Unexpected behaviour working with JITParticles and custom kernels

[VeckoTheGecko]

I've been encountering very strange behaviour when using the JITParticle combined with custom kernels.

Eg.

• Using a simple nested if statement in a kernel causes a bugged output file (part of the if statement getting hit regardless of the condition, causing all particles to change instantly), while collapsing the nested if statement to an equivalent <condition> and <condition> causes no issues
• Memory access violation bug appearing out of nowhere, despite not changing the kernels (duplicate of exception when runnning, in which both recovery and mask on #793)

I guess my question is more to get a feel of:

• whether these bugs are known/encountered by the developers often
• whether these bugs/unexpected behaviour is something that can be attributed to parcels, or is it moreso the dependencies/way that things are compiled to C
• whether working on Windows makes me more prone to running into these bugs (and if so, how much worse is it to work on Windows compared to Linux? Is it worth swinging across?)

PS:
I experimented using the ScipyParticle, and my simulation went from 2 minutes (100% completion) to 2hr 30mins (22% completion) so I don't think thats an option for me. Its "JITParticle or nothing" with my current kernels.

[erikvansebille]

Hi @VeckoTheGecko, comments to your two questions below

Using a simple nested if statement in a kernel causes a bugged output file (part of the if statement getting hit regardless of the condition, causing all particles to change instantly), while collapsing the nested if statement to an equivalent and causes no issues

Hmm, nested if-statements are not explicitly implemented in the codeconverter (which converts the python-code in Kernels to JIT-code. So I wouldn't be surprised that this goes wrong. I also don't know how to easily fix this, not how to give an explicit warning. I'll open an Issue and think about it. But much safer to use non-nested-if-statements!

Memory access violation bug appearing out of nowhere, despite not changing the kernels (duplicate of #793)

I have never experienced this Issue myself, but don't run Windows. I also haven't seen it from anyone else here in my group (none of whom use Windows). So he bug seems to be explicitly related to Windows? No idea how to even begin to fix this...