Double precision option for ropp_io_write
In the ropp_io_write the option " output_precision = 'double'" is not working properly. The problem arose when I tried to add the gradient of the cost function to the io output. To solve it(temporary) vlist%data is converted to real in the ropp_io_write_ncdf_put.f90 (Line 504).
It is necessary to add an automatic conversion (double->real) to prevent the system to misbehave in case of input which is not real (too small). The ROPP_io_write default should be single precision but the user should have the option to go for double precision.
Change history
(7)
| Cc: |
dave.offiler@… carlo.buontempo@… axel.vonengeln@… added
|
| Milestone: |
→ 1.0
|
| Owner: |
changed from to marq
|
| Version: |
→ 0.8
|
| Status: |
new → assigned
|
| Version: |
0.8
|
| Component: |
ROPP (all) → I/O
|
| Type: |
defect → enhancement
|
| Owner: |
changed from marq to Dave Offiler
|
| Resolution: |
→ fixed
|
| Status: |
assigned → closed
|
ropp_io write output precision option tested at v1.2 to write data (including vlist variables) and working ok. No evidence of 'manual' conversion of vlist%data in ropp_io code (or repository) at line indicated in initial ticket description. Method for assigning precision type is apparently similar for vlist as all other defined variables. Must therefore assume initial problem is fixed. Ticket closed.