Lean on compiler #215
Lean on compiler #215
Conversation
| @@ -23,7 +23,6 @@ program example | |||
| real(wp), dimension(:,:), pointer :: out_data | |||
| real(wp), dimension(n,m) :: expected | |||
| integer :: tensor_layout(2) = [1, 2] | |||
| integer :: i, j | |||
There was a problem hiding this comment.
Unused variables.
| do i = 1, ndims | ||
| if (i == 1) then | ||
| strides(layout(i)) = 1 | ||
| else | ||
| strides(layout(i)) = strides(layout(i - 1)) * tensor_shape(layout(i - 1)) | ||
| end if |
There was a problem hiding this comment.
The original version is a zero loop if ndims = 1.
There was a problem hiding this comment.
Good spot!
I suppose it did what was intended, but the rewrite is much more explicit and doesn't leave us open to unexpected intentions in future.
| @@ -38,7 +38,7 @@ subroutine main() | |||
| integer, parameter :: out_layout(out_dims) = [1, 2] | |||
|
|
|||
| ! Path to input data | |||
| character(len=100) :: data_dir | |||
There was a problem hiding this comment.
Number of characters was inconsistent with number of characters for command line input.
There was a problem hiding this comment.
Thanks @jwallwork23 all looks sensible to me and some good catches.
Slightly worrying that the C precision was working in the old version...?
Happy for you to merge.
| do i = 1, ndims | ||
| if (i == 1) then | ||
| strides(layout(i)) = 1 | ||
| else | ||
| strides(layout(i)) = strides(layout(i - 1)) * tensor_shape(layout(i - 1)) | ||
| end if |
There was a problem hiding this comment.
Good spot!
I suppose it did what was intended, but the rewrite is much more explicit and doesn't leave us open to unexpected intentions in future.
Agreed. |
jwallwork23
added
the
autograd
Building in debug mode with gfortran 11.4.0 on Ubuntu 22.04, I came across several warnings, most of which are fixed in this PR. They are mostly minor but the ones to do with overloaded operators are an oversight of mine from #139.