-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix tests #47
Fix tests #47
Conversation
all the |
the contribution is very much appreciated, of course. |
loc_buf was completely uninitialized, while the second and third loop nests are checking that the values are 1.0 + rank. With luck it would be zero, and thus the actual - expected > 1e-10 check would pass (since the difference is then negative). With less luck the check would spuriously fail for the part that is not overwritten by the MPI_Get. Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
In case the actual value is zero and the expected value is positive, the check would spuriously succeed. Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Now that it is fixed, we can make it actually perform strided accesses. Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
The initialize needs to be 1.0+rank, the second and third loop nests actually check that (to check that MPI_Get properly intermixes values reception, not overwriting everything). I have improved the log message to mention that. |
i recall i had to do a release for debian long ago, or something like that, because they don't just pull from git HEAD. do i need to do that again? |
Debian can cherry-pick fixes when they are committed upstream |
(I plan to do it actually) |
ok, cool. i think it's good now but please let me know if you need anything else. |
The
test_mpi_indexed_gets
test is currently failing spuriously in debian unstable due to an uninitialized value, e.g. https://ci.debian.net/packages/a/armci-mpi/testing/amd64/44486187/ , preventing the newer mpich version from migrating to unstable.