O1OD1 App version 0.03

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250536279
RAC: 34232
Topic 218332

We just issued a new app version for the "Gravitational Wave All-sky search on LIGO O1 Open Data" (einstein_O1OD1). This shouldn't change anything performance-wise. However it incorporates two new features that will become important in the future: 1. The computation is slightly changed. The results shouldn't differ enough to prevent validation, but it should be more compatible with the upcoming GPU version of the code. 2. The App can now report back additional timing information which will help us to develop better models for timing prediction ("flops estimation" and credit) in the future.

BM

Zalster
Zalster
Joined: 26 Nov 13
Posts: 3117
Credit: 4050672230
RAC: 0

Interesting, they get to

Interesting, they get to about 4 minutes then restart their progress...Going to need to keep an eye on them....

Bill
Bill
Joined: 2 Jun 17
Posts: 38
Credit: 328674797
RAC: 217545

Bernd, I assume you are aware

Bernd, I assume you are aware of the peak flops bug that Richard Haselgrove had brought up on a recent conference call.  I haven't experienced this problem with Einstein so hopefully it continues to be that way.

Zalster
Zalster
Joined: 26 Nov 13
Posts: 3117
Credit: 4050672230
RAC: 0

So this was an interesting

So this was an interesting bit in the result. I'm sure you already know about it but just thought I'd post it anyway.

 

DEPRECATION WARNING: program has invoked obsolete function FreeDopplerSkyScan(). Please see XLALDestroyDopplerSkyScan() for information about a replacement.

solling2
solling2
Joined: 20 Nov 14
Posts: 219
Credit: 1577584642
RAC: 20113

Great that a GPU version is

Great that a GPU version is now on the horizon!

 

Anyone without GPU would surely welcome a wisdom file as well.   :-)

Jim1348
Jim1348
Joined: 19 Jan 06
Posts: 463
Credit: 257957147
RAC: 0

solling2 wrote:Great that a

solling2 wrote:
Great that a GPU version is now on the horizon!

Exactly so.  I would think that it lends itself to parallel processing.  The fact that they are trying to make 0.03 consistent between the two indicates that they are the same thing.  Then I don't see much point in running the CPU version at all.  You might as well devote your CPU cores to something that needs them.  The GPU could do Einstein much more efficiently. 

 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.