Andy Porter

icon indicating a website link http://www.hartree.stfc.ac.uk/

icon company @stfc Science and Technology Facilities Council, UK icon location Sci-tech Daresbury, Warrington, UK icon location I specialise in the portable performance of scientific codes, particularly those in the earth-system modelling domain.

Results 748 comments of Andy Porter

FFSL transport in LFRic depends on a kernel that loops over halo cells

> I thought I'd just add another comment here, as there is a whole category of kernels that currently have psykal lite implementations because they loop to some depth in...

Pass read-only arrays of scalars to kernels

Priority wise, how does this compare with the 'stencils for domain (i-first) kernels' issue @andrewcoughtrie and @TeranIvy ?

Pass read-only arrays of scalars to kernels

Thanks both :-)

[LFRic] Support for user-defined kernels that operate on DoFs

Following on from our chat in the telco this morning, it sounds as though the random-number initialisation functionality needed for PSyAD should also be a dof-wise kernel. In fact, for...

[LFRic] Support for user-defined kernels that operate on DoFs

#2556 only added the documentation, not the implementation.

Incomplete variable access information for LFRic kernels, e.g. in LoopFusion

I'm probably missing something here but can't we use the kernel metadata - that's what it's there for?

Incomplete variable access information for LFRic kernels, e.g. in LoopFusion

In short, the `STENCIL` and function-space information in the metadata will tell us about horizontal dependencies. Certainly, there's no concept of 'North'/'South' etc.

Create PSyIR kernel arguments from new metadata classes (closes #2078)

There's quite a long way to go to get the tests to pass flake8 - they're very much a WIP with variables assigned to but never used.

Create PSyIR kernel arguments from new metadata classes (closes #2078)

Following what Rupert wrote in the description of this PR, I've modified gen_kernel_stub so that it uses the new functionality in this PR rather than the old, fparser1-based version. This...

Where statement array syntax is limited

While tackling reviewer comments on #1518, it was suggested that the `_array_notation_rank` method in the frontend ought really to be associated with a PSyIR node. I started to do this...