You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Bug]: ctest finds 10 cores on M2Pro instead of 6 performance cores #146
So I think there may be more to this than we realize; if I go back a few days to ba4145d it passes everything. So the GLOW path name changes from aa50fa1 are causing some kind of problem.
Here's the glow error introduced by that new commit; it's missing some file -- at least on my system:
32: At line 140 of file /Users/zettergm/Projects/gemini3d/build/_deps/glow-src/ssflux.f90
32: Fortran runtime error: Cannot open file '/Users/zettergm/Projects/gemini3d/build/_deps/glow-src/datassflux_hint.dat': No such file or directory
If possible, reach out to my email or cell and I can just send the entire directory. I've updated since the above. There were TONS of places where settings or values or other measurements could be ignored or dropped silently without warnings being reported.
This "gemini3d/build/_deps/glow-src/datassflux_hint.dat" no such file ... is due to a missing "/" file separator. I hadn't changed the code regarding that "/" but I think the easiest and robust solution is to add a "/" in the gemini3d/glow code, in case a "glow_data_dir" without the "/" is passed - good defensive programming.
I didn't notice this was trying to refer to a subdirectory of the source code; @scivision I'm assuming you'll add the robustness check to the glow fork? @guygrubbs what do you mean by "measurements"? Also are you able to fork the repo and apply your changes there? I have no doubt there a lots of omissions of that nature. I did some cleanup of the input options code a few weeks ago that may also have fixed a few things.
@mattzett Please try my fork that I pushed and see if the above or other issues are resolved and also that the update didn't break or change any physics at a large scale.
Activity
guygrubbs commentedon Sep 16, 2026
As part of my work I described for the discussion I opened, I did find multiple issues including the above which are worth examining and updating.
mattzett commentedon Sep 16, 2026
Yeah it used to detect performance vs. efficiency cores correctly for both apple silicon and intel chips so I'm not sure what changed.
mattzett commentedon Sep 16, 2026
So I think there may be more to this than we realize; if I go back a few days to ba4145d it passes everything. So the GLOW path name changes from aa50fa1 are causing some kind of problem.
mattzett commentedon Sep 16, 2026
Testing of ba4145d:
mattzett commentedon Sep 16, 2026
Testing of aa50fa1:
mattzett commentedon Sep 16, 2026
Here's the glow error introduced by that new commit; it's missing some file -- at least on my system:
guygrubbs commentedon Sep 16, 2026
I'm unable to push my new branch which may resolve a decent amount of these. Here's a patch file.
changes.patch
guygrubbs commentedon Sep 17, 2026
If possible, reach out to my email or cell and I can just send the entire directory. I've updated since the above. There were TONS of places where settings or values or other measurements could be ignored or dropped silently without warnings being reported.
scivision commentedon Sep 17, 2026
This "gemini3d/build/_deps/glow-src/datassflux_hint.dat" no such file ... is due to a missing "/" file separator. I hadn't changed the code regarding that "/" but I think the easiest and robust solution is to add a "/" in the gemini3d/glow code, in case a "glow_data_dir" without the "/" is passed - good defensive programming.
mattzett commentedon Sep 18, 2026
I didn't notice this was trying to refer to a subdirectory of the source code; @scivision I'm assuming you'll add the robustness check to the glow fork? @guygrubbs what do you mean by "measurements"? Also are you able to fork the repo and apply your changes there? I have no doubt there a lots of omissions of that nature. I did some cleanup of the input options code a few weeks ago that may also have fixed a few things.
scivision commentedon Sep 18, 2026
I added this script https://github.com/gemini3d/gemini3d/blob/main/scripts/macos_cpu_info.sh to help diagnose the CPU count issue, to see if my machine is similar to yours or if it has more than 2 CPU core levels.
It might output like
Indicating two CPU core levels, with 4 each of P performance and E efficiency CPU cores - total of 8 CPU physical cores.
guygrubbs commentedon Sep 20, 2026
@mattzett Please try my fork that I pushed and see if the above or other issues are resolved and also that the update didn't break or change any physics at a large scale.