Message boards :
Number crunching :
At what point does server Flop guesstimates get transmitted to the client?
Message board moderation
Author | Message |
---|---|
Keith White Send message Joined: 29 May 99 Posts: 392 Credit: 13,035,233 RAC: 22 |
In particular AP tasks (but I assume it's similar for all apps). Reason I'm asking is for a while I've been using <flops> in my CPU (and GPU) AP portion of app_info to established a flop estimate that won't put my CPU AP WU completion times in the 9+ day range. But now I have 19 completed (I was told for MB it needed more than 10 to use the server estimate) CPU WU so I removed my flop entry expecting the server to relay that sooner rather than later. Of course the estimated time remaining shot up 11x from 22 hours to 233 hours and now all my CPU WUs are in panic mode thinking they'll never get done now that I have 48 days (really 5) of AP WUs on my system. And since my GPU AP task has 11 completed I'm planning to pull those <flops> lines out of app_info as well (a completion time estimate that's 5x higher than reality). So how frequently does that info get transmitted back down to the client from the server. Please I hope it isn't when it fetches a new CPU AP WU because I don't think the scheduler will let one run until I'm out of MB V7. "Life is just nature's way of keeping meat fresh." - The Doctor |
cov_route Send message Joined: 13 Sep 12 Posts: 342 Credit: 10,270,618 RAC: 0 |
If you are using app_info.xml the Average Processing Rate (APR) seen on your user page is not used. You have to maintain the <flops> entry yourself. Good thing too because APR calculation is unreliable. Right now it shows my modest little 6670 gpu is chewing through MB at over 7 teraFLOPS. It would be nice but I don't think so. See also this thread. |
Keith White Send message Joined: 29 May 99 Posts: 392 Credit: 13,035,233 RAC: 22 |
Actually the APR does work some of the time as my MB V7 CPU is accurate and the values I had used as flop guesstimates for AP CPU was the APR being reported and while my AP GPU guesstimate was of my own construction it was within 3% of what the APR was for AP GPU was reporting. MB GPU is a different story altogether, wasn't being computed right for V6 or V7 for my system. In slightly brighter news two of the five "233 hour" AP WUs and now being processed as "high priority". Maybe the other three are next so the system will think it's OK to fetch more CPU work. The question would be will there be any AP units to fetch if the APR is only transmitted when data for that app is downloaded. "Life is just nature's way of keeping meat fresh." - The Doctor |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Actually the APR does work some of the time as my MB V7 CPU is accurate and the values I had used as flop guesstimates for AP CPU was the APR being reported and while my AP GPU guesstimate was of my own construction it was within 3% of what the APR was for AP GPU was reporting. I gave up on the add flops/remove flops thing some time ago. I just leave flops in for the GPU and currently everything is OK. If you remove or remove the value then everything gets all wonky again. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
©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.