[l008com]

I'm wondering about processor usage. One of the reasons I loved RC5 was because it used up my G4's to the MaX. Seti uses about 1/4 or less of my total CPU, probably closer to 1/6. Are any of the other clients for the other distributed programs Altivec Enhanced? And MP Aware?

[Scotttheking]

Originally posted by l008com:
I'm wondering about processor usage. One of the reasons I loved RC5 was because it used up my G4's to the MaX. Seti uses about 1/4 or less of my total CPU, probably closer to 1/6. Are any of the other clients for the other distributed programs Altivec Enhanced? And MP Aware?

They all use all of your CPU, anything less is a problem on the user end, other then the short time when uploading / downloading work.

No other client is altivec enhanced at this time.
distributed folding evaluated their code and found that there is no benefit to altivec. However, the L3 cache of the new G4s does benefit it.

No other client is MP aware, but that's not a big deal. You just run 2 copies of the client, and make a script to start or stop em both at once.

[Shaktai]

Folding@Home will eventually take advantage of altivec when running Gromacs cores, but not yet. Gromacs was rolled out first for Linux, and they have started occasionally rolling out some Windows Gromacs cores. OS-X will be third, but don't expect it for another month or two. Gromacs is a special type of computational core that utilizes 3dnow, SSE and Altivec. Eventually Folding@Home will make greater use of Gromacs across all platforms, but it is still in testing stages for now.

[l008com]

What about seti though? Its almost unfair, Macs could be totally dominating like they did in RC5, but now were stuck way down because of it.

Dose anyone have any idea how long until the next RC5 starts?

[mikkyo]

Originally posted by l008com:
What about seti though? Its almost unfair, Macs could be totally dominating like they did in RC5, but now were stuck way down because of it.
Dose anyone have any idea how long until the next RC5 starts?

<rant>
There is a longstanding issue with Seti - They dont care about processing all the data except in the sense that if they had enhanced clients all the data would get processed super fast and there wouldnt be any more data for a few days, so they care about that.
Seti doesnt have enough input to keep all its user running all the time with efficient clients.
Because of this I cant justify doing seti work.
They need to create a better solution to their problems.
Plus they complain about having to maintain multiple versions of the client(MMX, Altivec, etc plus the regular platform versions).
A mac with an altivec Seti client would indeed be blazing fast, but that will never happen. They had one to demo one time but it never got released, see all the mac users switching to it would have exhausted their input stream in a few weeks, ended the project for a time and made everyone bored. They like having the captive crowd processing all the time and being a popular DC example.
</rant>

[mikkyo]

Originally posted by Scotttheking:

No other client is altivec enhanced at this time.
distributed folding evaluated their code and found that there is no benefit to altivec. However, the L3 cache of the new G4s does benefit it.

D2OL and CommunityTSC use vector math, so if you want fast and useful DC projects they are the way to go. Team MacNN has a pretty good group running D2OL.

I'm curious how Distributed Folding "evaluated" their code, maybe Ill search their forums. They could definately benefit from coding for the G4 and altivec, I have a feeling the just dont want to really investigate it as it would take a lot of time maybe require rethinking their algorithms and the mac client would be very different than the others, plus then they have to test it out, make sure everything is working right and there is always that attitude - our stuff works fine now, why waste time and risk breaking with the chance of not really gaining anything - that most programmers justifiably have.
I think they currently use pointer math, which though efficent at doing some things, is not at all the best way for what they are doing and not the best way for macs.
Pointer math will always be faster at higher clcok rates(MHz) and with faster memory controllers(DDR, etc). This is why an Athlon with DDR screams and newer macs with DDR are faster too.
The L3 cache can help some, but not nearly as much as efficient altivec use.
Coding for altivec would be much, much, much faster - at least 2X.

The future for Folding@Home looks good too, with altivec enhancement planned.

[Shaktai]

Originally posted by mikkyo:
D2OL and CommunityTSC use vector math, so if you want fast and useful DC projects they are the way to go. Team MacNN has a pretty good group running D2OL.

The future for Folding@Home looks good too, with altivec enhancement planned.

D2OL started out as a Windows only project, but with its expansion to Mac and Linux, it has really started to take off. Funny thing is the windows versions are significantly slower then the Mac and Linux versions on a per megahertz basis. I suspect this is due to both the java and vector math implementations in Windows. Eventually Folding@Home will have the gromacs core rolled out across the board, but there is still extensive testing to do before that can be done. It will happen though.

Interestingly, in D2OL we have soared up the ranks with only 3 or 4 people crunching consistently. (I will be back to it soon). #2 is eventually attainable with only a moderate increase in power and #1 is doable but would require a major push. Of course I am sure that other teams will be redistributing their RC-5 power as well.

I however have picked Folding@Home and D2OL as the two projects I feel have the greatest potential to provide nearer term benefit to humanity. dFold is excellent basic science, but I don't see it providing any real benefit for quite a while. SETI? While I do believe that someday we will discover intelligent life elsewhere in the universe, I don't think it will be via the SETI@home project as it exists at this point in time. IMHO for benefit to humanity, F@H or D2OL.

Course, I have been wrong before.

[sideus]

Is D2OL multiple processor aware? If not, is running two clients on one G4 possible?

[Shaktai]

Originally posted by sideus:
Is D2OL multiple processor aware? If not, is running two clients on one G4 possible?

It is not multi-processor aware. I don't know if it is possible to run multiple instances or not?

[sideus]

Heh, guess I should have rephrased that to running multiple instances on a dual G4.

[l008com]

I'm having trouble just running one. The icon seems to have disappeared out of my dock, but the two programs that make up this patchwork of a distributed computing program still look like their running in ProcessViewer.

[Shaktai]

Originally posted by l008com:
I'm having trouble just running one. The icon seems to have disappeared out of my dock, but the two programs that make up this patchwork of a distributed computing program still look like their running in ProcessViewer.

That means you are using Jaguar. Sengent staff are aware of the problem and are working on a fix. D2OL does still run in the background though, as you have noticed. It continues to complete work. Kind of an annoyance though as it makes it difficult to check progress.

You have completed 23 candidates as shown at http://d2ol.sengent.com/memberServic...ile.jsp?id=631

[l008com]

Well as long as its working I like it better this way, keeps my dock nice and clean. :-) We'll see how many diseases I can cure before dnet starts up again :-D

[mikkyo]

Originally posted by sideus:
Heh, guess I should have rephrased that to running multiple instances on a dual G4.

Yes you can run two instances.
Just install one, shut it down after it has a node id and has run for a bit, then move that version somewhere else to a new directory and name if you like.
Install again and start up 2nd client.
Now run both.
Be sure to add the nodes to your account on the d2ol web site.

You may want to invesitgate the java tweaks I posted about awhile ago, though they probably wont give much of a boost.

[l008com]

Since some of you seem to know alot about D2OL, can someone explain something to me? What the hell is this client doing? I understand that were all looking for cures for diseases but beyond that I have no idea exactly what were computing?

[Shaktai]

Originally posted by l008com:
Since some of you seem to know alot about D2OL, can someone explain something to me? What the hell is this client doing? I understand that were all looking for cures for diseases but beyond that I have no idea exactly what were computing?

D2OL is actually DDOL an acronym for Drug Docking Optimization Laboratory. From the D2OL homepage " _ _The Drug Design and Optimization Lab (D2OL)™ works to discover drug candidates against Anthrax, Smallpox, and Ebola and will soon be adding targets for the top 5 major Bioterrorism organisms." (Eventually adding bottulism and plague)

Essentially we are taking known drug possibilities to see how the drug elements dock with Anthrax, Ebola and Smallpox elements. By doing this type of computer analysis we contribute by helping to determine the most likely candidates, out of many thousands, for further specific research as cures or treatments for these and eventually other extreme biohazards. In the realm of biohazards, you don't want to look at just "a single cure or treatment" because biological organisms can mutate or change. Hence you want to have as large a library of probable candidates as possible.

Because of the potential for these particular diseases to be used as weapons, the real world concern is to find many different likely drug candidates so as to provide multiple possiblities for specific research.