Mark Taylor

So the array is allocated outside the parallel region, but then marked thread private. So when the parallel region starts, each thread should get a copy of the array. when...

@ambrad explained this to me: for pointers which are declared thread private, then all threads must allocate them (since each thread has it's own version of the pointer). most of...

SCREAM has not yet been run on Intel PVC. So correct config options are unknown! Just to warn you, this will probably be lot of work. But if you do...

Correct - pack_size is for vectorization on CPU systems. It should be 1 on GPU systems. some documentation: doi: 10.5194/gmd-12-1423-2019

Fortran code has: #ifdef SCREAM adjust_ps=.false. ! Lagrangian case can support adjusting dp3d or ps #else adjust_ps=.true. ! Lagrangian case can support adjusting dp3d or ps #endif And thus hommexx...

adjust_ps=false: mass fluxes from physics are applied locally (removing the mass from the layer it came from) adjust_ps=true: mass fluxes are smeared over the whole column and removed from the...

this is one of these nuisance bug fixes. it should have been fixed a long time ago, but the hassle of evaluating it combined with the small impact means it's...

great, thanks. it will probably be a several more years before we next have to use this tool, but still good to resolve this PR.

We should discuss the async model. Years ago, we were spending most of our time aggregating the data down to the I/O processors, and it was clear that just putting...

I verified this conjecture by changing, in subroutine copy_pio_var01d, in pio_nf_utils.F90 ``` ! ierr = pio_put_var(ofh, ovid%varid, dval) do i=1,length index(1)=i ierr = pio_put_var(ofh, ovid%varid, index, dval(i)) enddo ``` and...