Average Credit Decreasing?

Message boards : Number crunching : Average Credit Decreasing?
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 9 · 10 · 11 · 12 · 13 · 14 · 15 . . . 32 · Next

AuthorMessage
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786099 - Posted: 9 May 2016, 10:22:52 UTC - in response to Message 1786094.  
Last modified: 9 May 2016, 10:25:26 UTC

After several years of sleepless nights, much research with a lot of help from friends, I may have found a simple way to restore credit to what it should be (or close enough), without disabling CreditNew. Should I talk to Eric about this ?

I hope this isn't related to the post on Link Time Optimisation which has just popped up on boinc_dev? Optimise the hell out of Dhrystone, and pretend it's still a benchmark within the meaning of the definition?


No Way, 'Optimising' DhryStone OR Whetstone, for that matter, is exactly the way to make things worse, as I am 99.9% sure you already know.

[Edit:] I unsubbed from alpha and dev actually, as soon as I recognised the stale threading concepts were incompatible with the way Linux and common sense were following Windows....
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786099 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14672
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1786100 - Posted: 9 May 2016, 10:33:34 UTC - in response to Message 1786099.  

... as I am 99.9% sure you already know.

I've posted a critical, but hopefully diplomatic, reply along those lines.
ID: 1786100 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786101 - Posted: 9 May 2016, 10:37:21 UTC - in response to Message 1786100.  

... as I am 99.9% sure you already know.

I've posted a critical, but hopefully diplomatic, reply along those lines.


Cheers. I'm pretty convinced the designer has his math arse bout face, and may actually believe a global benchmark hike would increase credit. It will of course just increase the normalisation reference and so drop every other app/device/planclass
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786101 · Report as offensive
Sleepy
Volunteer tester
Avatar

Send message
Joined: 21 May 99
Posts: 219
Credit: 98,947,784
RAC: 28,360
Italy
Message 1786102 - Posted: 9 May 2016, 11:03:48 UTC - in response to Message 1786093.  

I may have found a simple way to restore credit to what it should be (or close enough), without disabling CreditNew. Should I talk to Eric about this ?

Great! :-)

Why not talking to him? If correctly offered, it is all for the good of the project.

There may be many more people that we imagine and do not even know about this place of discussion that are seeing their credit vanish, are not so motivated by the scientific goals of SETI (or may equally like another project), and walk silently away from the project.

Sleepy
ID: 1786102 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14672
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1786103 - Posted: 9 May 2016, 11:14:00 UTC - in response to Message 1786093.  

I should have replied to the substantive point as well.

Should I talk to Eric about this ?

Actually, I don't think Eric - on his beleaguered own - is the right person to address first. Instead, run it past an ad-hoc peer review committee of the people you know are equipped to comprehend the complexities of the question. Catch any wrinkles there before you go to full publication.

Then, isn't the right venue for this the Project Management Committee? I know, I know - it's been given a stupidly restrictive remit by those with a vested interest in maintaining as much of the status quo as they can, and it's done nothing visible, and I suspect nothing substantive, since it was created.

So let's subvert it. Use the mechanism they created to return control of BOINC to the scientific research community that needs it - and take it out of the hands of Berkeley politics. Can't make the situation any worse than it is at the moment.
ID: 1786103 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786104 - Posted: 9 May 2016, 11:16:49 UTC - in response to Message 1786103.  

Alright, so let me get a beer, and then let's have a chat about this narcissistic infrastructure I don't know who they are (PMC)
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786104 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13835
Credit: 208,696,464
RAC: 304
Australia
Message 1786105 - Posted: 9 May 2016, 11:24:29 UTC - in response to Message 1786093.  

After several years of sleepless nights, much research with a lot of help from friends, I may have found a simple way to restore credit to what it should be (or close enough), without disabling CreditNew. Should I talk to Eric about this ?


If it fixes the actual problem and isn't just a work around.
Over the years I've found that no matter how good things might look, fixing the symptoms often leads to more grief. Better to fix the problem.
Grant
Darwin NT
ID: 1786105 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786106 - Posted: 9 May 2016, 11:28:20 UTC - in response to Message 1786105.  

Oh no, this one is the root cause of the main scaling error on CreditNew. Can't fix the variation part (which nobody really comments on that much yet), that requires actual engineering.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786106 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14672
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1786107 - Posted: 9 May 2016, 11:31:00 UTC - in response to Message 1786104.  

Alright, so let me get a beer, and then let's have a chat about this narcissistic infrastructure I don't know who they are (PMC)

Give me an hour - I need to get some lunch and vitamin D too. Then I'm all yours.
ID: 1786107 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1786111 - Posted: 9 May 2016, 12:29:15 UTC - in response to Message 1786103.  

I know, I know - it's been given a stupidly restrictive remit by those with a vested interest in maintaining as much of the status quo as they can, and it's done nothing visible, and I suspect nothing substantive, since it was created.

They had a BBQ. Or at least a get-together with food.

As for who they are"
Bruce Allen, Einstein@Home.
Christian Beer, RNA World and Rechenkraft.net e.V.
Matt Blumberg, GridRepublic.
Eric Korpela, SETI@home.
Keith Upplinger, IBM World Community Grid.
Rom Walton, Cloud Nine.
Willy de Zutter, BOINCstats and BAM!
And David Anderson, Whipmaster.

So in essence, walking it by Eric isn't a bad idea, as long as you CC at least two or three others of the list.
ID: 1786111 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786112 - Posted: 9 May 2016, 12:44:52 UTC - in response to Message 1786111.  

haha, well falling asleep from my own beer+BQQ committee meeting.

In the meantime, To fix the coarse scaling error, what's the alternative to including an allowance for SIMD vectorisation within creditnew for stock applications ? (Given that it is complex and been avoided)

[Hint: I know the answer already. and so do you]
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786112 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1786116 - Posted: 9 May 2016, 12:52:01 UTC - in response to Message 1786112.  

To fix the coarse scaling error, what's the alternative to including an allowance for SIMD vectorisation within creditnew for stock applications ?

[Hint: I know the answer already. and so do you]

More beer and BBQ, pedantically spoken?
ID: 1786116 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786117 - Posted: 9 May 2016, 12:53:38 UTC - in response to Message 1786116.  

To fix the coarse scaling error, what's the alternative to including an allowance for SIMD vectorisation within creditnew for stock applications ?

[Hint: I know the answer already. and so do you]

More beer and BBQ, pedantically spoken?


No simpler and stupider than that, you're trying to be too clever again :P
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786117 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1786119 - Posted: 9 May 2016, 12:57:53 UTC - in response to Message 1786117.  

The it's probably got to do with the removal of something, but I can't decide if that is of part of the credit generating code, or David.
ID: 1786119 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786120 - Posted: 9 May 2016, 12:59:18 UTC - in response to Message 1786119.  
Last modified: 9 May 2016, 13:00:09 UTC

The it's probably got to do with the removal of something, but I can't decide if that is of part of the credit generating code, or David.


Neither, It's not David's fault he doesn;t understand SIMD

[Well OK, CLose with removal of something]
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786120 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14672
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1786122 - Posted: 9 May 2016, 13:01:07 UTC - in response to Message 1786111.  

As for who they are...

Yes, that's what the page I linked says. Actually, since 9 October 2015, Christian Beer has been on the development staff at Einstein - and immediately became a welcome and active contributor to both the project science and its message boards. Though I believe he has also kept in touch with his previous projects.

Keeping its own web presence updated is one of the visible things that the PMC could - but didn't - do.

The lower part of the page has - since version 11, at least - listed some 120 volunteer contributors to the BOINC code base. It is for those practical skills and past contributions that I suggested the PMC as a possible route in - not because of their titular position.
ID: 1786122 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14672
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1786124 - Posted: 9 May 2016, 13:02:46 UTC

OK, do we have a quorum for the big reveal?
ID: 1786124 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786125 - Posted: 9 May 2016, 13:09:43 UTC - in response to Message 1786124.  
Last modified: 9 May 2016, 13:10:36 UTC

OK, Well first there is the preferred (complex) option (IMO) which would be to load the flops with a suitable multiplier for SIMD (as used to be done)

Simpler option (pending any actual fix for creditNew to account for SIMD vectorisation), is to simply disable ALL use of SIMD in the stock CPU application(s), that includes the core source and FFTW library etc.

Then your reference host claims will reflect reality (+/- 37% variation steady state because of other problems)
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786125 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1786126 - Posted: 9 May 2016, 13:18:09 UTC - in response to Message 1786125.  

But doesn't that affect only CPU tasks? The removal of the SIMD extensions at least?
ID: 1786126 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1786127 - Posted: 9 May 2016, 13:21:33 UTC - in response to Message 1786126.  

But doesn't that affect only CPU tasks? The removal of the SIMD extensions at least?


No, Stock CPU applciation becomes normalisation reference, by underclaiming due to AVX.

Explanation follows
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1786127 · Report as offensive
Previous · 1 . . . 9 · 10 · 11 · 12 · 13 · 14 · 15 . . . 32 · Next

Message boards : Number crunching : Average Credit Decreasing?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.