Windows S5R4 App 6.06 available for Beta Test

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

RE: RE: I suspended

Message 88705 in response to message 88704

Quote:
Quote:

I suspended becuase I'm running the AP opti app on a single core pc with the new Einstein Beta, and I was afraid that if there's no working checkpointing, I could be wasting my time with them because of the switching back and forth between the tasks on the pc. So I thought it would be better for me if I waited for an update with checkpointing.

If you leave the app in memory while suspended and the system doesn't crash or reboot or BOINC restart before WU completion, you should be OK.

A risk, I know. But if you've got a stable system and you don't mess around start/stopping BOINC, you should be able to get away with it.


I'll just suspend the AP till the Einstein completes. Long as the power doesn't go out I should be fine, and now the rains have stopped here in Southern Cal there's less chance of that happening. I'll try doing the switching between tasks if/when the new update comes

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5874
Credit: 118177420207
RAC: 23889877

RE: There shouldn't be any

Message 88706 in response to message 88702

Quote:

There shouldn't be any difference in speed comparing to the 6.05 if the 6.05 ran on your computer.

I'd be curious to know how the 6.04_0 compares to the 6.06_0 and the 6.04_1 to 6.06_1 (does s.o. get what I mean by this?).

BM

I'm not sure who s.o. is - significant other perhaps? :-). I know my s.o. wouldn't have a clue about anything to do with even the basics of this project, let alone the arcane differences between _0, _1, or _2 executables. :-).

I guess s.o. might be someone and someone being a very casual reader of these boards is bound not to be fully up to speed with the differences you allude to. So a quick summary:-

The Windows 6.04 version consisted of a "switcher app" which would determine the level of SSE instructions that any particular host was capable of and then run the appropriate executable that could use those instructions. The three executables (_0, _1, _2) could handle "no SSE", "SSE only", or "SSE2 and above" respectively.

After further optimisation of the Windows app, the 6.05 version was made available but only as an "_2" executable. Those of us with SSE only hosts and below were left out in the cold. :-). With the release of 6.06, once again there are the full range of SSE capabilities being catered for.

So, for someone who has an SSE2+ capable host that has been running the 6.05 executable, there should be no further speed improvement by running 6.06. On the other hand, anyone still running any of the 6.04 executables should see a very nice improvement by switching to the 6.06 Beta test suite.

However, reading between the lines, it looks like the old adage of fixing one bug only to see a fresh crop of new ones appear, seems to have come true yet again so perhaps 6.06 will be rather short lived. I have already converted some SSE hosts running Windows to the 6.06 app. If we have to go to 6.07, it would be good to have this confirmed fairly quickly.

Cheers,
Gary.

Novasen169
Novasen169
Joined: 14 May 06
Posts: 43
Credit: 2767204
RAC: 0

Indeed. No speedup compared

Indeed. No speedup compared to 6.05, getting the same errors. I've put resource share at 50% (on a dual core) and told it not to get new work. Hopefully we'll see 6.07 soon :)

John Clark
John Clark
Joined: 4 May 07
Posts: 1087
Credit: 3143193
RAC: 0

RE: Indeed. No speedup

Message 88708 in response to message 88707

Quote:
Indeed. No speedup compared to 6.05, getting the same errors. I've put resource share at 50% (on a dual core) and told it not to get new work. Hopefully we'll see 6.07 soon :)

The ones I produced overnight seemed to have the same overall time as for the 6.05 client, which is what, I believe, is expected.

Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4330
Credit: 251394672
RAC: 35874

The checkpointing problem

The checkpointing problem apparently arises from compiling the BOINC library for Windows with gcc. I've seen these "couldn't rename" messages occasionally with 6.05, but apparently the change I made to fix this in 6.06 made things rather worse than better. I'll continue to work on this.

I think for most people that have SSE2 machines for the time being it would be best to step back to 6.05, I hope to get a 6.07 out over the weekend, but am not sure.

I'd still be curious about performance comparisons between equally featured Apps of the 6.04 vs. 6.06 packages, e.g. if someone could test 6.06 on an old Athlon that is not capable of SSE2.

I actually wonder if there are still non-SSE machines attached to Einstein@home.

BM

BM

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

RE: I'd still be curious

Message 88710 in response to message 88709

Quote:
I'd still be curious about performance comparisons between equally featured Apps of the 6.04 vs. 6.06 packages, e.g. if someone could test 6.06 on an old Athlon that is not capable of SSE2.


Ok I'll hook up this one and let you know how it does.

Edit: My wingman isn't gonna like me though (sorry bout that) with GenuineIntel
Intel(R) Core(TM)2 Duo CPU T7300 @ 2.00GHz, running Linux and only 1 task...They'll be waiting on me ;P

[AF>Futura Sciences]click
[AF>Futura Scie...
Joined: 12 Apr 05
Posts: 34
Credit: 1923040
RAC: 0

Hi Bernd ! RE: I'd

Message 88711 in response to message 88709

Hi Bernd !

Quote:

I'd still be curious about performance comparisons between equally featured Apps of the 6.04 vs. 6.06 packages, e.g. if someone could test 6.06 on an old Athlon that is not capable of SSE2.

I did a quick run of refWU for you.
On XP3800+ AMD system, FSB 210MHz, low latency mem (2-2-2-5), XP SP3...
HostID : 0963368

Running only the ref WU, priority real-time.
Many runs with 6.04, but only had time for one on each 6.06. So i need to ensure that later.

6.04=>
_0 - Up time : 20min 06s CPU time : 19min 57s
_1 - Up time : 13min 14s CPU time : 13min 11s

6.06=>
_0 - Up time : 17min 16s CPU time : 17min 12s
_1 - Up time : 11min 19s CPU time : 11min 16s
_2 - Up time : 11min 07s CPU time : 11min 01s

So compiler change seems to down the crunch time.
Still very small boost for SSE2 version on AMD XP?+ systems (we already knew).

Hope it helps.
Let me know if you need me to test anything else.

Will try to update my "follow up" of every app versions (XP and Linux), for crunch time on the ref WU, a little bit later.
Need to test back previous apps since i switched from SP1 to SP3 in the mean time. But the effect seems to be small on crunch time from what i saw with 6.04. a few percents only.

Quote:
The checkpointing problem apparently arises from compiling the BOINC library for Windows with gcc. I've seen these "couldn't rename" messages occasionally with 6.05, but apparently the change I made to fix this in 6.06 made things rather worse than better. I'll continue to work on this.

I can see it in the stderr.txt of the ref WU too:

Output from " einstein_S5R4_6.06_windows_intelx86_1.exe " run : 
2008-11-28 16:21:55.9218 [normal]: This program is published under the GNU General Public License, version 2
2008-11-28 16:21:55.9218 [normal]: For details see http://einstein.phys.uwm.edu/license.php
2008-11-28 16:21:55.9218 [normal]: This Einstein@home App was built at: Nov 27 2008 02:54:57

2008-11-28 16:21:55.9218 [normal]: Start of BOINC application 'HierarchicalSearch'.
Activated exception handling...
Can't open init data file - running in standalone mode
command line: HierarchicalSearch einstein_S5R4_6.06_windows_intelx86_1.exe --method=0 --Freq=1005.20242518 --FreqBand=0.0161398745876 --dFreq=6.71056161393e-06 --f1dot=-1.58548959919e-09 --f1dotBand=1.74403855911e-09 --df1dot=3.88447721545e-10 --skyGridFile=skygrid_1010Hz_S5R3.dat --DataFiles1=h1_1005.00_S5R3;l1_1005.00_S5R3;h1_1005.05_S5R3;l1_1005.05_S5R3;h1_1005.10_S5R3;l1_1005.10_S5R3;h1_1005.15_S5R3;l1_1005.15_S5R3;h1_1005.20_S5R3;l1_1005.20_S5R3;h1_1005.25_S5R3;l1_1005.25_S5R3;h1_1005.30_S5R3;l1_1005.30_S5R3;h1_1005.35_S5R3;l1_1005.35_S5R3 --tStack=90000 --nStacksMax=84 --pixelFactor=0.500 --nf1dotRes=1 --ephemE=earth_05_09 --ephemS=sun_05_09 --nCand1=10000 -o Hough.out --gridType=3 --useWeights=0 --printCand1 --semiCohToplist -d1
2008-11-28 16:21:55.9375 [debug]: Set up communication with graphics process.
2008-11-28 16:21:56.4375 [debug]: Reading SFTs and setting up stacks ... done
2008-11-28 16:22:18.8750 [normal]: INFO: Couldn't open checkpoint Hough.out.cpt
2008-11-28 16:22:18.8750 [debug]: Total skypoints = 31. Progress: 0,
$Revision: 1.131 $ REV:$Revision, OPT:6, SCVAR:9, SCTRIM:2, HOTVAR:4, HOTDIV:0, HGHPRE:7, HGHBAT:2
1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, c
15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 2008-11-28 16:33:11.8750 [CRITICAL]: ERROR: Couldn't rename
Hough.out.cpt.tmp: line:580, ferr:0, errno:17: File exists
c
done.
FPU status flags: COND_2 PRECISION
2008-11-28 16:33:12.6250 [normal]: done. calling boinc_finish(0).
called boinc_finish

Quote:

I actually wonder if there are still non-SSE machines attached to Einstein@home.

Sometimes i run a few oldies for fun, as some others do.
But there shouldn't be much real crunchers on non-SSE.

Cheers,
Alex

God created a few good looking guys.. and for the rest he put hairs on top..

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

Just had the first result of

Just had the first result of the 6.06 on my AMD64. Same CRITICAL]: ERROR message as the others. I expected the time to be about the same but there was a possible slight drop in performance from the previous opti, so I decided to do another.

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 2975287886
RAC: 787703

RE: I actually wonder if

Message 88713 in response to message 88709

Quote:

I actually wonder if there are still non-SSE machines attached to Einstein@home.

BM


Host 476676 is still plodding away!

John Clark
John Clark
Joined: 4 May 07
Posts: 1087
Credit: 3143193
RAC: 0

SSE came in with the Intel

SSE came in with the Intel P3s, mainly on the Coppermine range. That means they need to be older than a P3 now?

Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!

Comment viewing options

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