Badly optimized Mac PPC app?

Martin P.
Martin P.
Joined: 17 Feb 05
Posts: 162
Credit: 40,156,217
RAC: 0

RE: Significant app

Message 73636 in response to message 73635

Quote:
Significant app optimization is on hold while the team trys to find the last handful of low occurence bugs.

Hmmm, hard to believe. Windows version 4.13 brought a 20% performance increase over version 4.07: http://einsteinathome.org/task/87874714 . They obviously DO work on optimizations for Windows.

So, did you mean "Significant app optimization for Mac PPC is on hold ..."?

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,364
Credit: 3,562,358,667
RAC: 0

I'm not certain what you're

I'm not certain what you're trying to tell me by linking to a single WU, but there's been no mention of a speedup either by bernd in the 4.13 thread, nor by any of the people replying to it. It's possible that somethin was done to result in a speadup, but if so it wasn't done intentionally.

the current results on my two machines are 50-62k vs 59-64, and 49-62 vs 60-76. In the first case the peak to valley spreads from the combined range are within the ranges seen from people charting runtime stats long enough to get the entire curve (28% vs 21% and 34% in the graph I found). The 55% spread in my second machine is somewhat larger, but I also use it much more heavily. Gaming sessions that max one core will badly slow the two einsein proceses down by forcing them to fight over cache resources. I've seen 33% increases in the amount of CPU time needed in earlier runs, so while somewhat suggestive the numbers from my second box aren't conclusive.

archae86
archae86
Joined: 6 Dec 05
Posts: 3,156
Credit: 7,179,524,931
RAC: 745,095

RE: Windows version 4.13

Message 73638 in response to message 73636

Quote:
Windows version 4.13 brought a 20% performance increase over version 4.07:


I agree that Windows 4.13 and 4.15 are substantially faster than 4.07, though on my Core 2 machines the difference is about 10%, definitely not 20%. I posted evidence I think pretty strong on that point here.

However, based on the introductory comment Bernd made for the 4.11 beta, I think the improvement was not a change to the primary science code, but rather to error-handling and containment.

My scan did not pick up enough 4.11 results to give an effective comparison for that specific release.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,522
Credit: 694,470,247
RAC: 119,568

RE: RE: Significant app

Message 73639 in response to message 73636

Quote:
Quote:
Significant app optimization is on hold while the team trys to find the last handful of low occurence bugs.

Hmmm, hard to believe. Windows version 4.13 brought a 20% performance increase over version 4.07: http://einsteinathome.org/task/87874714 . They obviously DO work on optimizations for Windows.

So, did you mean "Significant app optimization for Mac PPC is on hold ..."?

Most of the speedup we see in the Windows app now is the result of replacing debug code that introduced a slowdown earlier. The next PPC app will surely include these changes.

CU
Bikeman

Martin P.
Martin P.
Joined: 17 Feb 05
Posts: 162
Credit: 40,156,217
RAC: 0

RE: I'm not certain what

Message 73640 in response to message 73637

Quote:
I'm not certain what you're trying to tell me by linking to a single WU, but there's been no mention of a speedup either by bernd in the 4.13 thread, nor by any of the people replying to it. It's possible that somethin was done to result in a speadup, but if so it wasn't done intentionally.

Dan,

sorry, it was meant to be the link to the computer: http://einsteinathome.org/host/624455/tasks . There you can see that the speed-up is in fact 20% (67,575 seconds for 233.81 credit versus 84,203 seconds for 218.52 credits)

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,364
Credit: 3,562,358,667
RAC: 0

RE: Most of the speedup we

Message 73641 in response to message 73639

Quote:

Most of the speedup we see in the Windows app now is the result of replacing debug code that introduced a slowdown earlier. The next PPC app will surely include these changes.

That doesn't appear to be listed in the mini changelogs that came with any of the beta apps. Had it been I probably would've swtiched over instead of waiting for 4.13 to become hte official app.

Martin P.
Martin P.
Joined: 17 Feb 05
Posts: 162
Credit: 40,156,217
RAC: 0

RE: RE: Most of the

Message 73642 in response to message 73641

Quote:
Quote:

Most of the speedup we see in the Windows app now is the result of replacing debug code that introduced a slowdown earlier. The next PPC app will surely include these changes.

That doesn't appear to be listed in the mini changelogs that came with any of the beta apps. Had it been I probably would've swtiched over instead of waiting for 4.13 to become hte official app.

Dan,

4.13 IS the official version for more than a week now. Please do not pretend to know about future releases ("Significant app optimization is on hold while the team trys to find the last handful of low occurence bugs. ") without even using the current versions.

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,364
Credit: 3,562,358,667
RAC: 0

RE: RE: RE: Most of

Message 73643 in response to message 73642

Quote:
Quote:
Quote:

Most of the speedup we see in the Windows app now is the result of replacing debug code that introduced a slowdown earlier. The next PPC app will surely include these changes.

That doesn't appear to be listed in the mini changelogs that came with any of the beta apps. Had it been I probably would've swtiched over instead of waiting for 4.13 to become hte official app.

Dan,

4.13 IS the official version for more than a week now. Please do not pretend to know about future releases ("Significant app optimization is on hold while the team trys to find the last handful of low occurence bugs. ") without even using the current versions.

You're reading something other than what I was trying to say. What I meant was that if the changelog would've mentioned the speedup from removing unneeded debugging code I would have switched 2 or 3 days earlier, and flushed my queue (3 days) to bring the new app into play sooner. Since the official thread mentioned nothing about performance gains I assumed it would run about the same and waited for an offial rollout.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,522
Credit: 694,470,247
RAC: 119,568

Hi! The debug code wasn't

Hi!

The debug code wasn't so much removed but changed so that it is more useful (and happens to be faster). The focus of the Win beta tests are still on stability, I guess that's why the changelists don't dwell on performance so much.

CU
Bikeman

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282,700
RAC: 0

RE: The debug code wasn't

Message 73645 in response to message 73644

Quote:

The debug code wasn't so much removed but changed so that it is more useful (and happens to be faster). The focus of the Win beta tests are still on stability, I guess that's why the changelists don't dwell on performance so much.

Do you know if there has been more progress on finding out the wide variations in runtime?

Comment viewing options

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