Andy Porter

Results 789 comments of Andy Porter

I don't really have a problem (in the sense that I have it all working now), e.g.: real, dimension(2:8) :: actual call my_sub(actual) subroutine my_sub(x) real, dimension(3:) :: x x(3)...

We need this functionality in order to get useful coverage of the use of ASSOCIATE in NEMOVAR. We should maybe have a go at prioritising Issues once Sergi is back.

Currently, `Reference` has a `datatype`. We need to extend this to: - [x] Operations - [ ] IntrinsicCalls - [ ] Function calls.

I'm thinking about Operations first. Everything seems fairly straightforward unless the operands are of the same intrinsic type (e.g. real) but of different kinds. The standard says: ![image](https://github.com/stfc/PSyclone/assets/17409212/e283dc48-bdc6-4958-a31e-666c88783685) which is...

Hi @oakleybrunt, the rule for proper Kernels (as opposed to their PSyKAlite implementation) is that they must make no reference to parallelism (which, in this case, means no references to...

In the Builtin itself you will need to ensure that the LHS of the Assignment has the correct precision and that the RHS has a cast. The code you want...

The code generated by PSyclone (and that which is present in the psykalite you are looking at) is specifically designed to ensure that if you run the same code on...

See also `examples/lfric/eg6` - https://psyclone.readthedocs.io/en/stable/examples.html#example-6-reductions

> Questions: What do you mean when you say the reduction reference "doesn't yet use the PSyIR properly"? Why is REPRODUCIBLE_REDUCTIONS set to false? Would anything break if this was...

Well, it's not so much that it doesn't know about precision (I think it can be supplied), it's more that the code that currently uses it will be assuming/hardwired to...