Welcome to the MacNN Forums.

surfdog17 · Jun 16, 2002, 01:19 PM

Thats right the new Folding@Home 3.0 client has been released but there is some bad news, right now it is only for Windows. If I remember reading correctly somewhere a OS X client should be ported within a few days. This new core is about 10-15x faster and maybe more depending on how much Gromacs and Alitvec can speed it up.So keep you peepers on folding.stanford.edu for the new Mac client.

Shaktai · Jun 16, 2002, 04:51 PM

The new client is out of Beta and is available for both Linux and Windows. the plan was to debug the Linux version and then recompile it for OS-X. Since the Linux version is now final, the OS-X version is supposed to follow very shortly.

The new client showed about 10% speed gains across the board on windows and linux boxes. Hopefully that will carry over tot the Mac clients as well. The new Gromacs core which is supposed to provide the big speed gains, is still being rolled out slowly and in small quantities while it is tested. It was first released (briefly) for linux, and now is being released for Windows.

Since Gromacs supports SSE, 3D now and Altivec, all platforms will likely see gains on a near equal basis. However the Intel chips seem to gain the greatest advantage from early reports on the Gromacs core. <img border="0" alt="[Hmmm]" title="" src="graemlins/hmmm.gif" />

SkiBikeSki · Jun 17, 2002, 12:58 AM

Although SSE and 3DNow does help performance, just like Alti-vec. I bet Alti-Vec will provide far bigger performance boosts. I never heard of SSE or 3DNow making 3 or 4 times improvement like Alti-Vec did for RC5 or some Photoshop filters. So I think in the end we shall prevail and get a HUGE boost from the Gromacs and Alti-Vec improvements.

Shaktai · Jun 17, 2002, 01:09 AM

Just posted on the Folding@Home forum by Adam Beberg (The Coder).

<blockquote>quote:<hr />We're doing internal testing on OSX 3.0 now, so should have a expiring-beta around wednesday.<hr /></blockquote>

SkiBikeSki · Jun 17, 2002, 01:15 AM

I guess that's good news. But then again it's an expiring beta, that's bad news. When are they gonna have the full thing done?

Shaktai · Jun 17, 2002, 01:23 AM

<blockquote>quote:<hr />Originally posted by SkiBikeSki:
I guess that's good news. But then again it's an expiring beta, that's bad news. When are they gonna have the full thing done?<hr /></blockquote>If it follows pattern with the Linux and Windows betas, and there are no big problems (shouldn't be, it is a recompile of the Linux client) probably have a final by early July.

SkiBikeSki · Jun 17, 2002, 01:27 AM

Well I'll be sure to d/l and install them as soon as possible. I just hope that they've made some real alti-vec enhancements.

PS Look at the times of these last four posts.

Shaktai · Jun 17, 2002, 03:11 AM

<blockquote>quote:<hr />Originally posted by SkiBikeSki:
Well I'll be sure to d/l and install them as soon as possible. I just hope that they've made some real alti-vec enhancements.

PS Look at the times of these last four posts.<hr /></blockquote>The client doesn't offer any altivec enhancements, only the new Gromac scientific cores. The new client code did provide about a 10% speed improvement for Windows and Linux though. Currently the Gromac scientific cores that provide the big speed improvements are being rolled out gradually. For more info on Gromacs go to <a href="http://www.gromacs.org/" target="_blank">the Gromacs web site.</a>

Shaktai · Jun 19, 2002, 01:10 AM

FAH3 OS-X beta is released. You can get it here.
<a href="http://forum.folding-community.org/viewtopic.php?t=716" target="_blank">http://forum.folding-community.org/viewtopic.php?t=716</a>

However it has an expiration of June 24th. Because it is a recompile of the Linux client which has already been tested, they don't anticipate too many problems.

reader50 · Jun 19, 2002, 02:13 AM

posted, and thanks.

Shaktai · Jul 2, 2002, 01:52 AM

Okay, the second F@H OS-X beta has come and gone (expired today), and they still do not have a working client. The OS-X beta crunches just fine, and seems to also enjoy the same 10-15% performance boost that the Linux and Windows versions got, but it will not upload finished work. Apparently the problem is an "endian" issue. Not sure what that is, but when the client tries to upload finished work, it reverses the IP address, resulting in an bad address and can't upload. No word from Mr. Beberg for several days as to what is happening.

Recommendations? Do not use the Beta until the "endian" issue is addressed. 6 days of crunching down the drain.

However the old client still works just fine. It looks like rolling out the new client and the Gromacs cores is going to take longer then expected. Well it is back to TSC for me, until they get it fixed.

Whoa!!! Just noticed. 600 posts in less than 6 months. Hmm! Should I get a life???

[ 07-02-2002, 02:59 AM: Message edited by: Shaktai ]

The_Equivocator · Jul 2, 2002, 09:25 AM

<blockquote>quote:<hr />Originally posted by Shaktai:
Apparently the problem is an "endian" issue. Not sure what that is, but when the client tries to upload finished work, it reverses the IP address, resulting in an bad address and can't upload.<hr /></blockquote>Here's the scoop on the endian problem:

Motorola processors are big endian, meaning that when they store data, the left-most bytes are the most significant. Intel processors, on the other hand, are little endian, meaning that their right-most bytes are the most significant.

So, let's assume that the program is looking at the binary number

00011000 10001000 00010001 00000100

This is the same as the decimal numbers 24, 136, 17, and 4 (a good representation of an IP address).

All Macintoshes (since they use Motorola processors and are thus, big endian) will store that in memory as:

00011000
10001000
00010001
00000100

Most other computers will store it as

00000100
00010001
10001000
00011000

There have been huge debates about which is better, but there are advantages and disadvantages to both which I won't get into here.

So anyway, we get into a problem if we have the same piece of memory on a big endian and a little endian machine. Instead of reading 24.136.17.4, our computers are probably reading 4.17.136.24.

I really dont see why this would be hard to fix if the only problem is the fact that our application is messing up the IP address. Just reverse the order of the bytes of the address and we are golden. There's gotta be some other issues that are delaying the client.

Anyway, sorry for the long-winded post that no one probably cares about.

[ 07-02-2002, 10:27 AM: Message edited by: The_Equivocator ]

Shaktai · Jul 2, 2002, 01:28 PM

<blockquote>quote:<hr />Originally posted by The_Equivocator:
[QUOTE]

I really dont see why this would be hard to fix if the only problem is the fact that our application is messing up the IP address. Just reverse the order of the bytes of the address and we are golden. There's gotta be some other issues that are delaying the client.

Anyway, sorry for the long-winded post that no one probably cares about.

<hr /></blockquote>Thanks! That at least explains exactly what is happening. The IP address is being reversed when the client tries to send. This has been the case since beta 1. Beta 2 supposedly fixed some network issues, but the the endian issue still wasn't addressed.

I am guessing it shouldn't be a big issue, but they are slow getting it done. Beberg's comments were that the other coders developed on x86 boxes, and that he was the only one having to deal with "endian" issues.

Shaktai · Jul 4, 2002, 12:40 AM

The newest OS-X Folding@Home Client 3.x has been released. It can be downloaded at:

<a href="http://forum.folding-community.org/viewtopic.php?t=826" target="_blank">http://forum.folding-community.org/viewtopic.php?t=826</a>

The endian issue is supposed to be fixed, but be sure to read the message before downloading, there is a neat way to download a 5 minute test work unit, to test your connection and download / upload. Expiration on this beta is July 10th.