Race conditions in grid accumulator? about cilantro

[kzampog]

Thank you so much for reporting this! There were also performance issues in the parallel version of GridAccumulator on Windows (#29), so there's definitely some digging to be done there. A minimal example with data would be extremely helpful!

[kzampog]

Hi again,

Please have a look at ea57e47. Normal consistency was not properly enforced in the parallel version (when merging accumulators), so this might be the reason you were getting zero-valued normals. I think this should fix it, and performance is very close to what it was without the additional checks. In any case, I will refactor this to something cleaner in the future.

Please let me know of your findings!

[wannesvanloock]

Awesome, that fixed it indeed. Thanks for the lightning fast fix 🙂

To be honest, this part of the code is difficult to understand for me, so I didn't feel confident to tackle the issue myself. Looking forward to a refactor!

[kzampog]

Awesome, glad that fixed it! :D

Yes, that part is a bit messy. I'll try to improve it and make it cleaner!