CloudCompare forum

Hello all, first post ;-)

I want to run these usages of CloudCompare in command line mode in a Cloud (AWS, GCP, Azure or other):

some of the used Flags (over different calls):
-MERGE_MESHES
-MERGE_CLOUDS
-C2M_DIST with -MAX_DIST and -MAX_TCOUNT set

storage format is BIN files

So my questions are:

- Can these calculations benefit from having a GPU available or are they always running on the CPU?
I think CPU only, just wanted to make sure.

- How parallelized are these calls, what number of cores would be a sweet spot between performance (low runtime) and price?
Are there any benchmarks that show speedup vs number of cores?

- How is the scaling regarding RAM? Is more RAM always speeding up calculations or only up to a certain point?
Or should I just try to run my models and increase RAM if CC crashes? Does CC crash if there is too little RAM, or does it just work slower?

- Is there something else to consider when running CloudCompare in the Cloud? ;-)

thanks for your time and/sharing your experiences,
rookie

1) We don't rely on the GPU sadly

2) C2M_DIST relies on parallelization indeed. I don't have an answer to your question about the best number of cores (however, the more cores the less efficient each core is virtually)

3) Actually it's more the lack of RAM that would slow down the computation if the OS starts to 'swap' RAM contents on the disk. Of course if the system doesn't support this feature, you'll end with an error or a crash.

4) CC is not a real command line tool (it's still a graphical application). The lack of graphic card on some servers can prevent it to start.

Thank you for your quick response!

I'll try a few different configurations and see for myself and thanks for the hint about maybe having problems with scenarios with no GPU present (question 4).

best regards,
rookie