Hi :)

Do you prefer Windows crunchers to stay with Nbody app deselected, and announce when we shoudl resume work on this app? or would using the current version with errors help you in the development?

---

Hey,

I'd say keep it checked. When we next release we will make an announcement.

Thanks!
Sidd

Hello,

Have started today crunching MW on an ATI again. W7U.

No problem with MW.

But 42 INVALID Milkyway@Home Separation (Modified Fit) v1.36 (opencl_ati_101) so far.

Best,

Philippe

Hello,

Have started today crunching MW on an ATI again. W7U.

No problem with MW.

But 42 INVALID Milkyway@Home Separation (Modified Fit) v1.36 (opencl_ati_101) so far.

Best,

Philippe

Hello

219 INVALID Milkyway@Home Separation (Modified Fit) v1.36 (opencl_ati_101) so far.

Best,

Philippe

Hi Sidd.
I have windows 7 Home Premium (64 bit) running on a Toshiba laptop (computer 596332) with an Intel Celeron 900 2.20 Ghz processor with 2.00 GB RAM. The OS has Service Pack 1 installed on it along with all MS OS updates up to, but not including, those released on or after 2015-04-14.

To acquire data on the problem, I ran three tests. For the first two tests, I let a specific Nbody simulation run twice, with the second run continuing non-stop, lasting approximately 168 hours. During all three of these tests the laptop was left entirely to itself almost the entire run, with the units getting an average of nearly 95% CPU over the course of the tests. Here is what I observed.

1. Neither of the two units took any checkpoints during any of the runs.

2. Attempting to stop and restart the first unit on the first run resulted in no data saved from the first run. I restarted it by shutting down and restarting BOINC normally. The restarted unit began at 0.000% processed.

3. On both runs of the first unit it appeared to run fairly normally until it passed approximately 95% processed. At that point the percentage processed slowed to a crawl, advancing only .001% about every five minutes or so.

4. On both runs of the first unit it eventually reached the point where BOINC reported 100% processed with estimated time remaining showing "---". Still no checkpoint was taken, nor were any data files saved. This state of nearly 100% CPU processing the unit and 100% processed, with no time remaining, continued for dozens of hours.

5. Processing continued until the deadline had passed. At that time I shut down BOINC and restarted it to see if the unit would save any files and/or report as completed. Neither occurred. When I restarted BOINC the unit reported as a task that was "Not started by deadline - canceled".

6. I then started a second nBody unit. It ran for over 20 hours with over 18 hours CPU time. It exhibited the same symptoms. After 20 hours I aborted the unit.

I might add that I have had no observable problems processing any Milkyway@Home units on this laptop since I joined your project.

I hope this information might provide some insight as to what is occurring, Sidd. Good luck!

Cherokee150,

Thank you! Your info is remarkably helpful. I think I have a line on a bug that might be occurring. If there is no checkpointing at all then that means it is probably the initialization that is getting stuck somewhere. I am looking into where.

Thanks!
Sidd

Hello,

As I see you read this topic, you might have noticed my 2 posts related to
the invalid "Separation (Modified Fit) v1.36 (opencl_ati_101)"
I see in my stats.

Is this normal ?

Should I stop running this app ?

...

:/

Will add 2*HD7950 very soon, and your comment would be much appreciated.

Thank You

Best

Philippe

Does anyone have any idea about the impact of Windows 10 on BOINC and milkyway project?

---

Yes. BOINC and the projects appear to be working just fine with Windows 10, identically to previous versions of Windows.