Bruce, a question about An Optimized Application

Michael Roycraft
Michael Roycraft
Joined: 10 Mar 05
Posts: 846
Credit: 157718
RAC: 0

RE: Will 3DNow!

Message 23721 in response to message 23720

Quote:
Will 3DNow! optimizations also make it into the next offical client? Akosf's S39L worked great on my machine, but his D40 version with the 3DNow! optimizations was even better! I just hope we Athlon XP users aren't forgotten. :-)

Kerwin,

Who knows? Would be nice, though. If I ever get the OK to use the electricity to bring my old rig online, I'll probably use my backup XP_Mobile Barton in it, but if and when I get a place of my own, with no power restrictions, I have on or two old XP 2500 T'breds that may again see the light of electron flow. :-) For my Barton core (with SSE), the 3DNow! app is only about 1-2% faster than the SSE app.

Michael R.

microcraft
"The arc of history is long, but it bends toward justice" - MLK

Kerwin
Kerwin
Joined: 22 Jan 05
Posts: 7
Credit: 10120613
RAC: 0

RE: Kerwin, Who knows?

Message 23722 in response to message 23721

Quote:


Kerwin,

Who knows? Would be nice, though. If I ever get the OK to use the electricity to bring my old rig online, I'll probably use my backup XP_Mobile Barton in it, but if and when I get a place of my own, with no power restrictions, I have on or two old XP 2500 T'breds that may again see the light of electron flow. :-) For my Barton core (with SSE), the 3DNow! app is only about 1-2% faster than the SSE app.

Michael R.

Yes it would be nice! If the only changes to the next official app are the optimizations, then there should be no reason why us AMD folks can't continue using the D40 app...probably with an app_info.xml file of course.

[Edit]
Then again, there's nothing stopping akosf from taking the next official app and optimizing the hell out of that. ;-)

Michael Roycraft
Michael Roycraft
Joined: 10 Mar 05
Posts: 846
Credit: 157718
RAC: 0

RE: Yes it would be nice!

Message 23723 in response to message 23722

Quote:

Yes it would be nice! If the only changes to the next official app are the optimizations, then there should be no reason why us AMD folks can't continue using the D40 app...probably with an app_info.xml file of course.

[Edit]
Then again, there's nothing stopping akosf from taking the next official app and optimizing the hell out of that. ;-)

Well, that's the thing, though - what is being talked about here is the next official app incorporating Akos' opts, and (if everyone gets their wish) there being several official versions, each tailored for extensions. As you've probably noticed, the last few have been relatively tiny "tweaks", amounting to 1-5%, compared to the mighty leaps (20-30%) of the first 3 or 4 Akos Miracles. If the next official app(s) do come out this way, I wouldn't expect that there will be much more optimization possible on top of what he's already accomplished. We're getting pretty demanding, aren't we? Geez, give the guy a break :-) They're already going to go through hell, just trying to get Linux and Mac into the same neighborhood, or at least get them into the same TimeZone.

microcraft
"The arc of history is long, but it bends toward justice" - MLK

peterthomas
peterthomas
Joined: 6 Nov 05
Posts: 47
Credit: 23940
RAC: 0

RE: (edited for length)

Message 23724 in response to message 23723

Quote:
(edited for length) We're getting pretty demanding, aren't we? Geez, give the guy a break :-) They're already going to go through hell, just trying to get Linux and Mac into the same neighborhood, or at least get them into the same TimeZone.

While reading this post, a rather amusing but terrifing thought occured to me.
That being

"They could implement daylight saving to try and bring the mac's and lunix into line."

I know its a scarey thought, those who know about daylight saving will I hope understand while those who have never heard about this abomination should thank their lucky stars.

bloed_brot
bloed_brot
Joined: 5 Apr 05
Posts: 70
Credit: 91124558
RAC: 0

RE: RE: akosf said he

Message 23725 in response to message 23698

Quote:
Quote:

akosf said he looked at the disassembly of the linux binaries, but they were different enough he wasn't able to figure out what was going on.

Bruce's promised new, official applications based on akosf's algorithm improvements, but they haven't done so yet. Mainly I suspect the delay is in translating the assembly back into C to allow crossplatform development.

Here's an update.

We've now incorporated Akosf's improvements into our source code. But we haven't started distributing this faster application yet, for a simple reason. We are worried that our project server might break with the increased upload/validation disk load since the work will be getting done faster when we begin distributing new apps to all users. So we're upgrading the disk controllers and should be ready for this increased load soon.

Alright there,

I will now cease calculations on my linux machines, since my one win latop now is able to outnumber the entire RAC I used to have before. This removes the strain on the project server's disk controller at least a little.

From a discussion in our team board, voices suggested handing over the source code to akosf since he has done so much great work. This would enable him to have his optimizations running on all platforms, to my understanding. Would that be at all feasible?

:
your thoughts - the ways :: the knowledge - your space
:

bloed_brot
bloed_brot
Joined: 5 Apr 05
Posts: 70
Credit: 91124558
RAC: 0

RE: As you've probably

Message 23726 in response to message 23723

Quote:
As you've probably noticed, the last few have been relatively tiny "tweaks", amounting to 1-5%, compared to the mighty leaps (20-30%) of the first 3 or 4 Akos Miracles.

On my machine there was not a 20-30% increase, not even 200% but more like a whopping 350%! I was cruising at ~180 RAC, now I am well beyond 520 RAC and counting.

:
your thoughts - the ways :: the knowledge - your space
:

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284730859
RAC: 105773

Bear in mind that there has

Bear in mind that there has been a LSC ( Ligo Scientific Collaboration ) meeting in recent weeks ( see here ), so I'd guess that our fearless leaders may well have been distracted/busy with that....
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

bloed_brot
bloed_brot
Joined: 5 Apr 05
Posts: 70
Credit: 91124558
RAC: 0

RE: Bear in mind that there

Message 23728 in response to message 23727

Quote:
Bear in mind that there has been a LSC ( Ligo Scientific Collaboration ) meeting in recent weeks ( see here ), so I'd guess that our fearless leaders may well have been distracted/busy with that....
Cheers, Mike.

Lousy excuses, while we are in the trench of the frontline! :-D

No, I know they work as hard as they can and there are many more obligations to follow, so I just hang in there and wait.

:
your thoughts - the ways :: the knowledge - your space
:

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284730859
RAC: 105773

RE: Lousy excuses, while we

Message 23729 in response to message 23728

Quote:
Lousy excuses, while we are in the trench of the frontline! :-D
No, I know they work as hard as they can and there are many more obligations to follow, so I just hang in there and wait.

Actually I'll share some interesting stuff I found while browsing the presentations:

Quote:
[pre]
Up to Mar 20 2006 01:31:02 UTC
Elapsed run time = 3249.5 hours
Sample Hours Percent
------- -------- -------
H1 2075.2 63.9 since Nov 4, 2005 8:00 PST
H2 2356.7 72.5 since Nov 4, 2005 8:00 PST
L1 1658.2 55.1 since Nov 14, 2005 12:00 CST
H1+H2 1897.6 58.4 since Nov 4, 2005 8:00 PST
H1+L1 1261.4 41.9 since Nov 14, 2005 12:00 CST
H2+L1 1316.5 43.8 since Nov 14, 2005 12:00 CST
H1+H2+L1 1154.3 38.4 since Nov 14, 2005 12:00 CST
One or more 2768.9 85.2 since Nov 4, 2005 8:00 PST
Two or more 2166.8 66.7 since Nov 4, 2005 8:00 PST
(H1orH2)+L1 1423.6 47.3 since Nov 14, 2005 12:00 CST
[/pre]


These are the 'locked' hours for the interferometers - thus producing the 'good' data for further signal analysis (H1 and H2 are Hanford big & small, L1 is Livingstone). The triple co-incidence time (H1+H2+L1) is already at about 1200hrs - twice the length of the 600hrs we are analysing now for the S4 run. Someone here recently asked 'what next' - well it's coming! Also akosf's ( or any other ) optimisations are going to flow for S5 too and be very bloody useful!
Another table relates to increased detector sensitivity:

Quote:
[pre]
Before break After break (best)
L1 ~10 Mpc ~12 Mpc
H1 ~11 Mpc ~14.5 Mpc
H2 ~5 Mpc ~7 Mpc
[/pre]


(The 'break' is the period between the S4 and S5 runs ).
These are distance estimates ( in Megaparsecs or Mpc ) to detection limit on current sensitivity for a 'standard' binary inspiral event - discovery of which is keenly sought. Because volume goes as cube of distance, a 20% increase in radius, say, gives about a 70% increase in the sphere of space enclosed ( 30% gives about another 120%, 40% gives about another 170% ).
The upshot is that S5, as we speak, has collected significantly longer times on considerably larger volumes of space. So who wants their computer(s) to host the calculations that reveal some real breakthrough for science?!
Pick me... pick me... pick me... pick me... pick me... :-)
( .....hand upstretched in classroom.... )
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

rbpeake
rbpeake
Joined: 18 Jan 05
Posts: 266
Credit: 968423676
RAC: 1163715

RE: Actually I'll share

Message 23730 in response to message 23729

Quote:
Actually I'll share some interesting stuff I found while browsing the presentations....
Cheers, Mike.


Thanks for taking the time to review the presentations, this is very interesting information! I was wondering how much "good" triple-coincidence time data they were collecting, because the intruments are so incredibly sensitive they pick up the very least bit of earthly background noise. In fact, imho, the sensitivity is such and the planet is so noisey that a space-based instrument seems probably best (and I know one is planned eventually).

Updates like yours are encouraging, and certainly keep up my interest in the project! :)

Comment viewing options

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