Yesterday I got some compute errors on 2 of my machines. See http://milkyway.cs.rpi.edu/milkyway/results.php?userid=9, if they haven't been purged already <grmbl>. In all cases I got "process got signal 11".

This could indicate a hardware problem, if it weren't that it occurs on 2 different (physical) machines. Both clients are Ubuntu 8.04 64-bit running in VMWare Workstation 6.0.4-built 93057, hosted by 64-bit Vista.

Seems like a good idea to me. Most of the project scientists don't have the time to worry about credit rates. Having a client/server version that handles the adjustment for them automatically should help alot in the credit flame wars.

John,

Since this subject has been so hotly debated for so long, I doubt you would have recieved a reasonable response here no matter how you phrased your inital post, but it doesn't help that you made a completely outlandish claim concerning appropriate credit for MW tasks.

You stated, "The long tasks really should be granting about 50 CS per task."

If you really believe that, then you understand that you're stating that MilkyWay should grant approximately 30% the amount of credit per unit of time vs SETI. (I'm assuming that MW is currently granting at a rate of 1.64 v SETI which seems to be a good average of the cross project comparisons that I could find) A 260 cs (long) task here would therefore grant around 159 on your hallowed SETI. (But should only grant "about 50" here)

Seems either quite spiteful or quite ignorant to tell the system administrators here that they're being "anti-social" if they don't value their project at 1/3rd the value of SETI.

I wanted to make sure that this comparison made some sense, so I looked at the computer I'm working from now and discovered that an Einstein task that takes me around 9 hours grants me ~240 cs. A MilkyWay task that takes 8 hours grants me 260. So that 260 cs task here would grant me 213 at Einstein.

Just so I'm clear, let me confirm. In your way of thinking it's 3 times more valuable to look for signals from extraterrestrials than it is to map the structure of our galaxy and over 4 times more important to look for gravity waves than to map the structure of our galaxy, yes?

I already replied to this and admitted an error in calculation or input. Why are you bringing it up again?

John,

Since this subject has been so hotly debated for so long, I doubt you would have recieved a reasonable response here no matter how you phrased your inital post, but it doesn't help that you made a completely outlandish claim concerning appropriate credit for MW tasks.

You stated, "The long tasks really should be granting about 50 CS per task."

If you really believe that, then you understand that you're stating that MilkyWay should grant approximately 30% the amount of credit per unit of time vs SETI. (I'm assuming that MW is currently granting at a rate of 1.64 v SETI which seems to be a good average of the cross project comparisons that I could find) A 260 cs (long) task here would therefore grant around 159 on your hallowed SETI. (But should only grant "about 50" here)

Seems either quite spiteful or quite ignorant to tell the system administrators here that they're being "anti-social" if they don't value their project at 1/3rd the value of SETI.

I wanted to make sure that this comparison made some sense, so I looked at the computer I'm working from now and discovered that an Einstein task that takes me around 9 hours grants me ~240 cs. A MilkyWay task that takes 8 hours grants me 260. So that 260 cs task here would grant me 213 at Einstein.

Just so I'm clear, let me confirm. In your way of thinking it's 3 times more valuable to look for signals from extraterrestrials than it is to map the structure of our galaxy and over 4 times more important to look for gravity waves than to map the structure of our galaxy, yes?

This is getting seriously funny John

The fact that projects grant credit at seriously different rates is a problem in the perceived fairness of the system.

Seti grants 30% more credit to the same machine only based on whether it has an optimized app or not that the project refuses to deliver to all its users as 20 other projects do!...unfair! Solaris and osx my foot, these are mainstream hosts that get optimized.

By your own ommision,although you felt the need to edit your little factoid out but I quoted you! ;)

The stock applications at S@H are currently much better optimized than they were, and even the best optimized applications do not grant at 2 to one, but if I recall, only around 1.3 to one at the moment. Trust me on this. S@H and BOINC are being re-worked to make it much easier for projects to have optimized applicatons delivered to clients

30% more credit and you have the gall to intimidate other projects and participants.

Trust me on this.

I THINK NOT!

If Seti wants the rest of BOINC to change then it needs to change its own culture 1st to start leveling the playing field.

I am sorry that I really don't understand your insinuation. S@H grants the same credit for a single task to all that crunched that task. Some computers go through tasks more quickly. Some people have better computers. That still has no bearing at all on cross project parity.

Thanks, Voltron. You're right. I did take the bait. I shouldn't have given him a target. Obviously he hasn't considered our posts; certainly not mine, and certainly not the one you quoted above. Why anybody would feel so threatened because of a few differences in worthless credits, that clearly isn't impacting BOINC in any significant way, is a mystery.

So yes, he's either out just to stir up a fight, or he's in denial and cannot be reasoned with.

Thanks for the good advice.

Glad to know you've been continuing the optimization Crunch3r!

I must admit that I suspected there were some serious opportunities for it when I saw that a 6 year old Celeron 2Ghz was only taking about 40% longer to do a task than a Xeon 5120 on the stock app. ;)

So are you hopefully going to share with the whole class and provide these to the developers? :D

just take a look at this thread... i think the answer to your question becomes quite clear...

besides that the "developers" do know what can be done to improve the apps performance cuz i told them a few months ago...

The fact that projects grant credit at seriously different rates is a problem in the perceived fairness of the system.

The stock applications at S@H are currently much better optimized than they were, and even the best optimized applications do not grant at 2 to one, but if I recall, only around 1.3 to one at the moment. Trust me on this. S@H and BOINC are being re-worked to make it much easier for projects to have optimized applicatons delivered to clients

Trust me on this.

Trust me on this.

No, it is just reversing the side on a stupid argument that I am tired of getting. BOINC is SUPPOSED to work such that a particular host gets the same credit per hour no matter which project it is working on. People that say stop complaining about something that is not correct are in the wrong.

AHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH


Just stop already............You don't like the credits given here...

Go somewhere else!

Since this is a problem affecting the entirety of BOINC, NO.

Easy fix... go do folding@home only 1 project, no cross project problems there!

Leave the rest of us in peace.

PS: Take the sauger with you..

Since you don't seem to like the way the BOINC is supposed to work, I can say the same to you. Go someplace where there is not the problem and leave those of who LIKE the way that BOINC is SUPPOSED TO WORK alone.

I agree for once John....the original post needs to be deleted as non relevant.

It is really hard for the first thread to be irrelevant to the topic of the thread.

Hmm. Boycotted by a project with merely 5K members and only 3K active? Hardly much of a threat.

Must have you flustered John as I would expect a better response than that!

Are all participants of Cosmology only connected to 1 project? Do you think team members would not support unfair treatment? Force away......you will never achieve anything near submission going into project boards arguing this same point....The rally cry may sound at such dismissive behaviour! You think the usership here will listen with those kind of inflammatory comments?

We and Cosmology have now been dismissed by Boinc as hardly much of a threat....ok I am done ...I refuse to respond to this type of Geastapoism.

You never answered as to why SETI needs to have a private club of optimized...the crux of my arguement?

BOINC is not well designed to distribute optimized applications to different clients of the same basic CPU and OS. It was made barely possible when the CPU optimizations were reported back. Work is in progress for the BOINC client and server to make this relatively easy, and S@H is, I believe waiting for that. The anonymous client was implemented long before the detection code was implemented in the client to distinguish which FPU extensions were implemented. The anonymous client was implemented so that people could develop applications for hardware that the project could not devote the manpower to implement the application for (i.e. OS2). It has been used for this. Some users also figured out that the same mechanism could be used to implement optimized applications. S@H has implemented some of the optimizations that have been developed into the stock client. Others have not been incorporated. There was an enormous debate about credits granted to optimized apps, and another when S@H incorporated some optimization and cut the credit grant to keep the credits / hour on hosts with stock applications about the same.

Basically what has happened has been that better hardware gets more credit per hour (the optimized applications count as "better hardware"). However, when the stock application gets optimized that does not count as "better hardware" for everyone and the credit grant was reduced to match the increased performance. I expect that when (may be later as Astropulse is absorbing much of the programming time) optimized apps per FPU extension start being used the credit request will be reduced again to reflect the reduced FLOPS count and to keep the average credits per hour about the same.

BTW, did you know that about 1% of the time the optimized applications at S@H fail to generate the correct result and therefore get no credit granted? This is one reason that not all of the optimizations were incorporated into the stock client. Another is, of course, that the BOINC client was not reporting the FPU extensions and therefore the server could not send a specifically optimized client.

Hmm. Boycotted by a project with merely 5K members and only 3K active? Hardly much of a threat.

The stock applications at S@H are currently much better optimized than they were, and even the best optimized applications do not grant at 2 to one, but if I recall, only around 1.3 to one at the moment. Trust me on this. S@H and BOINC are being re-worked to make it much easier for projects to have optimized applicatons delivered to clients.

That still leaves some projects that are way out of line in granting credits.