Why isn't there a problems message category?

From the BOINC log:
DivMgr

59 Milkyway@home 11/26/2024 7:58:50 AM URL http://milkyway.cs.rpi.edu/milkyway/; Computer ID 962997; resource share 90000
60 11/26/2024 7:58:50 AM Setting up GUI RPC socket
61 11/26/2024 7:58:50 AM Checking presence of 432 project files
62 Milkyway@home 11/26/2024 7:58:50 AM Resetting file projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_1.87_windows_x86_64__mt.exe: permanent HTTP error
63 Milkyway@home 11/26/2024 7:58:52 AM Started download of milkyway_nbody_1.87_windows_x86_64__mt.exe
64 Milkyway@home 11/26/2024 7:58:53 AM Giving up on download of milkyway_nbody_1.87_windows_x86_64__mt.exe: permanent HTTP error

Would Milkyway please provide an app_config.xml file and use it to determine the number of threads to schedule, instead of using <ncpus>N</ncpus> in cc_config.xml? If one wants to run two projects simultaneously, one on the GPUs and Milkyway@home on the CPUs, Milkyway@home locks out the GPU tasks since they do require some CPU time, and BOINC thinks all the CPUs are taken.

Multi-threaded apps are significantly affected by the number of threads scheduled, perhaps due to cache thrashing. Would Milkyway please consider tuning logic such as this pseudocode:

bestPregressTime = 0
startProgress = record client progress
For cpus = 1; cpus <= <ncpus>N</ncpus>; ++cpus
Do for testTime minutes
run client with cpus threads
End Do
endProgress = record client progress
if( (endProgress -startProgress) > bestPregressTime),
bestPregressTime = (endProgress -startProgress);
bestNCpus = cpus;
End if
End For
Schedule bestNCpus threads

If the Milkyway client has multiple stages, code like the above could be executed for each stage, to develop an app profile, so-to-speak.

---

Would Milkyway please provide an app_config.xml file and use it to determine the number of threads to schedule, instead of using <ncpus>N</ncpus> in cc_config.xml? If one wants to run two projects simultaneously, one on the GPUs and Milkyway@home on the CPUs, Milkyway@home locks out the GPU tasks since they do require some CPU time, and BOINC thinks all the CPUs are taken.

You can set the number of threads per task in your MilkyWay@home preferences, no app_info.xml necessary and ncpus in cc_config.xml should not be used for that anyway.

I have no idea on why you can't download the executable, perhaps detach and reattach once you removed ncpus from cc_config and restarted your BOINC client to clean up everything.

---