Ready Reckoner Area

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284710168
RAC: 110462

RE: This is printed in the

Message 79429 in response to message 79428

Quote:

This is printed in the console window. Sorry I had to pastebin it because it really messed up the forums here.

Bottom of the window is cut off. (fairly large image warning)


Thanks Kath. ( I keep forgetting I'm developing on a high res screen ).

Please re-download as I've altered the window size presets. :-)

Cheers, Mike.

( edit ) It'll fit on 1024 x 768 or higher.

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

KSMarksPsych
KSMarksPsych
Moderator
Joined: 15 Oct 05
Posts: 2702
Credit: 4090227
RAC: 0

RE: Thanks Kath. ( I keep

Message 79430 in response to message 79429

Quote:


Thanks Kath. ( I keep forgetting I'm developing on a high res screen ).

( edit ) It'll fit on 1024 x 768 or higher.

Works find on my monitor now (I'm pretty sure it's 1024x768).

However, I can't seem to load data into it. There's no sample file in this one. And nothing shows up in the RR when I try to load in one of the files from Gary's script (even after choosing all files).

Kathryn :o)

Einstein@Home Moderator

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 686043288
RAC: 582127

Hi Mike! The Code seems to

Hi Mike!

The Code seems to use a Java 1.6 feature (FileNameExtensionFilter). Especially for MacOS this is a problem, since Java 2 SE version 6 isn't yet released for MacOS at all.

CU
Bikeman

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284710168
RAC: 110462

RE: Hi Mike! The Code

Message 79432 in response to message 79431

Quote:

Hi Mike!

The Code seems to use a Java 1.6 feature (FileNameExtensionFilter). Especially for MacOS this is a problem, since Java 2 SE version 6 isn't yet released for MacOS at all.

CU
Bikeman


Thanks, I'll drop that. You'll have to the type extension in yourself.... :-)

I didn't know Mac hasn't yet got 6, so I'll drop code generation back to 5 .... Mac has that?

Quote:
However, I can't seem to load data into it. There's no sample file in this one. And nothing shows up in the RR when I try to load in one of the files from Gary's script (even after choosing all files).


This is Fedora, right? What's the Java version on that? Could you try a keyboard only cut-n-paste into the 'Data Entry' tab text area?

Cheers, Mike.

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

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 686043288
RAC: 582127

RE: Thanks, I'll drop that.

Message 79433 in response to message 79432

Quote:

Thanks, I'll drop that. You'll have to the type extension in yourself.... :-)

I didn't know Mac hasn't yet got 6, so I'll drop code generation back to 5 .... Mac has that?


Yes, java 1.5 is shipped/updated with the OS. There's even a preview version of 1.6 but only for a limited range of hardware, so 1.5 is the safe option.

CU
Bikeman

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5385205
RAC: 0

I seem to get negative

I seem to get negative feedback on my feedback ... but ... here goes ...

I only have the vaguest idea of what you are attempting to accomplish with this tool. There is information in the beginning of this thread that hints at the point of it ... not sure I got it ... but ... why is not this in the read me ...

The read me tells me I am going to use CSV data, and I know what THAT is, what I do not know, and there is no hint of how I should obtain said information. So, even were I interested in trying the tool there is no clear path for the novice to get there ...

Nor, unless I missed it, a link to whatever (if any) documentation there is ... or to the master website.

Lastly, other than to request feedback, which I will see how it is received :) ... there is no hint as to what should be done with the data. The INTENT as I grasped it was to be able to help the project better "model" the fit between something and something ... again, maybe I have completely lost my edge, but, I have no clue what your end goal is ... or how to feed back to the project(s) ... or even if they are really interested ...

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2139
Credit: 2752760842
RAC: 1457500

RE: I seem to get negative

Message 79435 in response to message 79434

Quote:

I seem to get negative feedback on my feedback ... but ... here goes ...

I only have the vaguest idea of what you are attempting to accomplish with this tool. There is information in the beginning of this thread that hints at the point of it ... not sure I got it ... but ... why is not this in the read me ...

The read me tells me I am going to use CSV data, and I know what THAT is, what I do not know, and there is no hint of how I should obtain said information. So, even were I interested in trying the tool there is no clear path for the novice to get there ...

Nor, unless I missed it, a link to whatever (if any) documentation there is ... or to the master website.

Lastly, other than to request feedback, which I will see how it is received :) ... there is no hint as to what should be done with the data. The INTENT as I grasped it was to be able to help the project better "model" the fit between something and something ... again, maybe I have completely lost my edge, but, I have no clue what your end goal is ... or how to feed back to the project(s) ... or even if they are really interested ...


Paul,

Don't worry, you just stumbled on the tail-end of - sorry Mike, a brief hiatus in the middle of - an intercontinental search for the perfect prediction of the expected run-time for Einstein S5R3 workunits. I think it started six months ago here, and meandered through several threads (as is the custom with these message board rambles). You probably don't want to read them all (unless you've got a lot of time on your hands), but to answer your final question: yes, I think we can say the project are interested. This was Bernd Machenschalk's response when he came across the discussion, much the same as you did.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284710168
RAC: 110462

RE: I seem to get negative

Message 79436 in response to message 79434

Quote:
I seem to get negative feedback on my feedback ...


Hello Paul! Are you the Wiki Man? It's cool, I won't 'ramp the amp' on the feedback ... :-)

As Richard indicates this is a work in progress. To cut to the chase ( excluding detailed referencing ) :

- non official, laissez faire attempt at a client side tool to characterise host performance on tasks. It may well be a waste of time. Care has been taken not to consume project resources or of those who have no interest.

- already done in a limited sense as HTML ( try this as the most mature ). Feature-wise you can only do so much with HTML/Javascript. Various requests for improvements triggered the idea of using Java, as it's platform independent .... well, we are ironing that bit out now. :-).

- it has been ( after long ponderings ) discovered that the science apps vary in by as much ~ 30% in execution time. This depends in a semi-complicated way on which unit you get and the sky position that the search is focussed on. Hence any comparison of a prior versus an 'upgrade' version of an app execution wise, ie. Is the app faster and by how much? - is polluted by this effect. Most upgrades are (a) better and but (b) better by much less than ~30%. Ready Reckoner ( derived from 'rule of thumb', 'slide rule' and a cheat sheet I've seen bookies use ) is an attempt to cope with that.

- In the task ( or work unit ) files handed out one can tease out a factor, from the file name, now given the home-grown name of 'sequence number'. It is related to sky search position. For a given frequency if one plots the runtime ( on Y-axis ) and the sequence number ( on X-axis ) you get an upsidedown rectified sinusoid. It looks roughly like a set of drapes hanging from a curtain rail. The 'period', or distance in sequence numbers, between the attachment points at the rail depends on a fiddly function of the continuous wave signal frequency being used in the search template. All of this is artifact of the search strategies chosen to maximise use of computing resources at E@H - specifically minimising overlap of work done as the sky co-ordinates used are 'spherical' ( and the Earth moves around the Sun not along any line anywhere near parallel with it's axis of rotation, and other stuff .... ).

- of course the actual task execution time depends on a blizzard of other factors related to the host configuration. But for a given machine setup you'll get the above described behaviour. RR takes a set of sequence/runtime pairs and attempts to deduce the sinusoid curve they are lying on. You need two or more pairs and preferably with other features too .... it has been found harder than it first looked to reliably/sensibly/meaningfully do this! :-)

- personally I seriously doubt if there is a large audience for this tool. You don't have to be a 'hard core cruncher' to use it, but it may assist evaluation of the behaviour of the apps outside of the test beds used for their development. Even if subsequent strategies lose the sinusoidal feature it may be adapted for whatever follows. Who knows?

- as for me, I'm just tooling around and having fun in my spare time. I hope to mature this tool in Java shortly. The cross platform stuff has been the stumbling block, hence this thread .... :-)

Cheers, Mike.

( edit ) It is the difference between the rail at the top and the lowest part of the hanging drape, divided by the total height of the rail from floor level that gives you ~30%. But that also varies ......

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

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5385205
RAC: 0

RE: Don't worry, you just

Message 79437 in response to message 79435

Quote:
Don't worry, you just stumbled on the tail-end of - sorry Mike, a brief hiatus in the middle of - an intercontinental search for the perfect prediction of the expected run-time for Einstein S5R3 workunits. I think it started six months ago here, and meandered through several threads (as is the custom with these message board rambles). You probably don't want to read them all (unless you've got a lot of time on your hands), but to answer your final question: yes, I think we can say the project are interested. This was Bernd Machenschalk's response when he came across the discussion, much the same as you did.


Ok ...

Well, cool ... however, in light of that, my other points then have more weight ...

If I were interested in using this tool ... the read me is useless to my now feeble brain to jump in ... of course, maybe you don't want feeble brains ...

At the moment, I have "dialed back" on EaH for some other goals (its a surprise) ... :)

But, were I back in my "normal" mode of EaH having a goodly share of resources, well, who knows ...

KSMarksPsych
KSMarksPsych
Moderator
Joined: 15 Oct 05
Posts: 2702
Credit: 4090227
RAC: 0

RE: RE: However, I can't

Message 79438 in response to message 79432

Quote:
Quote:
However, I can't seem to load data into it. There's no sample file in this one. And nothing shows up in the RR when I try to load in one of the files from Gary's script (even after choosing all files).

This is Fedora, right? What's the Java version on that?

Yeah, Fedora 7.

I have no idea what version of Java, but it's pretty old. Like last summer old. I had so much trouble getting it to work in the first place I'm a bit gun-shy on trying to update it.

If someone can tell me how to find the version, I'm happy to look.

Kathryn :o)

Einstein@Home Moderator

Comment viewing options

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