just a feedback on the calculix_2.23_4win.zip zipped file (probably an oversight)
Thanks for the info. We replace the file soon.
Nice, itâs fixed now
Uhhh. Great!!. Thank very much Don Guido and congratulations for this new release.
Regards
ÂżHas someone else notice a substantial increment in the time required for complete the analysis?
Same file, same number of cores,.. Iâm comparing differences between version 2.22 and 2.23 in case there is some point that could be set up different in my computer. ÂżMaybe a new system variable required?. I would appreciate any advice.
Looking at the screenshot, the windows executable is built with PastiX 6.4.0.
How was this done?
Looking at the diff between 2.22 and 2.23, there does not seem to be a lot of change in pastix.c.
The README.INSTALL file still mentions GitHub - Dhondtguido/PaStiX4CalculiX, which hasnât been changed in four years and which uses a modified version of PastiX 6.0.1.
Look here:
This has not been merged and is not part of 2.23, as far as I can tellâŚ
Using my scripts (with some updates that I will push shortly) I can build ccx 2.23 with PaStiX4CalculiX as modified by Kabbone to work without CUDA (PaStiX4CalculiX-2.17_cudaless).
Given that the work done by Kabbone to make CalculiX work with PaStiX 6.4 has not been merged yet, the executable that Disla used (which I presume came from Calculix_2.23_4win.zip) is probably a modification of 2.23.
With an executable built from the 2.23 source (using PaStiX4CalculiX-2.17_cudaless), I donât see a significant difference in CalculiX time compared to an executable built from the 2.22 source using the same method.
Iâm using the last calculix_2.23_4win.zip downloaded from the official calculix webside. Uncompress and executed without any modification. 2.22 was also the oficial from calculix_2.22_4win.zip. Both versions are the ccx_dynamic.exe.
In my case, Calculix_2.23_4win.zip is twice slower than ccx2.22PaStiX.
same with pardiso/spooles? which OS, windows 10 or 11?
Mine is Win10. Pardiso doesnât seems affected ( itâs slightly faster than Pastix 2.22).
I never use spooles. I have tried again just for curiosity but itâs not comparable
in my case, on windows 11.
The change in run time for this specific executable is clearly due to newer PaStiX version 6.4.0 being used, which does not include performance enhancements from PaStiX4CalciliX. Also I wouldnât call this a 2.23 build since it obviously includes sources from an unmerged pull request.
