Rupert Ford
Rupert Ford
Assuming we are not going to get the side-by-side working for a while, an alternative to the checksum-all-invokes option would be to gradually narrow down the region e.g. do per-timestep...
sum2code also does this I think but replaces the final sum as it is not sum2sum! So this functionality could be used here or extracted from here.
Creating a new branch which will modify the adjoint metadata appropriately. This branch will create a new `ArgOrdering` class which uses the new metadata to call methods representing arguments in...
Created [1772_modify_adjoint_metadata](https://github.com/stfc/PSyclone/tree/1772_modify_adjoint_metadata)
Thanks for the heads up @hiker. As this is a new base class, I can develop it independently and happily introduce my ownr bugs :-). When we start working on...
That's a good point @arporter. The same is true for code with stencils. That would be a subsequent PR I think, but I'm not sure whether the Met Office's suggesting...
Just updated to use the new `ArgIndexToMetadataIndex` class to check this works OK, which it does (for the test case). Full implementation, tidying etc will be done once the new...
gocean kernel metadata has been updated. Closing this issue.
I think that shape also supports a tuple of values that allows both lower and upper bounds to be set.
Where possible, the current version of PSyclone extracts the precision of variables from the algorithm layer in LFRic. This information is required for declaring variables appropriately in the PSy-layer. Whether...