New optimized versions have been released


Message boards : News : New optimized versions have been released

Message board moderation

To post messages, you must log in.
1 · 2 · Next
AuthorMessage
Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar

Send message
Joined: 9 Jun 12
Posts: 584
Credit: 52,667,664
RAC: 0
Message 1450 - Posted: 7 Aug 2013, 19:50:44 UTC

Last modified: 16 Aug 2013, 17:48:28 UTC
There are new optimized versions for x86_64 windows, x86_64 linux and Raspberry Pi released. Computing times are shorten to half. I also thank to HA-SOFT for the code optimization.

Radim Vančo (Kyong)
ID: 1450 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Snow Crash

Send message
Joined: 10 May 13
Posts: 1
Credit: 2,500,560
RAC: 0
Message 1451 - Posted: 7 Aug 2013, 23:06:32 UTC
Please take a look at time estimates, on my machine the new optimized version 1.01 is estimated at 25+ hours and the old 1.00 were coming in at 1.5 hrs.
ID: 1451 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 19 Jun 12
Posts: 17
Credit: 18,669,890
RAC: 19,214
Message 1452 - Posted: 8 Aug 2013, 0:37:22 UTC - in response to Message 1450.  
There are new optimized versions for x86_64 windows, x86_64 linux and Raspberry Pi released. Computing times are shorten to half. I also thank to HA-SOFT for the code optimization.


Where?
Dublin, California
Team: SETI.USA

ID: 1452 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
rebirther
Avatar

Send message
Joined: 18 Jun 12
Posts: 10
Credit: 111,120
RAC: 0
Message 1453 - Posted: 8 Aug 2013, 4:25:48 UTC
Against the top host optimized app 4 times slower.
ID: 1453 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar

Send message
Joined: 9 Jun 12
Posts: 584
Credit: 52,667,664
RAC: 0
Message 1454 - Posted: 8 Aug 2013, 6:37:58 UTC
Snow Crash: I will change it but the estimates times will have affect on new batch in the queue.

zombie67: In the apps, just update you project and the new application will be downloaded.

rebirther: The top host optimized app uses AVX optimizations, these apps have the basic code optimized without SSE, SSE2 etc. so that computation is faster on processors without these instructions, especially ARM. More optimized applications that use these instructions are in preparation.
ID: 1454 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar

Send message
Joined: 9 Jun 12
Posts: 584
Credit: 52,667,664
RAC: 0
Message 1455 - Posted: 8 Aug 2013, 11:31:11 UTC
Added optimized x86 linux version.
ID: 1455 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
rebirther
Avatar

Send message
Joined: 18 Jun 12
Posts: 10
Credit: 111,120
RAC: 0
Message 1456 - Posted: 8 Aug 2013, 14:02:00 UTC - in response to Message 1454.  
Snow Crash: I will change it but the estimates times will have affect on new batch in the queue.

zombie67: In the apps, just update you project and the new application will be downloaded.

rebirther: The top host optimized app uses AVX optimizations, these apps have the basic code optimized without SSE, SSE2 etc. so that computation is faster on processors without these instructions, especially ARM. More optimized applications that use these instructions are in preparation.


Good to hear. I need AVX ;)
ID: 1456 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 19 Jun 12
Posts: 17
Credit: 18,669,890
RAC: 19,214
Message 1457 - Posted: 8 Aug 2013, 16:31:50 UTC - in response to Message 1454.  
zombie67: In the apps, just update you project and the new application will be downloaded.

Ah! I thought you meant apps that we manually used via Anonymous Platform. Got it.
Dublin, California
Team: SETI.USA

ID: 1457 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MarkJ
Avatar

Send message
Joined: 27 Jun 12
Posts: 129
Credit: 62,714,553
RAC: 0
Message 1458 - Posted: 8 Aug 2013, 22:42:47 UTC
Re: Est flops

On a couple of my Raspberry Pi's I got the 101 app and a single task.

The 100 app used to take 67 hours (80 hours without overclock). The 101 app is estimating 260 hours so they've been running high priority since. I realize you can't easily fix the flops for the current work in the queue but can you reduce it for the new ones please.
BOINC blog
ID: 1458 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ongjenyung

Send message
Joined: 9 May 13
Posts: 1
Credit: 4,046,651
RAC: 0
Message 1459 - Posted: 9 Aug 2013, 1:44:34 UTC
Asteroids@home 3:26:45 2013-08-09 星期日 14:26:15 period Search application 100.00 ps_130726_12670_166_1

Asteroids@home 1:44:47 2013-08-09 星期日 14:26:15 period Search application 100.00 ps_130726_12682_98_0

3:26:45VS1:44:47
ID: 1459 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[B^S]Beremat

Send message
Joined: 4 Sep 12
Posts: 3
Credit: 154,643
RAC: 0
Message 1460 - Posted: 9 Aug 2013, 3:34:46 UTC
Awesome! 1hr 33mins to 47 mins on a 3570K. Great work :)
ID: 1460 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar

Send message
Joined: 9 Jun 12
Posts: 584
Credit: 52,667,664
RAC: 0
Message 1461 - Posted: 9 Aug 2013, 9:22:32 UTC
We are now developing SSE3 and AVX version so it will be even faster. But we still need a time.
ID: 1461 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
biodoc

Send message
Joined: 26 Jan 13
Posts: 11
Credit: 13,065,884
RAC: 1,307
Message 1462 - Posted: 9 Aug 2013, 10:13:00 UTC

Last modified: 9 Aug 2013, 10:15:09 UTC
Any idea why the linux64 app isn't 2x faster? There's an improvement, but not a 2 fold decrease in crunch time like the windows app.

EDIT: link to my 2600K running linux

http://asteroidsathome.net/boinc/results.php?hostid=37254
ID: 1462 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar

Send message
Joined: 9 Jun 12
Posts: 584
Credit: 52,667,664
RAC: 0
Message 1463 - Posted: 9 Aug 2013, 10:45:14 UTC
Are you computing more projects there? What about HT, is it on or off? On this type of processor it should by 2x faster.
ID: 1463 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
biodoc

Send message
Joined: 26 Jan 13
Posts: 11
Credit: 13,065,884
RAC: 1,307
Message 1464 - Posted: 9 Aug 2013, 10:55:40 UTC
HT is on and I'm also running GPU grid on a GTX460. So 8 threads total and GPU grid is stealing processor time both on the old and new asteroids app.

I'll experiment a bit and get back to you.
ID: 1464 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
yeahway

Send message
Joined: 7 Feb 13
Posts: 1
Credit: 7,915,080
RAC: 0
Message 1465 - Posted: 9 Aug 2013, 11:26:20 UTC
Oh heck yes. Computation time on my windows and linux machines has been cut in half and then some, keep up the great work!
ID: 1465 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Andrew Dicker

Send message
Joined: 2 Aug 13
Posts: 6
Credit: 6,051,000
RAC: 0
Message 1469 - Posted: 9 Aug 2013, 22:38:01 UTC
What about a mac version? Or is it compiled with these optimisations already?
ID: 1469 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MarkJ
Avatar

Send message
Joined: 27 Jun 12
Posts: 129
Credit: 62,714,553
RAC: 0
Message 1472 - Posted: 10 Aug 2013, 10:34:56 UTC - in response to Message 1458.  

Last modified: 10 Aug 2013, 10:35:49 UTC
Re: Est flops

On a couple of my Raspberry Pi's I got the 101 app and a single task.

The 100 app used to take 67 hours (80 hours without overclock). The 101 app is estimating 260 hours


Two tasks on Raspberry Pi completed so far. 45:19 and 47:51 with a medium OC.
BOINC blog
ID: 1472 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
cyrusNGC_224@P3D

Send message
Joined: 1 Apr 13
Posts: 37
Credit: 153,496,537
RAC: 0
Message 1475 - Posted: 10 Aug 2013, 12:45:07 UTC - in response to Message 1462.  
Any idea why the linux64 app isn't 2x faster? There's an improvement, but not a 2 fold decrease in crunch time like the windows app.
Hyper Threading.


My crunch times:
Raspbian 2.6.11+, RaspberryPI@950MHz: 55h ... 63h to 45h
Debian Squeeze 2.6.32 i686, AMD Geode LX@500MHz: 90h ... 150h to <in-progress>
Debian Squeeze 3.2.0 amd64, AMD X2 4600+@2,4GHz(SSE3): 4,3h ... 5h to 3,1h
Debian Wheezy 3.2.0 amd64, Intel Core 2 6400@2,13GHz(SSE3): 5h ... 5,8h to 2,3h
Debian Lenny 2.6.39 i686, Intel P4 M@2GHz(SSE2): 11,1h to 5,9h
Debian Wheezy 3.2.0 i486, Intel Celeron M@1,4GHz(SSE2): 7,7h to 3,8h
Debian Wheezy 3.2.41 i686, Intel P3-S@1,4GHz(SSE): 8h to 4,4h
Debian Wheezy 3.2.0 amd64, AMD X6 1090T@3,8GHz(SSE3): 2,7h to 1,22h
Debian Wheezy 3.10.3 amd64, Intel i7-3632QM@2,88GHz(AVX): 2,5h ... 4h to 1,58h
ID: 1475 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Saenger
Avatar

Send message
Joined: 18 Jun 12
Posts: 23
Credit: 5,455,486
RAC: 0
Message 1477 - Posted: 10 Aug 2013, 20:03:51 UTC - in response to Message 1462.  

Last modified: 10 Aug 2013, 20:05:03 UTC
Any idea why the linux64 app isn't 2x faster?

Mine is two times faster, it's a C2Q9450 with ubuntu Version 12.04 (precise) (64-Bit).
Grüße vom Sänger
ID: 1477 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
1 · 2 · Next

Message boards : News : New optimized versions have been released