BRP4G WU

cliff
cliff
Joined: 15 Feb 12
Posts: 176
Credit: 283,452,444
RAC: 0
Topic 198050

Hi Folks,
Just got over 28 BRP4G WU... with an expiry of 7 days:-)

This is gonna be fun.. considering I run 7 hrs a day to reduce power bils..

Regards,

Cliff,

Been there, Done that, Still no damm T Shirt.

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 420,383,909
RAC: 37,090

BRP4G WU

Me too. Guys, are you tending to finish BRP4 asap now?

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4,289
Credit: 245,817,701
RAC: 10,709

We occasionally and rather

We occasionally and rather unpredictably get new data from Arecibo. "new" here means new to us, i.e. data that we haven't seen (and analyzed) before. In what we got recently there is data from pretty recent observations as well as data that's more than three years old.

Upon arrival, data is divided / prioritized in two categories: recent data where Einstein@Home has a reasonable chance of making a pulsar discovery before other PALFA search pipelines if we process the data fast enough, and rather old data that has already been seen by other PALFA pipelines and for which processing isn't that urgent. The first type data is fed into "BRP4G", i.e. processed by E@H GPUs with a rather short deadline to get results back fast. The second type of data is put on the "BRP4" pile of data, where is processed by smaller, slower (Intel) GPUs and (ARM) CPUs.

The (automatic) post-processing actually doesn't distinguish between whether a "beam" was processed by BRP4G or BRP4, hence there is no such distinction in the progress display on the server status page.

So whenever we get pretty recent "new" data from Arecibo, there are new BRP4G workunits for a while. BRP4 is the only type or work small enough for Android devices, so we try to keep a pile of such workunits in stock to feed our Androids for at least a year. If that buffer is full enough, we may shift some less urgent "beams" over to be processed by BRP4G, but that's pretty rare.

In some sense BRP4 isn't meant to ever be finished.

Actually BOINC should take into account your processing habits when fetching work and shouldn't get you work you can't finish within the deadline. If you're afraid you can't make 7d deadlines, you may de-select the BRP4G application. There's plenty of other (BRP6) work for your GPUs.

BM

BM

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 420,383,909
RAC: 37,090

Oh! E@H is now involved in

Oh! E@H is now involved in race for discoveries, isn't it?
There was no hurry for last 10 years seen by volunteers, especially between PALFA and E@H. Instead almost all the time we heard about scrupulous analysis being held over existing data to find something new in what is already superficially analyzed. Can you explain what for this hurry is when you have to move aside the main project purpose while concentrating on rather permanent and endless observation data crunching?

Mumak
Joined: 26 Feb 13
Posts: 325
Credit: 3,381,152,257
RAC: 419,953

RE: We occasionally and

Quote:

We occasionally and rather unpredictably get new data from Arecibo. "new" here means new to us, i.e. data that we haven't seen (and analyzed) before. In what we got recently there is data from pretty recent observations as well as data that's more than three years old.

Upon arrival, data is divided / prioritized in two categories: recent data where Einstein@Home has a reasonable chance of making a pulsar discovery before other PALFA search pipelines if we process the data fast enough, and rather old data that has already been seen by other PALFA pipelines and for which processing isn't that urgent. The first type data is fed into "BRP4G", i.e. processed by E@H GPUs with a rather short deadline to get results back fast. The second type of data is put on the "BRP4" pile of data, where is processed by smaller, slower (Intel) GPUs and (ARM) CPUs.

The (automatic) post-processing actually doesn't distinguish between whether a "beam" was processed by BRP4G or BRP4, hence there is no such distinction in the progress display on the server status page.

So whenever we get pretty recent "new" data from Arecibo, there are new BRP4G workunits for a while. BRP4 is the only type or work small enough for Android devices, so we try to keep a pile of such workunits in stock to feed our Androids for at least a year. If that buffer is full enough, we may shift some less urgent "beams" over to be processed by BRP4G, but that's pretty rare.

In some sense BRP4 isn't meant to ever be finished.

Actually BOINC should take into account your processing habits when fetching work and shouldn't get you work you can't finish within the deadline. If you're afraid you can't make 7d deadlines, you may de-select the BRP4G application. There's plenty of other (BRP6) work for your GPUs.

BM

Since we still seem to have BRP4G work (occasionally), will you consider to adopt the same improvements as done for BRP6 Beta into BRP4G too? Of course only if that would improve performance.

-----

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2,142
Credit: 2,812,931,675
RAC: 836,156

RE: Oh! E@H is now involved

Quote:
Oh! E@H is now involved in race for discoveries, isn't it?
There was no hurry for last 10 years seen by volunteers, especially between PALFA and E@H. Instead almost all the time we heard about scrupulous analysis being held over existing data to find something new in what is already superficially analyzed. Can you explain what for this hurry is when you have to move aside the main project purpose while concentrating on rather permanent and endless observation data crunching?


I don't think that's a fair objection. The primary purpose of this project is to study (potential) gravity wave detections by - yes - searching through the permanent and potentially endless observational output of the LIGO detectors. That goes on in the background, using the S6Bucket application(s), which for the time being only run on CPUs. We just don't talk about it much.

Alongside that, we can use different compute resources - GPUs - to search through different types of observation. It doesn't seem unreasonable to occasionally participate in the initial scan of brand-new observations from Arecibo, alongside the long-term re-analysis of the Parkes data - especially since neither of them is in competition with the resources used for the primary (LIGO) search.

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 420,383,909
RAC: 37,090

RE: That goes on in the

Quote:
That goes on in the background, using the S6Bucket application(s)

I can't agree with you, because S6 and other searches (s5, s4, s3 and perhaps future ones) were the main part of the project over last ten years, while BRP is an additional aim helping us to concentrate efforts on most promising candidates (e.g. points in the sky) due to huge amount of computation power needed even for current level of sensitivity of S6 run.
And I can agree with concentrating our efforts on any secondary purpose only if the main project doesn't have enough data to move forward. This in my opinion would be fair enough.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,564
Credit: 291,476,350
RAC: 110,028

The non-gravity work was

The non-gravity work was engaged for several reasons ( this is multi-year-old news to most regulars but I will repeat for the clarity of others ).

Firstly it was intended to be relevant for the period awaiting the LIGO upgrades now completed ie. give us here at E@H something to do while the advanced interferometers were brought into being. That purpose has been fulfilled. ( Alas the intermediate Enhanced LIGO configuration ( ~ 2010 ) really didn't technically perform well enough for E@H signal searches, regardless of whatever else was learnt during that time ).

Secondly the PALFA group had a need for signal analysis that they couldn't provide. This purpose has definitely been fulfilled, and then some !

Thirdly the astronomical objects of interest - pulsars - are the same target class as those the Continuous Wave Group of LIGO are studying. Especially non-axisymmetric pulsars ( 'bumpy' if you like ) that ought radiate GWs, and such signals are the ones we will soon be trawling the AdLIGO data streams for. We now have a well practised and understood pipeline awaiting this role.

[ The gamma-ray work was later added in with much the same logic. ]

As for permanent and endless crunching, that was and always will be true. There is a vast parameter space out there to explore for any particular data set. This is a simple consequence of the blind search strategy, by definition one is choosing as yet untried signal templates in the hope of gaining matches. This is akin to listening for a bird call that you have never before heard*. I urge you all to interpret this quality as an exciting aspect of the project indeed ! Even if E@H is not the first to detect a GW that doesn't mean any subsequent work we do won't be novel, interesting and/or groundbreaking in it's own right. Far from it. We are on the verge of a really fascinating era of discovery !!! Come on fellow crunchers, get your grins out in anticipation please .... :-) :-)

[ Yes I know. For many of us it has been a long 10 years. As decades tend to be. So it may be hard to rustle up even a wee bit of 'go fever' at this stage. But I try .... :-) ]

As for which 'direction' the searchlight is pointed in parameter space for any given phase of this project, then I trust the overseeing scientists to judge that. Evidently they seem to have an effort vs. reward tension in such decisions. I'd be a fool to attempt a second guess.

Personally I only crunch for E@H, and somewhat mono-manically to boot. I've ticked all the boxes for task type availability, hence my rigs will follow the flavors as determined by project managers. But that's just my ( lazy ) choice. I can't be bothered managing below that level.

Cheers, Mike.

( edit ) IF there is an implicit criticism lying around somewhere along the lines of the initiators of E@H being unable to foresee in detail what the project would be doing 5, 10+ years hence of ~ 2003, THEN I can only say it really sucks to predict the future !! Please stop reading now, go and do something far more fruitful and deterministic with your life/brief-mortal-coil : pick & bet on some winners for your local horse races this weekend coming up. Bleeding edge astrophysical investigation is not for you !! :-)

* In a very bloody noisy forest where you can't see any of them.

( edit ) A few years into my involvement here I remember clearly being heavily struck one day with the realisation of how ambitious it is to attempt the measurement of effects of order 10^[-24]. Thus I cannot understate how amazing it is that both LIGO IFOs have recently demonstrated prolonged lock, and certainly good enough quality for serious analyses, at that design spec. WOW !! :-)

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Comment viewing options

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