives me the right times. It reported 10:57 for my last task when the log shows it's been running for 17:13.


You mean the estimated time to completion, or the CPU time?


Estimated doesn't make any sense at all! It might say 8 hours done with 8 hours to go. Never could tell - not on this project anyway. But the log says it start at 10:19pm and finished 17 hours later the next day (log gone now). It even looks like Boinc's clock runs slow (or not at all sometimes).

Just blew another task. :( Should have reported it last night during SNL, but didn't. Some 'Anonymous' came in and did it in 34 minutes! But got awarded less credits than I had claimed. Interesting.
ID: 5556 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Al

Send message
Joined: 9 Aug 08
Posts: 10
Credit: 46,370
RAC: 0
Message 5557 - Posted: 19 Oct 2008, 15:40:55 UTC - in response to Message 5555.  

Not scared off. Just pointless in continuing a conversation with fundamentalist. And besides the current topic has nothing to do with the thread.


But it's such a hostile thread, Zed. It made for interesting reading (some of it). It needed that dog!

I agree with your position, but it's not worth raising your blood pressure. I wish I could crunch a task like that, but I can't - oh, well. Maybe next one...
ID: 5557 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Brian Silvers

Send message
Joined: 21 Aug 08
Posts: 625
Credit: 558,425
RAC: 0
Message 5558 - Posted: 19 Oct 2008, 16:07:56 UTC - in response to Message 5556.  
Last modified: 19 Oct 2008, 16:32:18 UTC


Just blew another task. :( Should have reported it lastnight during SNL, but didn't. Some 'Anonymous' came in and did it in an hour! My bad


No, that host didn't do it in "an hour", it did it in 258.77 seconds, which is 4 minutes, 18.77 seconds. Since it was a gs_373 (medium length), most likely it is one of the people with a fully optimized application. My system does those in about 10,000 seconds, and even if a Xeon 5460 was 10 times faster, and I doubt that it is, that would still be 1,000 seconds, so it would be someone with an optimized (or de-unoptimized) application. Whether that is someone with the project or one of the people that folks will like to say are "cheating", is anyone's guess... On the positive side (for the rest of us who are credit conscious), they are only getting about 4X the cr/hr of my system. Mine will get around 50/hr, and they got 216/hr for that task. My (revised) guess is a 5460 is about 4-6 times better than my system, so in a sense, it is "fair" (for the rest of us)...even though I disagree with the cap...
ID: 5558 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Al

Send message
Joined: 9 Aug 08
Posts: 10
Credit: 46,370
RAC: 0
Message 5559 - Posted: 19 Oct 2008, 17:15:09 UTC - in response to Message 5558.  
Last modified: 19 Oct 2008, 17:17:13 UTC


Just blew another task. :( Should have reported it lastnight during SNL, but didn't. Some 'Anonymous' came in and did it in an hour! My bad


No, that host didn't do it in "an hour", it did it in 258.77 seconds, which is 4 minutes, 18.77 seconds. Since it was a gs_373 (medium length), most likely it is one of the people with a fully optimized application. My system does those in about 10,000 seconds, and even if a Xeon 5460 was 10 times faster, and I doubt that it is, that would still be 1,000 seconds, so it would be someone with an optimized (or de-unoptimized) application. Whether that is someone with the project or one of the people that folks will like to say are "cheating", is anyone's guess... On the positive side (for the rest of us who are credit conscious), they are only getting about 4X the cr/hr of my system. Mine will get around 50/hr, and they got 216/hr for that task. My (revised) guess is a 5460 is about 4-6 times better than my system, so in a sense, it is "fair" (for the rest of us)...even though I disagree with the cap...


Rats! It was a 373 too - probably why I didn't worry about it. I can do those in time. In "real-time" it took me 29471 seconds to do that one! Oh, well. DSL comes Tuesday, if the phone co. is true to their word and I can figure it out.
ID: 5559 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Milksop at try

Send message
Joined: 1 Oct 08
Posts: 106
Credit: 24,162,445
RAC: 0
Message 5560 - Posted: 19 Oct 2008, 22:18:04 UTC - in response to Message 5558.  

No, that host didn't do it in "an hour", it did it in 258.77 seconds, which is 4 minutes, 18.77 seconds. Since it was a gs_373 (medium length), most likely it is one of the people with a fully optimized application. My system does those in about 10,000 seconds, and even if a Xeon 5460 was 10 times faster, and I doubt that it is, that would still be 1,000 seconds, so it would be someone with an optimized (or de-unoptimized) application.

Actually it isn't a very fast version. Crunch3r can do a long WU in less than 120s (two minutes), a medium one takes just slightly more than a minute. And that on a 3GHz 65nm Core 2. A 3.16GHz 45nm Core2 would be a bit faster.

Nevertheless, if I would have to bet, I would still think it is Crunch3r. That he does not use the fastest version has a very simple explanation. As weird as it sounds, on a sufficiently fast system (anything from AMD or intel with more than 1.6GHz, maybe besides an Atom ;) you get less credits per day if you use the fastest version. That behaviour of the cap is really a punishment. The faster you crunch, the less credits you get. And remember, I'm not speaking about the credit/WU level (that goes down too of course) but the credit/day number.
ID: 5560 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Al

Send message
Joined: 9 Aug 08
Posts: 10
Credit: 46,370
RAC: 0
Message 5561 - Posted: 19 Oct 2008, 23:37:45 UTC - in response to Message 5560.  
Last modified: 19 Oct 2008, 23:39:23 UTC


Actually it isn't a very fast version. Crunch3r can do a long WU in less than 120s (two minutes), a medium one takes just slightly more than a minute. And that on a 3GHz 65nm Core 2. A 3.16GHz 45nm Core2 would be a bit faster.


That is just too fast! I'll bet the developers should be quite leary of releasing something like that (though it needs to be done). It might totally change their way of doing things. It could work better - or it could break it.
ID: 5561 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 · Next

Message boards : Number crunching : I think you made your point !!

©2024 Astroinformatics Group