[Todd Madson]

re: And a third for random praise.
Heh. Yeah, no kidding. I still can't believe my G5 is churning out 90%
of the blocks it does in under an hour and sometimes much less.

I'm #1 on my team (which dwindled to four people from eleven) and
am responsible for over half of the work units in the new regime.

Maybe I need to find a different hobby?

[TiloProbst]

imho Snake_doctor is right, so i dont know if this one still belongs here .. maybe. anyways.

i started a thread in the official seti@berkeley forums, because the stats of my dual g5 (id 2012766) have not been updated since January 1st. uploading/downloading/crunching/connecting seemingly work fine. it cant be the internet connection or something, since my second computer on the same connection works as usual including updatet stats.

the thread url is: http://setiathome.berkeley.edu/forum...d.php?id=26839

i wonder if that could be an issue with the beta clients? i mean, my second computer works fine with the same software, but who knows :/
i use BOINC client version 5.2.13 and the seti@home-G5-a5 workers.

[halimedia]

Tilo, have you tried resetting the project (S@H)? You have nothing to lose except the WUs in progress (downloaded ones will be processed fine after the reset). I'd try this first, if you haven't yet...

HTH,

Ron

[TiloProbst]

from the BOINC wiki:

"One of the more important aspects of a Project "Reset" is that it will abort any Work Units/Results in progress, including those that have been completed but not "Reported" (see Reporting Process)."

that means, i gonna lose all result credits since January 1st .. is there no other way??

EDIT: Boinc Wiki says:
http://boinc-doc.net/boinc-wiki/inde...t_%22Update%22

so how do i "update" in order not to lose the credits of the recent ten days?

and finally, how do i actucally reset?

maybe i am not the one to ask, but why is there no sticky thread regarding those questions?

[beadman]

TiloProbst:

If you are using BOINC Manager, click on the "Projects" tab, and you'll see the Update and Reset buttons.

If you are using CLI in Terminal, enter a Control-C to interrupt BOINC, then enter

./boinc... -help (the "boinc..." is whatever you normally type in to get it to start running, then a space, and then the -help)

BOINC will respond with a list of allowable statements. It's something like
./boinc... -update_prefs http://url_of_the_project_you_are_updating

Hope this helps...
beadman

[BTBlomberg]

Do you have a bunch of 100% WUs in the work screen of the Boinc Manager? I often go to the menu bar and do "Commands -> Retry Communication". That typically uploads any stuck ones for me. Also, the project has been up and down (as usual) since last Wednesday and my Dual G4 hung up after the down time and I had to quit Boinc and restart to get it going again. I had some 20 WU sitting there at 100%. No new crunching and my others had been back at it for days.

Also, Berkley has had Network issuers after the downtime. Lot of rejected connections.

The messages screen on Boinc manager will tell you a history of it's activity. could be it is postponed communication 24 Hours or more bacause of rejections.

[TiloProbst]

Originally Posted by beadman

BOINC will respond with a list of allowable statements. It's something like
./boinc... -update_prefs http://url_of_the_project_you_are_updating

Hope this helps...

helps a lot thx you! bash looks good
btw .. -update_prefs without -dir <..> does nothing, just dumps file to ./

EDIT
my machine is back on the team stats page, RAC jumped from 300+ to 500+!
Thank you guys, you are great

q033:/ tiloprobst$ /Applications/Boinc5/boinc_5.2.13ppcG5gui -dir /Applications/Boinc5/ -update_prefs http://setiathome.berkeley.edu/
^[[D2006-01-10 02:44:57 [---] Starting BOINC client version 5.2.13 for powerpc-apple-darwin
2006-01-10 02:44:57 [---] libcurl/7.14.0 OpenSSL/0.9.7b zlib/1.2.2
2006-01-10 02:44:57 [---] Data directory: /Applications/Boinc5
2006-01-10 02:44:57 [SETI@home] Found app_info.xml; using anonymous platform

(...)

2006-01-10 02:44:58 [SETI@home] Resuming computation for result 18fe05ab.3121.9649.573572.1.26_0 using setiathome version 418
2006-01-10 02:44:58 [SETI@home] Resuming computation for result 14fe05ab.25600.32272.509654.1.141_1 using setiathome version 418
2006-01-10 02:44:58 [---] Suspending work fetch because computer is overcommitted.
2006-01-10 02:44:58 [---] Using earliest-deadline-first scheduling because computer is overcommitted.
2006-01-10 02:45:00 [SETI@home] Sending scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi
2006-01-10 02:45:00 [SETI@home] Reason: Requested by user
2006-01-10 02:45:00 [SETI@home] Reporting 130 results

[reader50]

Originally Posted by Snake_doctor

Folks,

I am certain that most of the readers of this thread would agree that we could not be any happier for you guys that are running the new G5 Quads. Clearly discussion of those systems is on topic if Rick or Alex ask you questions related to the SETI development. But this thread is now really, really, really long, and it is almost impossible to follow the discussion of the SETI worker development. Many people have posted on topic questions which get quickly buried in off topic discussion and go unanswered. It would be considerate if you could either stay on the titled subject, go to PM mode, or perhaps create a new thread about G5 dual core systems and noise/cooling issues.

Alternatively, Rick and or Alex might consider creating a new thread starting with the links to Rick's Website at the top of this one, covering just the Altivec SETI worker development.

Either of these options would be a great service to people who are interested in SETI development.

Regards
Phil

This was well said.

Normally, generic questions about Quads (or other PowerMacs) go in our PowerMac Forum. If people posted such questions in any other MacNN forum, they would be moved to PowerMac. If the post mentioned it was related to DC, it might be moved here, or to PowerMac, depending on if the Moderator who moved it noticed the reference.

However, we don't enforce forum-specific posting too much in the Team forum because this is an open contact zone for people from all over. Rival teams, client development experts, all are welcome. So long as it's somehow related to Distributed Computing, we don't move the threads or posts. Off-topic posts in threads are seldom a problem either, because we have few long threads. And a few off-topic posts by regulars are fine too.

There *have* been an awful lot of Quad posts unrelated to the alpha client development, I hadn't noticed just how many. Most likely because I'm not lucky enough to own a Quad. Yet.

So, to those who want to quiet their Quads (or other PowerMacs) down, please make future posts on that topic in a separate thread. It can be in this forum since it's definitely DC related. I could move the existing Quad-noise posts into a separate thread, but the missing posts would confuse people who only visit this thread. They may think their posts were deleted (censored) for no obvious reason, several weeks after being posted.

BOINC configuration problems that are unrelated to the alpha client belong in their own threads too. If anyone with BOINC experience would like to write up a FAQ for BOINC installation and problem solving, it can be linked from our Resource Pages thread. That thread is a sticky already. We'd even be happy to host such a guide on our team site, alongside our other help pages. If anyone writes a guide and needs it hosted, send a PM my way and we'll go on from there.

I'd rather not create a mess of stickies for very specific subjects, other sites' forums have made that mistake. They end up with a page of stickies before you reach any live threads. There is no need to make this thread a sticky, it stays high because it is the highest-traffic thread in the Team Forum's history.

[TiloProbst]

does that one help?

http://apfelwiki.de/wiki/Main/Nap

Originally Posted by halimedia

Two related questions:
1.) Esa - is it possible to switch Nap off/on via CLI? If so -how?

[halimedia]

Originally Posted by TiloProbst

does that one help?

http://apfelwiki.de/wiki/Main/Nap

Thanks, Tilo! That's a goodie! I have been running all my boxes that support Nap with Nap disabled for over two days now, and guess what - it doesn't affect WU times a bit. So I'll forego this one...

Edit: for those curious, it's 'hwprefs cpu_nap=false' (or true to enable Nap again).

[Todd Madson]

Eh, something broke here. Or rather I broke it I believe.
I was running the 4.44 superbench app with the alpha 5 client and updated the client
to 5.213 in order to gain the "retry communications" option in the menu.

Now my math performance and cpu benchmarking is horrid. I still have the alpha 5
data in the library / application support / boinc data / projects folder as before but it
appears to not want to use this.

I suppose I'll have to try and go back to 4.44 to make it work again eh?

[Todd Madson]

Horrible. I tried going back to 4.44 and now all it says is:

"Boinc Manager - Connection Error:
The Password You Have Provided is Incorrect, Try Again" with an OK dialog.

But there's no obvious place to re-enter this in the 4.44 connection manager.
Help!

[Todd Madson]

Apparently I can't go back, I can only use the 5.213 version - how on earth do I get this
darn thing to use the Altivec optimized version? It's sitting in the same directory but is
not using A5?

Update: the 5.213 version is indeed running the A5 client on the two processor(s) but
it is producing MUCH longer runtimes for the estimate of completion. Three and five
hour run times? Also, floating point and integer speed are WAY down (integer is in
the 4200 range, was formerly 16000-ish, etc.)

Yipes.

[beadman]

Todd:
Did you install the "plain Jane" version of 5.2.13, or the Team MacNN SuperBench version?

The SuperBench version here should give you the higher benchmarks...

beadman

[Todd Madson]

THAT'S IT!

Geez. I was terrified for a second thinking the new software update (10.44)
or the updated boinc manager mangled my math libraries or reduced overall performance.

Thank you!

As of 10:11 pm. 10January:
6626 double precision MIPS (Whetstone) per CPU
21951 integer MIPS (Dhrystone) per CPU
DP 2.5 + 2.5 GB ram

(The numbers seem higher than under 10.43 - can anyone verify they are experiencing
better math performance under 10.44? I maxed out around 16k integer and around 5k
for double precision under 10.43 I believe - can anyone with a quad do a little test?)

[halimedia]

Originally Posted by Todd Madson

The numbers seem higher than under 10.43 - can anyone verify they are experiencing better math performance under 10.44? I maxed out around 16k integer and around 5k for double precision under 10.43 I believe - can anyone with a quad do a little test?

I'd say it's the change from 4.44 Superbench to 5.2.13 Superbench. Both my Quad and the Dual 2.5 GHz benchmark at around 6600 Whetstone and 22500 Dhrystone per CPU (both running Menubar 5.2.13 Superbench under 10.4.3). These figures were clearly lower under 4.44 Superbench.

Also see my post at the bottom of page 11 regarding being able to claim higher credit with 5.2.13 Superbench compared to the stock client or even 4.44 Superbench.

Have yet to install 10.4.4 - have to do some Aperture 'visual benchmarks' first to see if 10.4.4 solves the issues I'm having. Incidentally, Todd: what's the build of 10.4.4 GM?

[halimedia]

Hope this is not too off-topic, but I think it will help many here to understand the BOINC credit granting system better:

http://en.wikipedia.org/wiki/BOINC_Credit_System/

Got to love Wikipedia!

[TiloProbst]

after I forced Boinc to -update on both of my computers everythin worked fine a few days. now Terminal output says that my WUs are outdated. I did an -reset project on one of the computers .. now what happens is exactly NOTHING, I just doesn't fetch any new work.

Terminal looks like this:
2006-01-12 22:54:31 [SETI@home] Resetting project
2006-01-12 22:54:31 [---] request_reschedule_cpus: exit_tasks
2006-01-12 22:54:31 [SETI@home] PERS_FILE_XFER_SET::remove(): not found
2006-01-12 22:54:31 [SETI@home] PERS_FILE_XFER_SET::remove(): not found
2006-01-12 22:54:31 [SETI@home] PERS_FILE_XFER_SET::remove(): not found
2006-01-12 22:54:31 [SETI@home] PERS_FILE_XFER_SET::remove(): not found
2006-01-12 22:54:31 [SETI@home] Project http://setiathome.berkeley.edu/ has been reset
(...)
2006-01-12 22:54:56 [---] General prefs: from SETI@home (last modified 2005-12-15 18:24:27)
2006-01-12 22:54:56 [---] General prefs: no separate prefs for home; using your defaults

[halimedia]

Hey, what's the matter with Steve Zaharias' Wintel box? It's in <a href="http://setiweb.ssl.berkeley.edu/top_hosts.php" target="_blank">second place</a>!!! Alex, Rick: Is there a new Wintel worker with optimizations we don't have? Can't let them beat the Quads now, can we?

Edit: Ooops! I guess that 'box' has <a href="http://setiweb.ssl.berkeley.edu/show_host_detail.php?hostid=2055955" target="_blank">32 CPUs</a>. Nothing to worry, then... Sorry for wasting bandwidth...

[TiloProbst]

well, the Windows clients are updated very often, and there is a wide range of different clients.

in my team I am the only mac user. I somehow dont like the clients that my team mates use. for example our number one computer is a 3.2 Ghz HT Pentium4 running WinXP, overclocked to 3,8 Ghz HT. updating the client maybe two months ago improved WU times only a little, while the credit granted jumped by almost 40% ... RAC was around 800 before, jumped to 1300.

[TiloProbst]

Originally Posted by TiloProbst

.. now what happens is exactly NOTHING, I just doesn't fetch any new work.

hmm ok, now the next morning Boinc fetched at least one WU .. looking at -help there doesn't seem to be a command to force Boinc to ask for work ..

what does -exit_when_idle do? is there a documentation of all CLI comands somewhere?

[Todd Madson]

Notice too that that 32 processor windows box is running the datacenter edition of
windows 2003 server. Interesting. (Update: 32 processors and 20 gigs of ram, that
would explain it).

I noticed now that my average RAC has risen on the DP 2.5. Cool. But I wish it
would be greater than 1K all by itself.

Originally Posted by halimedia

Better than staging a car race on public roads an endangering innocent bystanders in the process, IMO

Edit: I have a confession to make myself... I'm trying to claim one of the top three spots <a href="http://www.boincsynergy.com/stats/country.php?country=Switzerland&project=sah" target="_blank">of Swiss SETI crunchers</a>. Slooowly getting there...

LOL! I'm a complete petrolhead and 'cyberface' is my name on a few car-fanatic forums, not sure if that's what you had in mind (I don't road race but do enjoy track driving)

Anyway thanks for all your help, I'm now sitting at #7 worldwide for recent average credit based on your and ET's recommendations.

Unlike ET, my Quad is my primary workstation and so Boinc doesn't get all the cycles... however still very impressed with the raw performance of this machine, new Intel Core Duo Macs or no.

The Quad is a massive improvement over the previous Dual 2.5... I've got that running the same config and it's currently RAC 892.83 versus 2616.54 for the Quad - and the Dual CPUs are pushing 85-90˚C whereas each Quad core never goes above 66˚C.

Impressive given they must be concentrating most of their resources on Intel hardware now.

Anyway sad geekery fulfilled, all I was aiming for was top 10 and I've got 7th at the mo....

[halimedia]

Alex, Rick, here's another datapoint: DA G4 DP 533 MHz, 768 MB CL 2.5 RAM upgraded from OSXS 10.3.9, BOINC CLI Superbench 4.44 and alpha-4 to OSXS 10.4.4, BOINC CLI Superbench 5.2.13 beta and alpha-5. WU times shrank from 11300 to 9500 seconds (<a href="http://setiweb.ssl.berkeley.edu/show_host_detail.php?hostid=1415895" target="_blank">link to host</a>).

Edit: this box crunches on both CPUs.

[halimedia]

Originally Posted by cyberface

LOL! I'm a complete petrolhead and 'cyberface' is my name on a few car-fanatic forums, not sure if that's what you had in mind (I don't road race but do enjoy track driving)

I've got an amazing talent of putting my foot in my mouth lately. Clearly, I had no knowledge of your 'petrolheadedness'. No harm intended - just sad geekery, as you said

[Lauger]

Which is faster? The Superbench version or the alpha-5 version? I'm running 5.2.5 and the a-5 version on my 2.0GHZ DP. Should I upgrade to the 5.2.13 client?

http://setiathome.berkeley.edu/resul...hostid=1596056

[halimedia]

The version of the BOINC client has no impact on the crunching speed of the SETI worker. alpha-5 is as fast as it gets at the moment. What difference the Superbench client can make you can read in a post I made on Page 11.

HTH,

Ron

[Karl Schimanek]

Originally Posted by halimedia

Hey, what's the matter with Steve Zaharias' Wintel box? It's in <a href="http://setiweb.ssl.berkeley.edu/top_hosts.php" target="_blank">second place</a>!!! Alex, Rick: Is there a new Wintel worker with optimizations we don't have? Can't let them beat the Quads now, can we?

Edit: Ooops! I guess that 'box' has <a href="http://setiweb.ssl.berkeley.edu/show_host_detail.php?hostid=2055955" target="_blank">32 CPUs</a>. Nothing to worry, then... Sorry for wasting bandwidth...

But the guy on fifth place is terrible fast. 1100 seconds per WU
But his problem is the worse claimed credit

Regards
Karl

[Todd Madson]

10.44 is 8G32 btw. (1100 seconds. Yikes).

[Todd Madson]

Thinking. My G4/400 now gets under 3 hours per work unit (!) (optimized worker G4 A5),
generally about 10,000 seconds plus or minus a few (I keep seeing about 2 hrs 30 minutes
to about 2 hours 46 minutes frequently).

Under Seti Classic it was always 14 hours or so. I'm still stunned by the performance here.

I wonder if A6 will be able to increase performance by any amount?

The G5 DP 2.5 is taking about 50 minutes to an hour and 5 minutes per work unit now.


My guess is some of the issues are hardware limitations (cache in the G5 chips, the authors
of the alphas don't have access to a G5 to play with to optimize the routes, or at least didn't
have access at one time).

Obviously a Quad benefits from larger caches, two more processor chips, and the ability
to divide the work into four chunks for higher throughput.

I'd be real interested to see how the new Core Duo iMacs do on this particular application.

[TiloProbst]

what exactly does this Terminal output mean:

2006-01-15 20:15:31 [SETI@home] Temporarily failed upload of (..): error 500

and this one:

2006-01-16 01:40:15 [---] Resuming round-robin CPU scheduling.

[virex]

The first one means that the connection errored out while trying to upload. makes sense since berkeley has been having issues.

The second one means that it is going back to crunching according to your priorities. It must have had some debt to a project that it had to force crunching for, or the deadline was approaching so it focused on one project. Now it's back to switching between them round-robin style.

[TiloProbst]

thank you, but what exactly does this "round-robin style" mean?
(<- not a mothertongue english speaker)

i guess it came from the approaching deadline as a result from the problems i had on january 1st (producing even some overdue results).

on my other computer i get this one since a few days when i start up boinc:
[SETI@home] Result 17fe05aa.16807.21650.965922.1.204_0 is 3.10 days overdue.
[SETI@home] You may not get credit for it. Consider aborting it.

what can i do about it? is there any way of aborting a WU with the CLI client?

[Knightrider]

Originally Posted by TiloProbst

thank you, but what exactly does this "round-robin style" mean?
(<- not a mothertongue english speaker)

round robin

A round robin is an arrangement of choosing all elements in a group equally in some rational order, usually from the top to the bottom of a list and then starting again at the top of the list and so on. A simple way to think of round robin is that it is about "taking turns." Used as an adjective, round robin becomes "round-robin."

In computer operation, one method of having different program process take turns using the resources of the computer is to limit each process to a certain short time period, then suspending that process to give another process a turn (or "time-slice"). This is often described as round-robin process scheduling.

courtesy http://whatis.techtarget.com/

K.

[TiloProbst]

Originally Posted by Knightrider

round robin
http://whatis.techtarget.com/

K.

cool, thx

so anybody knows how to abort those two overdue WUs as described above?
seems I don't get any credit since this one appeared .. o_O

WOOOOO I'm number 2 at the mo, behind ET of course. All I can say is WOW this Quad is impressive, it's my primary workstation and not a dedicated 'seti' cruncher.... (I'm 'lx' by the way)

Can't see it being there for long due to that rather meaty 32 CPU Intel box so must take a screenshot

[E.T from tellus]

Party time!!!!!!
top 4 -> all Macs

[virex]

well my quad will be up there shortly....so lets hope the top 5 will be soon!

[TiloProbst]

Originally Posted by E.T from tellus

Party time!!!!!!
top 4 -> all Macs

indeed. and a macguardians.de news a few days ago (german):

http://www.macguardians.de/index.php?p=4088

hey who is that Anonymous owner of a Top10 Quad? reveal yourself

[halimedia]

Originally Posted by TiloProbst

hey who is that Anonymous owner of a Top10 Quad? reveal yourself

It's me - and top-5 should be happening today, I hope

[TiloProbst]

uhm, so why do you appear as Anonymous, but on your previously linked 'Swiss Crunchers' statistics page you were halimedia?

[halimedia]

Originally Posted by TiloProbst

uhm, so why do you appear as Anonymous, but on your previously linked 'Swiss Crunchers' statistics page you were halimedia?

That's because I've set my SETI preferences at 'Should SETI@home show your computers on its web site?' to 'No'. This setting has no bearing on the public availability of a user's total stats, which is why you've seen my stats at the BOINC Synergy site (and TeamMacNN, SETI User stats itself, etc.).

HTH,

Ron

[[email protected]]

I'm a new MacNN forum user. I read with interest about the enhanced SETI worker programs optimized for the G4 and G5's.

I am a Boinc menubar user at the moment. Can I run these enhanced SET worker programs under the Boinc menubar program somehow or are these for the command line or another Boinc manager only? I'm looking for instructions as to how to run the enhanced SETI workers.

Thanks for any help you can provide.

[halimedia]

Originally Posted by [email protected]

I am a Boinc menubar user at the moment. Can I run these enhanced SET worker programs under the Boinc menubar program somehow or are these for the command line or another Boinc manager only? I'm looking for instructions as to how to run the enhanced SETI workers.

See my post about 1/3 of the way down on page 7 - the instructions are geared towards BOINC Menubar specifically. But you don't need to follow all the steps, since you already have BOINC Menubar installed. You'll want to make sure to use a Superbench version of BOINC Menubar (just trash your original BOINC Menubar.app and replace it with the new one), and you can simply stick the optimized worker in the place mentioned in the instructions.

HTH,

Ron

Hi,

I have a question regarding the performance of the Quad.
There is a Dual G5 2Ghz (2Gb Ram) and a Quad (6.5Gb Ram) running on my account.
Both under 10.4.4 with 5.2.13 BoincManager superbench and G5-a5-worker.
My Dual is performing as expected

http://setiathome.berkeley.edu/show_...hostid=1621209

but the Quad seems a bit slow with only 15720.45 million ops/sec

http://setiathome.berkeley.edu/show_...hostid=2118879

from what I read it should be well within the 22000 million ops/sec ?

Energie saver is set to highest performance.

any ideas ?

Thanks in advance

[virex]

is the a5 a lot faster than the a4? is there a binary of a5 or do i have to compile?

EDIT
wow i'm an idiot and apparently didn't check the beginning of this thread...well anyway, on a4 i was getting about 45 minutes per work unit.....lets see how much faster a5 is.

[Todd Madson]

Cardiel: see http://members.dslextreme.com/~reade...boincbeta.html and get the superbench version of the software there (5.2.13).

The version at Seti produces considerably worse results than the superbench version.

My DP 2.5 rates 6628.73 million ops/sec fp and 23584.2 million ops/sec integer with
the superbench version at the macnn page above.

If you read my post above, I was getting significantly worse results with the standard
version.

Originally Posted by Cardiel

with 5.2.13 BoincManager superbench and G5-a5-worker.

Thats my problem I have the superbench version the fp numbers are correct but the int ones are low.

[alexkan]

No, I'm not posting to announce the release of alpha-6...however, I am still posting to ask for testing help. I'm looking for more data regarding whether or not using FFTW in our code is a good thing, performance-wise. (As it turns out, vDSP is not as thoroughly optimized as we originally thought, at least for the FFT sizes that matter to SETI performance.)

How can you help with this? Download and decompress this file, run it in Terminal, and paste the program's output from "Apple Altivec DSP" on down in a forum post, along with as many specifics you can gather about what CPU you're running. Don't worry if the program seems to stop at FFTW3--it'll take more than a few minutes, but in a real client, this won't happen.

From what I can already tell, older (read: non-dual-core) processors, G4 and G5 alike, may stand to gain from the switch from vDSP to FFTW. Who knows, dual-core G5s may gain from this as well!

Originally Posted by alexkan

No, I'm not posting to announce the release of alpha-6...however, I am still posting to ask for testing help. I'm looking for more data regarding whether or not using FFTW in our code is a good thing, performance-wise. (As it turns out, vDSP is not as thoroughly optimized as we originally thought, at least for the FFT sizes that matter to SETI performance.)

How can you help with this? Download and decompress this file, run it in Terminal, and paste the program's output from "Apple Altivec DSP" on down in a forum post, along with as many specifics you can gather about what CPU you're running. Don't worry if the program seems to stop at FFTW3--it'll take more than a few minutes, but in a real client, this won't happen.

From what I can already tell, older (read: non-dual-core) processors, G4 and G5 alike, may stand to gain from the switch from vDSP to FFTW. Who knows, dual-core G5s may gain from this as well!

Hi Alex, I'll give it a run on my G4 dual 1.33 tomorrow and post the results. Glad to help. BTW, you see this? http://fftw.org/release-notes.html Thought this might interest you.
Greetings from San Diego.