Hi everyone,
So ah, I played with the rtrace -n option a bit, and got results I don't understand. Anyone interested in having a look, please visit this page:
http://www.rumblestrip.org/interests/light/rtrace-multiprocessing-option-initial-test-results/
- Rob
Hi Rob,
When there is no ambient file there is usually an ambient cache created in
memory. If you don't have an ambient file each rtrace process will build
its own cache. I'm guessing that each process spends most of the time
building the cache with a first point and the subsequent calculations are
very quick. Because getting the cache to saturation is the majority of the
calculation time and each processor is doing this independently before
really getting going on the points it won't matter if you have 1 or 16
processors, the simulation will take the same amount of time.
To completely turn off the ambient cache use -aa 0. I expect you will see a
linear correlation between number of processes and simulation times, but the
sim times will be much much longer.
On a side note, I'm very jealous of your 16 core mac pro!
Andy
On 12/17/09 8:30 PM, "Rob Guglielmetti" <[email protected]> wrote:
Hi everyone,
So ah, I played with the rtrace -n option a bit, and got results I don't
understand. Anyone interested in having a look, please visit this page:http://www.rumblestrip.org/interests/light/rtrace-multiprocessing-option-initi
al-test-results/- Rob
_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general
____________________________________________________________
Electronic mail messages entering and leaving Arup business
systems are scanned for acceptability of content and viruses
I agree with Andy's assessment. I also don't think you'll see a linear acceleration past 8 processes, as the MacPro doesn't really have 16 processors, only 16 threads on 8 cores. It's not nearly the same, which is why I'm holding out for more processors before I upgrade my old PowerMac G5 Quad.
-Greg
From: Andrew McNeil <[email protected]>
Date: December 17, 2009 11:00:00 PM PSTHi Rob,
When there is no ambient file there is usually an ambient cache created in
memory. If you don't have an ambient file each rtrace process will build
its own cache. I'm guessing that each process spends most of the time
building the cache with a first point and the subsequent calculations are
very quick. Because getting the cache to saturation is the majority of the
calculation time and each processor is doing this independently before
really getting going on the points it won't matter if you have 1 or 16
processors, the simulation will take the same amount of time.To completely turn off the ambient cache use -aa 0. I expect you will see a
linear correlation between number of processes and simulation times, but the
sim times will be much much longer.On a side note, I'm very jealous of your 16 core mac pro!
Andy
On 12/17/09 8:30 PM, "Rob Guglielmetti" <[email protected]> wrote:
Hi everyone,
So ah, I played with the rtrace -n option a bit, and got results I don't
understand. Anyone interested in having a look, please visit this page:http://www.rumblestrip.org/interests/light/rtrace-multiprocessing-option-initi
al-test-results/- Rob
_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general____________________________________________________________
Electronic mail messages entering and leaving Arup business
systems are scanned for acceptability of content and viruses_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general
OK, thanks guys. It's not as if I wouldn't use caching normally, I just saw something that didn't make sense to me. Thanks for straightening me out.
As far as the hyperthreading, I did do some tests and it seemed to be nearly linear to 15 or so, for mkillum at least. But those weren't as controlled as the testing I did yesterday with rtrace -n. I'll have to look at that some more. Anyway thanks guys!
- Rob
On Dec 18, 2009, at 8:47 AM, Greg Ward wrote:
I agree with Andy's assessment. I also don't think you'll see a linear acceleration past 8 processes, as the MacPro doesn't really have 16 processors, only 16 threads on 8 cores. It's not nearly the same, which is why I'm holding out for more processors before I upgrade my old PowerMac G5 Quad.
-Greg
From: Andrew McNeil <[email protected]>
Date: December 17, 2009 11:00:00 PM PSTHi Rob,
When there is no ambient file there is usually an ambient cache created in
memory. If you don't have an ambient file each rtrace process will build
its own cache. I'm guessing that each process spends most of the time
building the cache with a first point and the subsequent calculations are
very quick. Because getting the cache to saturation is the majority of the
calculation time and each processor is doing this independently before
really getting going on the points it won't matter if you have 1 or 16
processors, the simulation will take the same amount of time.To completely turn off the ambient cache use -aa 0. I expect you will see a
linear correlation between number of processes and simulation times, but the
sim times will be much much longer.On a side note, I'm very jealous of your 16 core mac pro!
Andy
On 12/17/09 8:30 PM, "Rob Guglielmetti" <[email protected]> wrote:
Hi everyone,
So ah, I played with the rtrace -n option a bit, and got results I don't
understand. Anyone interested in having a look, please visit this page:http://www.rumblestrip.org/interests/light/rtrace-multiprocessing-option-initi
al-test-results/- Rob
_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general____________________________________________________________
Electronic mail messages entering and leaving Arup business
systems are scanned for acceptability of content and viruses_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general
Rob,
I too was somewhat bemused/confused re: virtual cores. If you find the opportunity, could you also test it with totally independent jobs? In fact, varying multiple instances of the same job would do. I'd dearly like to have my scepticism refuted.
Cheers,
-John
-----------------------------------------------
Dr. John Mardaljevic
Reader in Daylight Modelling
Institute of Energy and Sustainable Development
De Montfort University
The Gateway
Leicester
LE1 9BH, UK
+44 (0) 116 257 7972
+44 (0) 116 257 7981 (fax)
Lemme see what I can do. Turns out, my recollection of rpict scaling was related to a different test I did, where I subdivided a view into 40 tiles and sent them off to a cluster, not to my local machine. So in that case it should have been expected that I got a linear speed-up because it was distributed to different physical cores. But I did do a test of mkillum with the -n option and I thought I saw a linear increase to 15 on my local machine. This was a little while ago and I'll have to dig it up and look closely at it, as well as do a proper test of rtrace. I'll let you know what I find out but I'm not optimistic at this point. =8-/
- Rob
On Dec 18, 2009, at 11:15 AM, John Mardaljevic wrote:
Rob,
I too was somewhat bemused/confused re: virtual cores. If you find the opportunity, could you also test it with totally independent jobs? In fact, varying multiple instances of the same job would do. I'd dearly like to have my scepticism refuted.