Message boards :
Number crunching :
v7 anonymous -9s
Message board moderation
Author | Message |
---|---|
David S Send message Joined: 4 Oct 99 Posts: 18352 Credit: 27,761,924 RAC: 12 |
I thought one of the advantages of the Lunatics apps was that they would handle high counts of possible signals better than simply giving a -9 overflow error. This WU came up with -9s for both of us, and with different counts adding up to 30. BTW, now that we're looking for more types of signals, mightn't it be a good idea to increase the maximum number of them before an overflow is triggered? David Sitting on my butt while others boldly go, Waiting for a message from a small furry creature from Alpha Centauri. |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
I thought one of the advantages of the Lunatics apps was that they would handle high counts of possible signals better than simply giving a -9 overflow error. And how will they validate against stock if they report more (e.g. 60) signals?? Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
I thought one of the advantages of the Lunatics apps was that they would handle high counts of possible signals better than simply giving a -9 overflow error. This WU came up with -9s for both of us, and with different counts adding up to 30. I'm sure Eric considered raising the <max_signals> setting, but it wouldn't make a big difference. Most overflows happen very early (like that example), so it is likely that even boosting the setting to hundreds or thousands would not keep overflow from happening. It would mainly require more storage space in the science database. The attempted validation of those tasks being inconclusive is indeed puzzling. Although known differences in processing order can give that effect when comparing a CPU overflow to a GPU overflow, that is not the case here because Spike processing is completed before Gaussian processing starts for both. *** a few minutes later... I downloaded a copy of the WU, checked with an offline test, and it seems there's a consistent difference between stock and optimized for that WU. The Windows CPU stock 7.00 overflowed with 29 Spikes and 1 Gaussian while my AKv8c SSE2 got 26 Spikes and 4 Gaussians. Thanks for pointing out the WU, we'll figure out if there's a definite bug or it is just a case where the signals are sitting right on the threshold levels. Joe |
David S Send message Joined: 4 Oct 99 Posts: 18352 Credit: 27,761,924 RAC: 12 |
The attempted validation of those tasks being inconclusive is indeed puzzling. Although known differences in processing order can give that effect when comparing a CPU overflow to a GPU overflow, that is not the case here because Spike processing is completed before Gaussian processing starts for both. I'm glad one of my "this seems weird to me" posts seemed weird to someone else too, for once. I'm going to keep an eye on this one too. I was the fourth one to get it, but the first to run it on CPU, first to run it with Anonymous, and first not to get an error or overflow. David Sitting on my butt while others boldly go, Waiting for a message from a small furry creature from Alpha Centauri. |
David S Send message Joined: 4 Oct 99 Posts: 18352 Credit: 27,761,924 RAC: 12 |
I'm going to keep an eye on this one too. I was the fourth one to get it, but the first to run it on CPU, first to run it with Anonymous, and first not to get an error or overflow. This one has validated, the 5th host's CPU stock app agreeing with my CPU Lunatics app. Two that ran it on GPU overflowed and one had too many exits. Still waiting for the one I started this thread with. David Sitting on my butt while others boldly go, Waiting for a message from a small furry creature from Alpha Centauri. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.