[RFC] Test the trajectories and proxy forces using stub#852
Conversation
|
|
||
| #include <iostream> | ||
|
|
||
| #define STUB_OUTPUT_FORCE |
There was a problem hiding this comment.
Are there strong reasons not to make this a runtime flag passed to read_frame_xyz? It would seem simpler to me, and I don't see a critical performance consideration there.
There was a problem hiding this comment.
There is already an optional argument for the output prefix. I wonder if I add another one, it may become difficault to maintain in the future due to added complexity, if a 3rd argument parser is not used. What's the best way to add such a runtime flag?
There was a problem hiding this comment.
I think you're right. With 3 or more arguments, it's hard to remember how to use them in the correct order.
Maybe we can use the gflags library or something similar to allow for flags?
There was a problem hiding this comment.
There are many choices. Personally I would like to use CLI11 since it is header only and can be used with a single header (https://github.com/CLIUtils/CLI11?tab=readme-ov-file#install). I am not familiar with gflags but it looks like linking to the library is required.
There was a problem hiding this comment.
Sure, I think that qualifies as "something similar" :-)
A header only library that is only used by our command-line tool should be no trouble at all.
There was a problem hiding this comment.
Done in 80d012e.
Currently, the tests failed because
- The Clang++ 3.4 is too old;
- The Oracle/Sun compiler is too old.
I suggest to give up the two outdated and buggy compilers, and use -Wshadow in GCC and Clang. However, this is up to @giacomofiorin .
There was a problem hiding this comment.
I suggest to give up the two outdated and buggy compilers, and use
-Wshadowin GCC and Clang. However, this is up to @giacomofiorin .
I checked: we do have this already #557 Although it's not quite the same thing as the Sun checks, we are probably quite close to abandoning C++11 as well, which would rule that out.
Dropping all CentOS 7 stuff in an upcoming PR.
HanatoK
force-pushed
the
test_stub_output
branch
from
cb833fa to
c4ec63a
Compare
giacomofiorin
force-pushed
the
test_stub_output
branch
2 times, most recently
from
abc6f6c to
5c00187
Compare
The test data are all NaNs. Spiff fails to compare -nan and nan.
This reverts commit 18dd35e. Giacomo said that the oneapi.sh had been updated and the issue should be solved without changing test-library.yml.
HanatoK
force-pushed
the
test_stub_output
branch
from
5c00187 to
c9d7089
Compare
3 participants
This PR tries to compare the stub test results with reference files in
AutoDiff, and also outputs the stub proxy forces for testing. The reference files will be useful for testing the GPU stub proxy in #826.