Welcome to the MacNN Forums.

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

You are here: MacNN Forums > Community > Team MacNN > Enhanced Optimized

Enhanced Optimized (Page 4)
Thread Tools
Knightrider
Dedicated MacNNer
Join Date: Sep 2004
Location: London
Status: Offline
Reply With Quote
May 31, 2006, 03:00 AM
 
Here is my first fully validated result with V4. I am number 2427722

For some reason the wu has been issued seven times.

K.
     
cwasko
Senior User
Join Date: Jul 2000
Status: Offline
Reply With Quote
May 31, 2006, 07:44 AM
 
I keep getting stuff similar to this on v4: "Wed May 31 07:10:53 2006|SETI@home|Unrecoverable error for result 14ja99aa.6153.17921.517322.3.154_3 (process exited with code 2 (0x2))"

It does this for WUs that I download. I ran the fft_test_shell and copied the wisdom.sah to the seti projects dir as well. I have a Rev a PowerBook 17", 1 GHz G4, 2 GB RAM.

Edit: Also running 10.4.6
( Last edited by cwasko; May 31, 2006 at 07:55 AM. )
     
linsyloo
Fresh-Faced Recruit
Join Date: Aug 2006
Status: Offline
Reply With Quote
May 31, 2006, 10:12 AM
 
i've also been getting loads of these with v4 :

Wed 31 May 11:40:22 2006|SETI@home|Checksum or signature error for 23mr99aa.20172.4225.561076.3.89

Wed 31 May 11:39:54 2006|SETI@home|Unrecoverable error for result 03mr99ab.11329.13297.336072.3.108_3 (process exited with code 2 (0x2))

and am using a fft_test_shell generated wisdom.sah

this is the first problem i've had using optimised clients, v3 was runing nicely on my 1.25 mac mini 10.4.6 1 gig ram

otherwise, the other clients rocked
     
Karl Schimanek
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Reply With Quote
May 31, 2006, 10:36 AM
 
Holy sh*t!

PPC970MP 2.5GHz (MacOS) faster as an Opteron 280 2.4GHz (Windows). V3 on PPC side and Crunch3r on x86.

8,668.18 vs. 9,401.09
http://setiathome.berkeley.edu/worku...?wuid=80003455

Keep on rockin'
Karl
     
arkayn
Dedicated MacNNer
Join Date: Aug 2005
Location: Golden Valley, AZ
Status: Offline
Reply With Quote
May 31, 2006, 10:44 AM
 
     
cwasko
Senior User
Join Date: Jul 2000
Status: Offline
Reply With Quote
May 31, 2006, 10:48 AM
 
Originally Posted by arkayn
Looks good. Unfortunately, I'm not even getting that far. As soon as the v4 client starts to process, it bails out immediately with the errors I noted above.
     
Knightrider
Dedicated MacNNer
Join Date: Sep 2004
Location: London
Status: Offline
Reply With Quote
May 31, 2006, 11:06 AM
 
Re:Error Messages.

Here is a url to couple of Wiki pages about error messages. Look under "U', a couple of options shown. Of course none of these may be the right ones for your error instance.Take a look in the stderrdae.txt to see if there is any additional info.

http://boinc-wiki.ath.cx/index.php?t...:BOINC_Message

K.
     
cwasko
Senior User
Join Date: Jul 2000
Status: Offline
Reply With Quote
May 31, 2006, 12:45 PM
 
Ok. I feel like a dope. I was suspending Boinc and then starting it again. Even after clearing out WU's etc. I then killed Boinc, and life is good. I've been through the process before so I'm not sure what the heck I was thinking...

I'll let ya'll know how it goes when I actually process some WU's.
     
BTBlomberg
Forum Regular
Join Date: Sep 2005
Location: Chicago Suburbs
Status: Offline
Reply With Quote
May 31, 2006, 03:12 PM
 
Here is my first WU with v4 only. It took 55298.94 sec. (15.36 hours) on the PowerBook G4 500 X 10.4.6. My Previous 1/2 v3 / 1/2 v4 WU took 63546.70 sec. (17.65 hours), so this is at least 2.25+ hours of improvement there, maybe 4.5 when you consider 1/2 & 1/2 on the WU.

I expected it to be faster, but each WU can be a little different it seams. More so than with the old SETI. I will see with more WUs, but I am gald about the improments so far. A separate G4 & G5 version may help next round.

In aditional observations of SETI Enhanced, it seams the first 1/2 of a WU processes slower than the second half. The opposite seamed to be true of the old SETI, but it may be based on what tasks are more optimized than others and the order of the processing tasks, but I don't know.
     
Todd Madson
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status: Offline
Reply With Quote
May 31, 2006, 03:27 PM
 
Here's one for the DP 2.5 with Alex's v4:
http://setiathome.berkeley.edu/resul...ltid=330299597

About 12,588 seconds. And a tad more credit than actually claimed.

RAC is now about 1100, down from around 2K on that machine.
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Reply With Quote
May 31, 2006, 04:58 PM
 
Originally Posted by BTBlomberg
As of Ver 3 he has a lot in the "strerr out". Here is an example from a failed WU:
Code:
<core_client_version>5.4.9</core_client_version> <message> Maximum disk usage exceeded </message> <stderr_txt> ar=0.422310 NumCfft=72953 NumGauss= 464507884 NumPulse= 89036768383 NumTriplet= 7691545591808 ar=0.422310 NumCfft=72953 NumGauss= 464507884 NumPulse= 89036768383 NumTriplet= 7691545591808 Crashed executable name: seti_enhanced-ppc-v3 built using BOINC library version 5.5.0 Machine type PowerPC 7400 System version: Macintosh OS 10.4.6 build 8I127 Sun May 28 22:04:06 2006 Stack frame backtrace: # Flags Frame Addr Caller PC Return Address Symbol === === ========== ========== ===================== 1 --- 0xf0100560 0x000d5374 MoreBacktraceMachSelf + 0x12c 2 --- 0xf01005e0 0x000a81e8 PrintBacktrace + 0x6c 3 --- 0xf0101930 0x00084dd8 _Z26handle_process_control_msgv + 0x110 4 --- 0xf0101d80 0x00084f00 _Z12worker_timeri + 0xc0 5 --- 0xf0101de0 0x0008512c _Z12timer_threadPv + 0x30 6 --- 0xf0101e30 0x9002ba68 </stderr_txt>
It does not reflect the FFT info though.
I can confirm that neither v3 nor v4 include any such output by default. This apears to be debugging output (noticed 'Crashed executable name...'?) that Alex decided to write to stderr_txt. This information does not appear in the stock worker.

HTH,

Ron
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Reply With Quote
May 31, 2006, 05:06 PM
 
Originally Posted by Gecko_r7
Halimedia: Have you seen the description we're referring in v3 or v4 WU report summary? I have not.
Neither have I - see my post above.
Another reason that it will be helpful is to better identify and compare processing times and differences between different AR wus and ap versions. Reported crunch times can be highly variable w/ Enhanced (and confusing if using opt aps) because WUs can be of substantially different length as described by the WU's AR.
Example: (not actual WUs)
v3 WU 18,000 secs, AR= .417
v4 WU 18,000 secs AR = .324 (on v3, time = 21,000)
Please excuse my ignorance, but what does AR stand for and how is it calculated?
     
adream
Junior Member
Join Date: Jun 2005
Status: Offline
Reply With Quote
May 31, 2006, 05:18 PM
 
sofar so good here on my dual 2.7 G5:

http://setiweb.ssl.berkeley.edu/resu...hostid=2406731

the last 23 results or so are with v4

no errors so far, and the 60ish credit units are coming in between 10-11000 seconds which is stonking fast as far as I'm concerned

many thanks alex for your time and effort, much appreciated, also a vote of thanks for boog too who put in some sterlin work in the early days

regards

adream
63. (1) (b) "music" includes sounds wholly or predominantly characterised by the emission of a succession of repetitive beats
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Reply With Quote
May 31, 2006, 05:20 PM
 
Status report v4: after crunching through more than 100 WUs on just about all flavors of G4 and G5, I haven't seen a single invalid result. Quality job, Alex!
     
liebsmaschine
Registered User
Join Date: Jul 2006
Status: Offline
Reply With Quote
May 31, 2006, 05:35 PM
 
Well, I finally found the reference WU, but I have not had a chance to run it yet. I've also had v4 running for about 30 hours now, so I've crunched several WUs. It appears that the *average* time per WU (yes, I know it's highly variable) has settled down to about 4-5 hours (see <a href="http://setiathome.berkeley.edu/results.php?userid=8344130">here</a>).

<a href="http://setiathome.berkeley.edu/workunit.php?wuid=80196983">Here's</a> an interesting one for speed comparison: I'm the third result down at 13,319 seconds. The first result, at 19,140 seconds, is a "PowerMac11,2" with a BOINC benchmark that's about double mine (some sort of G5 or G5 dual, maybe). I'm running a PowerBook G4 1.67, so it's cool to see that v4 on my machine is beating out the stock worker on a much faster machine.

I'm also consistently competitive with Intel processors running from 2-3GHz, although depending on the specific setup and if they're using optimized apps, mine sometimes takes the longest.

My PM 350 also just started crunching with v4. It has not yet finished a WU yet, though. I think it did get through one or two WUs with v2, but I haven't kept specific track of which WU was done by what worker. Also, S@H's results page seems to clear out old results much faster than it used to...any idea why?

All of my WUs for both v2 and v4 (I was about to install v3 when I saw that Alex had just posted v4) have validated--not a single failure. Congratulations, Alex!
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Reply With Quote
May 31, 2006, 06:09 PM
 
Here's a potentially interesting bit of food for thought: today, I have taken a 1.66 GHz Core Duo Mac mini into service, and it's been crunching SETI with stock worker and BOINC client for a number of hours now. Some comparative data of WUs that are worth around 62 credits:

Quad G5, stock worker: 15,500s
Quad G5, v4: 10,500s
CD Mac mini, stock worker: 13,600s

Impressive if not disturbing, I'd say. Ref-wu comparison will follow at some point soon.
     
TiloProbst
Junior Member
Join Date: Jul 2005
Status: Offline
Reply With Quote
May 31, 2006, 07:05 PM
 
Originally Posted by Gecko_r7
Reported crunch times can be highly variable w/ Enhanced (and confusing if using opt aps) because WUs can be of substantially different length as described by the WU's AR.
Example: (not actual WUs)
v3 WU 18,000 secs, AR= .417
v4 WU 18,000 secs AR = .324 (on v3, time = 21,000)
I'd like to have an explanation on those .xxx numbers too, since there is some comparison thread at planet3dnow.de mentioning comparison WUs with 0,42x AR.
     
beadman
Dedicated MacNNer
Join Date: Nov 2004
Location: Virginia
Status: Offline
Reply With Quote
May 31, 2006, 08:57 PM
 
Originally Posted by halimedia
Here's a potentially interesting bit of food for thought: today, I have taken a 1.66 GHz Core Duo Mac mini into service, and it's been crunching SETI with stock worker and BOINC client for a number of hours now. Some comparative data of WUs that are worth around 62 credits:

Quad G5, stock worker: 15,500s
Quad G5, v4: 10,500s
CD Mac mini, stock worker: 13,600s

Impressive if not disturbing, I'd say. Ref-wu comparison will follow at some point soon.
Here's another comparison for you. I have a loaner 2.0 GHz Mac Book Pro duo core running the stock BOINC and stock SETI (5.4.9 universal). http://setiathome.berkeley.edu/resul...4031&offset=20 and it's coming in around 12,000s on the average.

My iBook G4, 1.33 GHz, is coming in around 43,000s using the stock 5.4.9 universal SETI client. http://setiathome.berkeley.edu/resul...hostid=1360780

My PowerBook G4, 1.67 GHz, has finished one stock WU for 33,800 http://setiathome.berkeley.edu/resul...hostid=2041993. I just switched it over today to Alex's v4 after about 27% computed on a WU with the stock client. Switchover went very smoothly and no hiccups. This is the same computer that was getting around 5,000s per WU using Alex and Rick's G4-A5 client on the old SETI WU.

beadman
     
Gecko_r7
Forum Regular
Join Date: Oct 2005
Location: Las Vegas, NV
Status: Offline
Reply With Quote
May 31, 2006, 10:49 PM
 
Originally Posted by TiloProbst
I'd like to have an explanation on those .xxx numbers too, since there is some comparison thread at planet3dnow.de mentioning comparison WUs with 0,42x AR.
Hi Gents:
AR = Angle Range. I have just enough understanding to be dangerous, but am not confident that I can do the subject appropriate justice. Needless to say, the days of seeing consistent crunch times w/ only minor variation between WUs (except the occasional short/noisy WU) are over. Enhanced is a different animal. WU completion time is heavily influenced by the AR of the WU and identical AR's can still differ in length, though they do trend to be highly similar. My best advice is to read a few of these links: Just as an FYI, I have been on Enhanced since the public rollout w/ my x86 rig and noticed pretty quickly that ARs in the vicinity of .417 were the most common. These threads mention same.
Personally, I find and catalog these specific .417-.421 ARs for my "sample" group and average the times to represent an expected time of the most typical WU.

http://setiathome.berkeley.edu/forum...d=30625#305440
http://www.marisan.nl/seti/
http://www.teamlambchop.com/runtime303.htm
http://je2bwm.at.infoseek.co.jp/i04-whats-AR.html

Hope this helps.

BTW, v4 is working flawlessly. No validation issues, no errors.
Also, I'm happy to report that when comparing similar ARs with v4 on the G4 to Crunch3r's 5.12 ap on M-1.6 Banias, the perfomance gap has narrowed to only a 45% difference per WU (M being faster), down from 60% diff. w/ previous A5.3 and Crunch3r's best ap for PentM-SSE2....a 15% differential IMPROVEMENT vs. the best we had w/ Seti Std. All things being equal (and if my rigs are good enough representative examples...G5 should be even better yet as SSE3 on x86 is worth only 2-3% more real-world), it appears PPC is now BETTER optimized than current x86 ap. Way to go Alex!

Can anyone else confirm/invalidate this generalization with what you're seeing on your x86s and PPCs? Cheers!
( Last edited by Gecko_r7; Jun 1, 2006 at 12:51 AM. )
     
sdubz  (op)
Junior Member
Join Date: Jun 2006
Status: Offline
Reply With Quote
Jun 1, 2006, 05:58 AM
 
Originally Posted by Gecko_r7
Can anyone else confirm/invalidate this generalization with what you're seeing on your x86s and PPCs? Cheers!
Looking thru the top computers is starting to look like the G5 may be faster now.

http://setiathome.berkeley.edu/top_hosts.php

It also looks like Bob Palmer is starting to climb back up again, and looking thru his results quickly, it looks like he is the fastest. http://setiathome.berkeley.edu/show_...p?userid=24298
     
Karl Schimanek
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Reply With Quote
Jun 1, 2006, 09:45 AM
 
     
TiloProbst
Junior Member
Join Date: Jul 2005
Status: Offline
Reply With Quote
Jun 1, 2006, 10:56 AM
 
thx for the links @gecko

finally, the wisdom files:
for Single 1,8 GHz G5 / 512 MB RAM / OS 10.4.0 / 900 MHz FSB (Summer 2003)
LINK

for Dual 2,0 GHz G5 / 1024 MB RAM / OS 10.4.6 / non Dual Core
LINK
     
Karl Schimanek
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Reply With Quote
Jun 1, 2006, 11:40 AM
 
Here another interessting one:
http://setiathome.berkeley.edu/worku...?wuid=80292816

Intel P4 3.4GHz (Crunch3r/Win): 8,341.75 sec.
PPC970MP 2.5GHz: (v4/MacOS): 8,925.02 sec.
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Reply With Quote
Jun 1, 2006, 11:55 AM
 
A good place to do performance comparisons is the SETI top computers listing:

http://setiweb.ssl.berkeley.edu/top_hosts.php

If you ignore #1 (this is most likely a fellow who merged the RACs of many P4 boxes into one - his RAC has been dropping like a rock since he showed up at the top spot out of nowhere a few days ago), you'll see that only the fastest Pentium 4 and Pentium D boxes are able to get close to or beat a current G5 quad crunching with v4.
     
BTBlomberg
Forum Regular
Join Date: Sep 2005
Location: Chicago Suburbs
Status: Offline
Reply With Quote
Jun 1, 2006, 12:27 PM
 
Originally Posted by halimedia
I can confirm that neither v3 nor v4 include any such output by default. This apears to be debugging output (noticed 'Crashed executable name...'?) that Alex decided to write to stderr_txt. This information does not appear in the stock worker.
Yes, after seeing regular WUs from v4 I saw this. Hopefully it will be added like this next round. As said in old SETI Optimized thread the PC Clients have their Sig and Info in it and the Error output does so we know it can be done.

I now have a second full v4 WU that completed in 55,098.07 sec. (15.305 Hours) so that may be the time for this machine (PBG4 500) with v4.
     
alexkan
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Reply With Quote
Jun 1, 2006, 01:05 PM
 
Own a G5? Have Xcode installed? Know how to use Terminal? If so, I need your help!

I've been having trouble profiling reliably over SSH (nothing sucks more than having a 2+ hour profiling run die without any data recorded), so if you can gather some traces for me, it will make it that much easier for me to make the client faster. PM me for instructions and links to the relevant files. As an added bonus, you'll get to test what will eventually become v5.

Here's a status update, though. There's a bunch of small things that I can tweak that will provide an incremental performance improvement on both platforms. Of these, one is painful, and the rest should be simple enough for me to do. On the G4 side, there's room for a sizable performance gain if I make a special compile, but it's less clear to me whether or not this is true of G5s as well. (This is why I need your help, after all, since I'm finding that some optimization flags actually make the code slower)

I will try to get some nicer Crunch3r-style output to stderr in the next release, since you all seem to be clamoring for it. Hopefully v5 will be done enough and fast enough for me to feel confident about taking a break from optimization for a while, since I start work on Monday and probably will be too tired to put any real time into optimization.
     
Todd Madson
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status: Offline
Reply With Quote
Jun 1, 2006, 02:52 PM
 
Alex: I own a G5 (2.5 dual), have Xcode installed and know how to use terminal.

But I'm no programmer, especially not these days.

Let me know what I can do to help!
( Last edited by Todd Madson; Jun 1, 2006 at 08:07 PM. )
     
adream
Junior Member
Join Date: Jun 2005
Status: Offline
Reply With Quote
Jun 1, 2006, 04:37 PM
 
pmed you alex about volunteering my dual 2.7 g5

regards

adream
63. (1) (b) "music" includes sounds wholly or predominantly characterised by the emission of a succession of repetitive beats
     
liebsmaschine
Registered User
Join Date: Jul 2006
Status: Offline
Reply With Quote
Jun 1, 2006, 06:25 PM
 
I found an interesting way to compare v4 to the stock worker without a whole lot of hasslese : compare my S@H results with my own S@H Beta results--the Beta project is still crunching with the stock 5.13 worker (the newer Astropulse will be released to the Beta team in the next couple of weeks, hopefully--if you're viewing this more than a few weeks after I've posted it, beware that when Astropulse comes out, the results for the S@H Enhanced worker *may* be erased).

So, here are my:

<a href="http://setiathome.berkeley.edu/results.php?userid=8344130">S@H results</a>
<a href="http://setiathome.berkeley.edu/beta/results.php?userid=1977">S@H Beta results</a>

(The results that took an insanely long time are from my G4 350 desktop--there are only a couple.)

The Beta WUs seem to be a little larger but more consistent (averaging 55 credits) than the widely-ranging regular WUs. But, here's a comparison of a few close WUs:

standard: 60.72 credits; 24,772 seconds
standard: 61.27 credits; 20,622 seconds
standard: 61.33 credits; 20,701 seconds
Beta: 58.50 credits; 32,804 seconds
Beta: 57.88 credits; 32,234 seconds
Beta: 57.88 credits; 32,692 seconds

I think the first standard WU I listed was an aberration, so I'd trust the second one (since it and the third one match up better, timewise). All three Beta WUs are close, so I'd say they're good for a comparison.

So, figuring the second standard WU versus the first Beta WU (and factoring in the difference in credits, although it's minimal), and if I did the math right (it wasn't my strongest subject), I figure that v4 is just under 38% faster than the stock worker on my PowerBook G4 1.67. (Someone with better math skills can probably verify my calculations.) Nice job, Alex! Looking forward to v5 and then some rest for you...
     
Anner
Fresh-Faced Recruit
Join Date: Aug 2006
Status: Offline
Reply With Quote
Jun 1, 2006, 09:47 PM
 
@alex
anything you want....
G5 Dual 2.5 (3 gig)
Mini Intel Duo (2gig ( from this week i hope))
Mini G4 (512)

Sorry:
Introduction...
Been looking at macnn for a long time, using its optimized versions,
Macnn seems to be the only active mac/seti-community....soooo...
finally...let's join this team, so here's another aprox. 200K of seti. enjoy.....

he whow, i'm number 21 in the stat's....
( Last edited by boerke; Jun 1, 2006 at 09:54 PM. )
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Reply With Quote
Jun 2, 2006, 07:24 AM
 
A few observations on WUs of differing 'value'. By value I mean how much credit they're worth per unit time spent crunching them. I've noticed that there are four broad categories of WUs in terms of credit claimed/granted:

Category 1: 14-19 credits
Category 2: 28-34 credits
Category 3: 45-52 credits
Category 4: 58-64 credits

While categories 1, 3 and 4 are roughly equivalent in 'value' (i.e. crunching time per credit), category 2 is significantly worse. Below some numbers from my quad (smaller numbers are more advantageous). Ymmv as there are architecture-specific differences in crunching efficiency, but the general trend will probably be observable on all machines.

Category 1: approx. 170 seconds per credit
Category 2: 200-270 seconds per credit
Category 3: approx. 175 seconds per credit
Category 4: approx. 170 seconds per credit

Note that WUs of category 2 are both more variable and less 'valuable' than those of the other three. These calculations are based on a relatively small number of samples, but I believe the overall observation to be valid. I have not tried comparing these observations to WU ARs.

FWIW,

Ron
     
Karl Schimanek
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Reply With Quote
Jun 2, 2006, 11:20 AM
 
Alex, what's about DTrace? Maybe a better option than Shark?

Regards
Karl
     
alexkan
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Reply With Quote
Jun 2, 2006, 01:26 PM
 
Originally Posted by Karl Schimanek
Alex, what's about DTrace? Maybe a better option than Shark?
Shark is already really good for what I do (it shows instruction scheduling for all the processors I'm targeting and allows me to see where dL1 and dL2 misses are happening), and I can't imagine DTrace being any better for the task, especially if it didn't start out on OS X.

Thanks to all who contacted me about getting Shark traces. I really only need one or two profiles on G5 to know where I need to work next. However, since these category 2 WUs earn credit more slowly than WUs in the other categories, if one of you can catch one of these WUs and profile them for me, that would be great. I'll send you the details on what profiles I need once you have one.
     
weimings
Fresh-Faced Recruit
Join Date: Jul 2006
Status: Offline
Reply With Quote
Jun 2, 2006, 03:49 PM
 
Hi,
good job
But I hope can be made more

These is my tests, in v4 I don't use specific wisdom.

QUAD

Hardware Overview:

Machine Name: Power Mac G5 Quad
Machine Model: PowerMac11,2
CPU Type: PowerPC G5 (1.1)
Number Of CPUs: 4
CPU Speed: 2.5 GHz
L2 Cache (per CPU): 1 MB
Memory: 2.5 GB
Bus Speed: 1.25 GHz
Boot ROM Version: 5.2.7f1

roy:~/Test/seti_test administrator$ time ./setiathome_5.13_powerpc-apple-darwin -nographics

real 192m20.927s
user 192m9.627s
sys 0m12.321s
roy:~/Test/seti_test administrator$

user+sys > real !?!

init_data.xml:
<wu_cpu_time>11519.701573</wu_cpu_time>

stderr.txt:
Can't set up shared mem: -1
ar=0.775000 NumCfft=58347 NumGauss= 304962714 NumPulse= 584441
96736 NumTriplet= 4979562790912

roy:~/Test/seti_test administrator$ time ./seti_enhanced-ppc-v4 -nographics
real 136m52.816s
user 136m44.371s
sys 0m7.215s
roy:~/Test/seti_test administrator$

<wu_cpu_time>8158.913671</wu_cpu_time>

% of time compared to standard 5.13

wu_cpu 70.8257%
Real 71.1625%
User 71.159%
Sys 58.5585%
User+Sys 71.1455%

DUAL

Hardware Overview:

Machine Name: Power Mac G5
Machine Model: PowerMac7,3
CPU Type: PowerPC G5 (3.0)
Number Of CPUs: 2
CPU Speed: 2 GHz
L2 Cache (per CPU): 512 KB
Memory: 1.5 GB
Bus Speed: 1 GHz
Boot ROM Version: 5.1.8f7

A first test without time:
<wu_cpu_time>16889.663244</wu_cpu_time>

gici:~/Test/seti_test administrator$ time ./setiathome_5.13_powerpc-apple-darwin -nographics

real 283m27.175s
user 281m8.752s
sys 1m26.637s
gici:~/Test/seti_test administrator$

<wu_cpu_time>16932.497229</wu_cpu_time>

gici:~/Test/seti_test administrator$ time ./seti_enhanced-ppc-v4 -nographics
real 184m51.092s
user 179m11.313s
sys 1m18.448s
gici:~/Test/seti_test administrator$

<wu_cpu_time>10783.098098</wu_cpu_time>

% of time compared to standard 5.13

wu_cpu1 63.8443%
wu_cpu 63.6828%
Real 65.2141%
User 63.735%
     
weimings
Fresh-Faced Recruit
Join Date: Jul 2006
Status: Offline
Reply With Quote
Jun 2, 2006, 04:08 PM
 
Originally Posted by alexkan
Own a G5? Have Xcode installed? Know how to use Terminal? If so, I need your help!
At work I have a DUAL G5 2GHz, Xcode installed and know how to use Terminal.
But I have few time and my English is bad :-(

Let me know what I can do to help!

Originally Posted by alexkan
I've been having trouble profiling reliably over SSH (nothing sucks more than having a 2+ hour profiling run die without any data recorded),
You have tried with "screen"?
     
Thanar
Junior Member
Join Date: Feb 2002
Location: Kozani, Greece, EU
Status: Offline
Reply With Quote
Jun 2, 2006, 05:00 PM
 
Had this one for quite sometime now... wisdom file for iBook G4 @ 1.2.

The client works like a charm, btw... What's more, the first WU that was uploaded when running alex' v4 worker was partially done by boog's and then by the v3 worker and it validated fine... Nice...

Keep up the good work, everybody!
     
BTBlomberg
Forum Regular
Join Date: Sep 2005
Location: Chicago Suburbs
Status: Offline
Reply With Quote
Jun 2, 2006, 07:19 PM
 
Thanar,

You are right, all acted as if one in crunching for me too. Far cry from the old SETI days, Change the Client, blow a WU. All are based on same client version though, so that's why. It appears the WUs are attached to the cleint version, so when I switched my PCs to Crunch3r's 5.12 version from 5.15 the downloaded WUs all got flushed out and new ones downloaded.

Speaking of Crunch3r's app and his departure from SETI@Home because of many issues with people, I want to express my gratitude to Alex and Boog as well as all others like Rick and the folks here that do their part to help them optimize this client to it's full (or near full) potential.
     
Fortilan
Fresh-Faced Recruit
Join Date: Feb 2006
Location: Sydney, AUS
Status: Offline
Reply With Quote
Jun 3, 2006, 07:22 PM
 
Hi all,

I have the v4 enhanced worker running on my Mac Mini with a wisdom file supplied by boog on this forum. I used to run the same before the enhanced workers came along and it would indicate that the wisdom file was being used in the output, now I dont see anything.

my reading of a post said that this might be fixed in v5? or maybe I just have it set up incorrectly.

can someone document the right setup for v4 i.e. all the files, locations and contents? I think it would help me (and others?) to know we are running everything OK. My setup is running under the SETI/Boinc supplied Boinc Manager.

Also, I am not yet running v4 on my G5 iMac. Is it possible to run v4 on a G5 as well, or is it only for a G4? The standard enhanced worker and Boinc Manager running on my G5 is painfully slow. Sometimes I think my G4 mini is doing better than the iMac G5

Any help appreciated.

lastly, I would like to join in the thank you's to the people who have supported MacNN 'crunchers' with the many innovative SETI workers to keep us *ahead* of those poor people with PC's

*** update: I have started running v4 on my iMac G5 1.8 Ghz --- I'll see how it goes compared to the standard enhanced worker, thanks for the help
( Last edited by Fortilan; Jun 3, 2006 at 11:08 PM. )
     
Thanar
Junior Member
Join Date: Feb 2002
Location: Kozani, Greece, EU
Status: Offline
Reply With Quote
Jun 3, 2006, 07:40 PM
 
OF COURSE you should switch to v4 on the G5! What Alex said is that there are no G5-specific optimizations in v4 yet, but there are more than enough when compared to the stock worker.

Regarding the wisdom.sah file, it should be placed inside BOINC Data/projects/setiathome.berkeley.edu folder. Not sure if it does any good, though, since it seems like the workers create a somehow complete wisdom.sah file of their own when the custom one doesn't exists. That wasn't the case with classic BOINC S@H, so a well-compiled wisdom.sah file would do great difference!
     
BTBlomberg
Forum Regular
Join Date: Sep 2005
Location: Chicago Suburbs
Status: Offline
Reply With Quote
Jun 3, 2006, 10:13 PM
 
I have found the Wisdom file I created for my PowerBook G4 has been overwriten by the SETI/BOING app and it's different than the origional one that the stock setup created when I first installed BOINC 5.4.9 with stock SETI 5.13. Not sure why that happened.

I will try moving it back and see if it sticks. It almost looks like the program keeps rewriting it. Maybe it is a cumulative thing, give it that FFT3-test generated base and it alters things as it find new stuff out on it's own. From what I have read on the SETI forums though, the SETI team may not be that smart.

My iMac G3 333 with stock client also rewrites the wisdom all the time. Just ran FFT3-test on that poor kid to see if it would help it any.

That leads me to a question for Alex to weather you improved on any Mac code while in there that may help old G3s too. I have heard at the SETI forums that the new enhanced code still needs some basic work to it, but there are a lot of things being said their now so who knows. Some where complaints from Crunch3r about it having flaws he pointed out a long time ago.
     
alexkan
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Reply With Quote
Jun 3, 2006, 11:47 PM
 
Originally Posted by BTBlomberg
I have found the Wisdom file I created for my PowerBook G4 has been overwriten by the SETI/BOING app and it's different than the origional one that the stock setup created when I first installed BOINC 5.4.9 with stock SETI 5.13. Not sure why that happened.

I will try moving it back and see if it sticks. It almost looks like the program keeps rewriting it. Maybe it is a cumulative thing, give it that FFT3-test generated base and it alters things as it find new stuff out on it's own. From what I have read on the SETI forums though, the SETI team may not be that smart.

My iMac G3 333 with stock client also rewrites the wisdom all the time. Just ran FFT3-test on that poor kid to see if it would help it any.
This is a side effect of how the SETI code reads and writes wisdom. It will rewrite the wisdom file with whatever wisdom it has in memory, be that the wisdom you generated in advance or the wisdom it generated for itself on the fly. (The wisdom you generate will always supercede the on-the-fly wisdom, though.) For some reason, the individual values change, but the performance remains the same.
That leads me to a question for Alex to weather you improved on any Mac code while in there that may help old G3s too. I have heard at the SETI forums that the new enhanced code still needs some basic work to it, but there are a lot of things being said their now so who knows. Some where complaints from Crunch3r about it having flaws he pointed out a long time ago.
The SETI code does a lot of quirky things, that's for sure. I wound up fixing some of them in order to get around bottlenecks they were creating for the PPC architecture (like a bunch of unnecessary type conversions). However, it'd take a little bit of reconfiguration before I could get a G3-compatible compile up and running, and at that point, I would no longer be sure if a lot of the optimizations I put would actually be of any help, since I know almost nothing about the G3's performance characteristics.
     
BTBlomberg
Forum Regular
Join Date: Sep 2005
Location: Chicago Suburbs
Status: Offline
Reply With Quote
Jun 4, 2006, 01:18 AM
 
Thanks Alax, on both counts.

I tried v4 on my G3 and as I expected it faultered. Likely could not find the addresses or instructions to run right. Not a big suprise. I don't expect an optimized version specifically for G3, just hoped some core fixes would help it out.
     
liebsmaschine
Registered User
Join Date: Jul 2006
Status: Offline
Reply With Quote
Jun 4, 2006, 04:32 AM
 
I ran the reference WU three times: once with the stock worker, once with v4, and once with v4 with the custom-built wisdom file. Here are the results. I used "time ./[worker app name] -nographics -standalone -verbose &".

Machine Name: PowerBook G4 17"
Machine Model: PowerBook5,7
CPU Type: PowerPC G4 (1.2)
Number Of CPUs: 1
CPU Speed: 1.67 GHz
L2 Cache (per CPU): 512 KB
Memory: 1 GB
Bus Speed: 167 MHz
Boot ROM Version: 4.9.1f1

1) STOCK
real 457m59.500s
user 443m53.667s
sys 3m15.059s
init_data.xml wu_cpu_time: 43859.885467
stderr.txt: ar=0.775000 NumCfft=58347 NumGauss= 304962714 NumPulse= 58444196736 NumTriplet= 4979562790912

2) V4
real 292m67.061s
user 291m38.086s
sys 2m42.570s
init_data.xml wu_cpu_time: 17056.853851
stderr.txt: ar=0.775000 NumCfft=58347 NumGauss= 304962714 NumPulse= 58444196736 NumTriplet= 4979562790912

3) V4+WISDOM
real 286m22.711s
user 276m57.943s
sys 2m36.918s
init_data.xml wu_cpu_time: 33816.762545
(My thinking on this is that I must havecopied the whole test folder from the #2 test intact. I believe I read somewhere that the init_data.xml WU time counter is incremented from any preexisting number, so the init_data.xml file must have already been in there. Therefore, if I subtract the 17,056 seconds from Test #2, I get 16759.9087 seconds for the third test.)
stderr.txt: ar=0.775000 NumCfft=58347 NumGauss= 304962714 NumPulse= 58444196736 NumTriplet= 4979562790912

I actually ran the V4 test without the wisdom file on accident at first, and then I decided to run it with the wisdom file to see how much the wisdom file really did for it. Then, I read BTBlomberg's post, and looked at the wisdom file in the test directory and found that it was indeed different than my original custom-built wisdom file (the first line of the custom one says "fftw-3.1-beta2 fftwf_wisdom," whereas the first line of the one I found in the test directory says "fftw-3.1.1 fftwf_wisdom"). Perhaps it was overwritten by the client, or perhaps I was careless in my Terminal work and copied the wrong wisdom file. Or, perhaps I accidentally copied the wisdom.sah file from the #2 test--my eyes are too blurry to compare those two to see if they're the same (although the first lines of them are). In any case, I'm much too in need of my computer to run another test WU (to be safe, I was running these in text-only console mode, just to make sure that VM pages, etc., wouldn't slow the worker down). Someone else (preferably with a faster machine that crunches WUs in less than an hour...) should try the reference WU with and without the custom wisdom file and see what sort of difference it really makes.

Oh, all right, in the interest of my own curiosity, I'll probably run it again, this time ensuring that the proper wisdom.sah file is copied (and that the init_data.xml file is fresh and zeroed out prior to beginning)...maybe tomorrow while I'm running some errands. I'll edit this post and put the result below.
( Last edited by jackal; Jun 4, 2006 at 05:27 AM. Reason: Hit submit accidentally)
     
liebsmaschine
Registered User
Join Date: Jul 2006
Status: Offline
Reply With Quote
Jun 4, 2006, 05:30 AM
 
OK, I couldn't help myself and started to run another test. This time, I made sure that I copied the correct wisdom file, and I zeroed out the timer on the init_data.sah file.

A minute or so after I started, I decided to check: sure enough, the wisdom file had been overwritten/modified (the modification date was newer, and the first line changed to read fftw-3.1.1 instead of 3.1-beta2.

I aborted the test run and am back to my normal BOINC crunching.

The wisdom file in my BOINC installation of the v4 worker is my custom-created one--it's only when I run it standalone that it generates a new one on top of the existing one.

Anyway, hope the above info helped you, Alex. G'night!
     
liebsmaschine
Registered User
Join Date: Jul 2006
Status: Offline
Reply With Quote
Jun 5, 2006, 12:49 AM
 
<a href="http://setiathome.berkeley.edu/result.php?resultid=335561196">Here's</a> an interesting WU. It did not validate with the v4 client (running on my desktop G4 350). It's an extremely small WU (the <a href="http://setiathome.berkeley.edu/workunit.php?wuid=80454441">other two who have completed it</a> so far both requested similar credit and spent a small amount of time on it, so the low credit request is not due to it aborting early. Weird...
     
BTBlomberg
Forum Regular
Join Date: Sep 2005
Location: Chicago Suburbs
Status: Offline
Reply With Quote
Jun 5, 2006, 11:11 AM
 
I think the Out of Space Error is a somewhat common one. I almost think it may be a Bug in the main app, but after optimization it may stand out more in some cases. I got a few of those before with different versions of Enhanced, but all others also failed. I think it may be kin to the Overflow error of the previous version of SETI.

I get the fealing it may be interpreted as the app has detected more points of interest (gausians/peaks/etc.) than the DB has allocated for. I may be wrong, but I saw a link to a list of errata with SETI Enhanced over at the SETI forums and it was pointed to after a post listing an error line this. I did not follow the link though. Heck, it may have been here and this tread for all I know.
     
linsyloo
Fresh-Faced Recruit
Join Date: Aug 2006
Status: Offline
Reply With Quote
Jun 5, 2006, 11:50 AM
 
Originally Posted by cwasko
Ok. I feel like a dope. I was suspending Boinc and then starting it again. Even after clearing out WU's etc. I then killed Boinc, and life is good. I've been through the process before so I'm not sure what the heck I was thinking...

I'll let ya'll know how it goes when I actually process some WU's.
ok, so i saw this post and tried doing the same, and now v4 is zooming along with no problems. i have my little mac mini fan roaring like a jet engine again
     
Knightrider
Dedicated MacNNer
Join Date: Sep 2004
Location: London
Status: Offline
Reply With Quote
Jun 5, 2006, 12:09 PM
 
Originally Posted by jackal
<a href="http://setiathome.berkeley.edu/result.php?resultid=335561196">Here's</a> an interesting WU. It did not validate with the v4 client (running on my desktop G4 350). It's an extremely small WU (the <a href="http://setiathome.berkeley.edu/workunit.php?wuid=80454441">other two who have completed it</a> so far both requested similar credit and spent a small amount of time on it, so the low credit request is not due to it aborting early. Weird...
Information on the -9 error is HERE

I have had results where I validate and the others did not, failing with this message. Also the other way around.

K.
     
rhettmaxwell
Fresh-Faced Recruit
Join Date: Jun 2006
Status: Offline
Reply With Quote
Jun 5, 2006, 01:27 PM
 
Look at this WU!

http://setiathome.berkeley.edu/worku...?wuid=80752856

My 2.5 GHz G5 is the one with 1,952 seconds. I wish, ALL would be like this.
     
Knightrider
Dedicated MacNNer
Join Date: Sep 2004
Location: London
Status: Offline
Reply With Quote
Jun 5, 2006, 01:57 PM
 
Originally Posted by Bad to the bone
Look at this WU!

http://setiathome.berkeley.edu/worku...?wuid=80752856

My 2.5 GHz G5 is the one with 1,952 seconds. I wish, ALL would be like this.
OMG de dah de dah de dah

K.
     
 
Thread Tools
 
Forum Links
Forum Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Top
Privacy Policy
All times are GMT -4. The time now is 12:59 AM.
All contents of these forums © 1995-2017 MacNN. All rights reserved.
Branding + Design: www.gesamtbild.com
vBulletin v.3.8.8 © 2000-2017, Jelsoft Enterprises Ltd.,