Welcome to MilkyWay@home

v0.18/v0.19 issues here

Message boards : Number crunching : v0.18/v0.19 issues here
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 . . . 6 · Next

AuthorMessage
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 11000 - Posted: 16 Feb 2009, 7:53:26 UTC - in response to Message 10999.  

Seems the server can be finicky about updating the application for some reason. I deleted the previous versions (instead of deprecating them), so hopefully it should send them out now.

Can you also delete the old old aplication as well?
MilkyWay@Home Old 1.22


Yeah, just cleaned it out of the database ;) Don't really need it there anymore.
ID: 11000 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
cwhyl

Send message
Joined: 11 Nov 07
Posts: 41
Credit: 1,000,181
RAC: 0
Message 11005 - Posted: 16 Feb 2009, 12:49:13 UTC

Funny thing that my AMD X2 @2600 MHz now is faster than a Q6600 @3300 MHZ, 20 minutes compared to 22 minutes with Linux 32-bit, how come?

64-bit Linux is crunching fine here and does them in 18 minutes on a Q6600 @3200 MHz.

ID: 11005 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Cori
Avatar

Send message
Joined: 27 Aug 07
Posts: 647
Credit: 27,592,547
RAC: 0
Message 11006 - Posted: 16 Feb 2009, 13:07:34 UTC

Needing around 18.5 minutes on my C2D lappy with v0.19 (64 bit).
Compared to the old standard app it's good (~36-40 minutes with 0.17) but still bad compared to the opti app (~7-8 minutes with opti 0.16). ;-)))
Lovely greetings, Cori
ID: 11006 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 11009 - Posted: 16 Feb 2009, 13:25:55 UTC - in response to Message 11006.  

Needing around 18.5 minutes on my C2D lappy with v0.19 (64 bit).
Compared to the old standard app it's good (~36-40 minutes with 0.17) but still bad compared to the opti app (~7-8 minutes with opti 0.16). ;-)))


I don't think the windows apps are compiled with any vectorization, so that's probably where you're getting the rest of the benefit from :( I wonder if we could compile the 64 bit windows app with -SSE2 or -SSE3, which probably would bring it pretty close to that.
ID: 11009 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Debs

Send message
Joined: 15 Jan 09
Posts: 169
Credit: 6,734,481
RAC: 0
Message 11010 - Posted: 16 Feb 2009, 13:28:59 UTC - in response to Message 11009.  

Needing around 18.5 minutes on my C2D lappy with v0.19 (64 bit).
Compared to the old standard app it's good (~36-40 minutes with 0.17) but still bad compared to the opti app (~7-8 minutes with opti 0.16). ;-)))


I don't think the windows apps are compiled with any vectorization, so that's probably where you're getting the rest of the benefit from :( I wonder if we could compile the 64 bit windows app with -SSE2 or -SSE3, which probably would bring it pretty close to that.


-SSE2 would guarantee it can still run on all 64-bit systems. -SSE3 excludes at least some early AMD 64-bit systems (I know, because I am running 2 of them!).
ID: 11010 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Cori
Avatar

Send message
Joined: 27 Aug 07
Posts: 647
Credit: 27,592,547
RAC: 0
Message 11011 - Posted: 16 Feb 2009, 13:42:24 UTC - in response to Message 11010.  

Needing around 18.5 minutes on my C2D lappy with v0.19 (64 bit).
Compared to the old standard app it's good (~36-40 minutes with 0.17) but still bad compared to the opti app (~7-8 minutes with opti 0.16). ;-)))


I don't think the windows apps are compiled with any vectorization, so that's probably where you're getting the rest of the benefit from :( I wonder if we could compile the 64 bit windows app with -SSE2 or -SSE3, which probably would bring it pretty close to that.


-SSE2 would guarantee it can still run on all 64-bit systems. -SSE3 excludes at least some early AMD 64-bit systems (I know, because I am running 2 of them!).

But SSE3 is quite a lot faster. ;-)
I don't know how many old 64-bit system are out there though.
If it's not just a very small minority maybe having two 64-bit apps would be a solution?
That would imply of course one has to download one of those apps manually but with a little How-To added it would be easy for most users I guess (installing opti apps is just the same. *grin*)
Lovely greetings, Cori
ID: 11011 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
cwhyl

Send message
Joined: 11 Nov 07
Posts: 41
Credit: 1,000,181
RAC: 0
Message 11012 - Posted: 16 Feb 2009, 13:44:53 UTC

Yeah, the Linux 64-bit could also have SSE2 optimizations.
ID: 11012 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 11013 - Posted: 16 Feb 2009, 13:47:37 UTC - in response to Message 11012.  

Yeah, the Linux 64-bit could also have SSE2 optimizations.


The linux 64-bit app does have SSE2. It's using: -O2 -msse2 -ftree-vectorize -funroll-loops
ID: 11013 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
cwhyl

Send message
Joined: 11 Nov 07
Posts: 41
Credit: 1,000,181
RAC: 0
Message 11014 - Posted: 16 Feb 2009, 13:54:45 UTC

Ah, OK
ID: 11014 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Gavin Shaw
Avatar

Send message
Joined: 16 Jan 08
Posts: 98
Credit: 1,371,299
RAC: 0
Message 11015 - Posted: 16 Feb 2009, 14:17:30 UTC

Just added my wife's laptop back to MW and it is using the stock MW 0.19 Win32 app.

First unit took 21:17 and granted 31.56 credits. This is running on a AMD Mobile Sempron 3100+.

Never surrender and never give up. In the darkest hour there is always hope.

ID: 11015 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Labbie
Avatar

Send message
Joined: 29 Aug 07
Posts: 327
Credit: 116,463,193
RAC: 0
Message 11016 - Posted: 16 Feb 2009, 14:23:39 UTC

Switched over my Q9450 @ 3.36GHZ and it ran the first 4 v0.19 WUs in 832 to 837 seconds and still hit the limiter. They were granted from 24.97 to 25.12 credits.

Can't wait to see what this box will do with an opti app (Win32 please) and no limiter.

And my i7 parts come in today. Woo Hoo!!!


Calm Chaos Forum...Join Calm Chaos Now
ID: 11016 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile speedimic
Avatar

Send message
Joined: 22 Feb 08
Posts: 260
Credit: 57,387,048
RAC: 0
Message 11021 - Posted: 16 Feb 2009, 15:02:14 UTC

just posted the new v18d Linux-apps in the code forum...

mic.


ID: 11021 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Brian Silvers

Send message
Joined: 21 Aug 08
Posts: 625
Credit: 558,425
RAC: 0
Message 11025 - Posted: 16 Feb 2009, 15:12:27 UTC - in response to Message 11013.  

Yeah, the Linux 64-bit could also have SSE2 optimizations.


The linux 64-bit app does have SSE2. It's using: -O2 -msse2 -ftree-vectorize -funroll-loops


You should check with Bernd Machenschalk (Einstein@Home) in regards to doing a wrapper that has CPU feature detection...
ID: 11025 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Cluster Physik

Send message
Joined: 26 Jul 08
Posts: 627
Credit: 94,940,203
RAC: 0
Message 11026 - Posted: 16 Feb 2009, 15:34:20 UTC - in response to Message 11009.  

Needing around 18.5 minutes on my C2D lappy with v0.19 (64 bit).
Compared to the old standard app it's good (~36-40 minutes with 0.17) but still bad compared to the opti app (~7-8 minutes with opti 0.16). ;-)))


I don't think the windows apps are compiled with any vectorization, so that's probably where you're getting the rest of the benefit from :( I wonder if we could compile the 64 bit windows app with -SSE2 or -SSE3, which probably would bring it pretty close to that.

Visual Studio does not support auto vectorization (which gives close to 100% gain on Core2/i7 and K10 class CPUs). Therefore, another compiler would definitely help for builds using SSE2.
ID: 11026 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ebahapo
Avatar

Send message
Joined: 6 Sep 07
Posts: 66
Credit: 636,861
RAC: 0
Message 11027 - Posted: 16 Feb 2009, 15:52:15 UTC - in response to Message 11011.  

But SSE3 is quite a lot faster. ;-)
I don't know how many old 64-bit system are out there though.

Per my off-the-cuff analysis here, only 17% of hosts support SSE3. However, I couldn't break this figure out between 32 and 64-bit hosts.

HTH

ID: 11027 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 11028 - Posted: 16 Feb 2009, 16:16:47 UTC
Last modified: 16 Feb 2009, 16:21:20 UTC

Just swapped from the opti 0.16 client to the new stock 0.19 client, and first 4 away but I couldn't get the accurate figure for time or credit due to insta-purge.

I assume, like Labbie, it is hitting the credit limiter, but the 0.19 is fast compared to the previous stock client (0.17).

On 0.17 my QX9650 took 31 minutes, and I think I saw about 12.5 minutes on 0.19. Yes it is fast, but I was killing WUs in 3.2 minutes with the opti 0.16.

[EDIT]
Just saw another 3 through and report. Got to the results before insta-purge and saw them complete in 741 - 743 seconds and awarded 22.50 CS.
[/EDIT]
ID: 11028 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile GalaxyIce
Avatar

Send message
Joined: 6 Apr 08
Posts: 2018
Credit: 100,142,856
RAC: 0
Message 11030 - Posted: 16 Feb 2009, 16:50:40 UTC - in response to Message 11021.  

just posted the new v18d Linux-apps in the code forum...

I'd best get this fantastic news posted up on zslip then ;)



ID: 11030 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MB Atlanos

Send message
Joined: 2 Sep 07
Posts: 18
Credit: 180,611
RAC: 0
Message 11032 - Posted: 16 Feb 2009, 18:40:42 UTC
Last modified: 16 Feb 2009, 19:01:16 UTC

The new v0.18 stock app is also significant faster at PPC Macs to:
Mac mini G4 1,5 GHz are now at 1:15h from 2:15h in the past.
iMac G3 350 MHz now 4:20h vs. 8:15h before - very nice ;)
ID: 11032 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 11033 - Posted: 16 Feb 2009, 18:49:18 UTC

I was unable to get to check times as I was asleep when it crunched through a few of the 0.18 units and they purged too fast for me to see them when I got up this morning.

There are 2 running at the moment and looks like they will come in around 22 minutes.
ID: 11033 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Arion
Avatar

Send message
Joined: 10 Aug 08
Posts: 218
Credit: 41,846,854
RAC: 0
Message 11037 - Posted: 16 Feb 2009, 19:27:12 UTC - in response to Message 11030.  
Last modified: 16 Feb 2009, 19:30:05 UTC

just posted the new v18d Linux-apps in the code forum...

I'd best get this fantastic news posted up on zslip then ;)




All we need now is the Windows apps. <grin>


Have to admit the latest stock windows app. is a significant improvement in time reduction. 45-55 minutes on 2 systems down to 13-15 minutes.

Thanks to those folks who were generous enough to share their code with the project.
ID: 11037 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 . . . 6 · Next

Message boards : Number crunching : v0.18/v0.19 issues here

©2024 Astroinformatics Group