Closed juanlucasrey closed 2 weeks ago
Also worthy of notice, I had to add erfc_op at the end in ADOL-C/src/oplate.h. If I add it right after erf_op, the build fails.
The last commit fixes this. I don't know why those numbers are there, but I updated them to the current order and then it builds.
Apart from the comment, looks good. Thank you very much!
Could you provide us the tests? I would like to add them to the boost-test
file.
Apart from the comment, looks good. Thank you very much!
Could you provide us the tests? I would like to add them to the
boost-test
file.
I tested with my own repo. I have not tried to build with boost yet, so I have not ran the test suite. I can look into it.
This implementation has been tested and working with:
-tensor_eval -tape-less forward order 1 -fov_forward -zos_forward + fos_reverse
in particular it has NOT been implemented with CUDA, since I cannot it test at the moment.
also, the "(int*)" changes are to avoid clang warnings.
Also worthy of notice, I had to add erfc_op at the end in ADOL-C/src/oplate.h. If I add it right after erf_op, the build fails.