LonelyCat124

I worked out why nothing was happening. Now the code breaks because when looking in a kernel the checks I did for Reference's being in the parent private region don't...

I came across another issue while progressing here. Inside even simple NemoLite2D kernels there are two extra challenges: ``` SUBROUTINE bc_flather_u_code(ji, jj, ua, hu, sshn_u, tmask) USE physical_params_mod, ONLY: g...

Ok, I will look at that for kernels. I hope that avoids a lot of issues for non-inline code and makes life easier. I'm then just left with the challenge...

@arporter I'm still struggling to find how to access the stencil info. I've found the `GOStencil` class (for GOcean) and some others, but its not clear to me how to...

> You'll be pleased to hear that Rupert has just completed refactoring the handling of kernel metadata (for GOcean so far - LFRic is now in progress) so life should...

> Only the PSyKAl APIs currently have this information. Ideally we would attempt to use the dependence analysis to construct it for a CodedKern in NEMO but we don't do...

> Yes.. the get_kernel_schedule() method Ok, i thought i'd tried this and it not worked but I'll test again! > It also looks like you have a very thin transformation...

> The problem with doing things when lowering (or gen_code) is that is not composable with other transformations. If it is a transformation it can be applied and then explored/walked...

Aha no, I circled back round and the issue I had with nemo was a kernel that has: ``` jiu = ji + 1 array[jiu] = ... ``` as I...

> > as I can't track what jiu is right now, and it doesn't appear as a private variable (because its local to the routine). I need to think about...