Message boards :
Number crunching :
New WUs
Message board moderation
Author | Message |
---|---|
Send message Joined: 28 Aug 07 Posts: 49 Credit: 556,559 RAC: 0 |
I see that we must be crunching a new type of WU. I say this because I see that must of the WUs are being granted 4 credits each, up from 2. They do not seem to run twice as long as the last batch ;) |
Send message Joined: 8 Oct 07 Posts: 289 Credit: 3,690,838 RAC: 0 |
I see that we must be crunching a new type of WU. I say this because I see that must of the WUs are being granted 4 credits each, up from 2. They do not seem to run twice as long as the last batch ;) LOL I am seeing + and- twice as long...go figure...hey Acmefrog we have to get on the same page sooner or later? eh? |
Send message Joined: 28 Aug 07 Posts: 49 Credit: 556,559 RAC: 0 |
Actually, it looks like the gs_24 WUs do take about twice as long as the gs_22 WUs. What threw me off is the fact that my quad crunches about twice as fast as the Core2 I have running. |
Send message Joined: 30 Aug 07 Posts: 2046 Credit: 26,480 RAC: 0 |
Actually, it looks like the gs_24 WUs do take about twice as long as the gs_22 WUs. What threw me off is the fact that my quad crunches about twice as fast as the Core2 I have running. Yep :) gs_24 is using a different sized volume than gs_22 (and the new ps_1). As part of the assimilator updates in the last few days i made the volume size an input to the search being done (as opposed to being statically in the assimilator). This'll give us a lot more flexibility in our searches. Also, we're thinking about testing doing searches that adaptively modify the volume as the search progresses -- ie, the fitness calculation will be initially less accurate, but faster, but while the search starts to converge it'll progressively get more accurate. |
Send message Joined: 10 Nov 07 Posts: 96 Credit: 29,931,027 RAC: 0 |
I’ve noticed that the short WUs are taking less time on my G5 Mac than they did before: today’s times are coming in around 175 s, down from ~205 s. But the longer ones do take very nearly double what the short ones used to, averaging about 415 s. |
Send message Joined: 30 Aug 07 Posts: 2046 Credit: 26,480 RAC: 0 |
I’ve noticed that the short WUs are taking less time on my G5 Mac than they did before: today’s times are coming in around 175 s, down from ~205 s. But the longer ones do take very nearly double what the short ones used to, averaging about 415 s. i think most of the parameter sets generated by ps_1 and ps_2 got messed up. I've stopped those searches and started up ps_3, ps_4 and ps_5. These still should be taking half the time as gs_25, but as much time as the work units used to be taking. |
©2024 Astroinformatics Group