[Users] local and global reduction handles
Roland Haas
rhaas at illinois.edu
Tue Feb 2 12:19:21 CST 2021
Hello all,
Cactus's reduction interface provides two different set of reduction
operations:
1. reduction of grid variables (grid functions, grid arrays) that are
identified via their variable index
2. reduction of local arrays that are identifies using the local memory
address
These two operations use different reduction operator handles returned
by the CCTK_ReductionHandle (grid variables) and
CCTK_ReductionArrayHandle (local arrays) functions, respectively.
In the past the numerical values returned by these functions often
happened to be identical for identical string input (eg "maximum"
might be handle number 1 in both functions).
This was, however, an accident due entirely to the order in which the
individual operator names for each type of reduction were registered
with Cactus.
As of Cactus commit 32def5eb "Cactus: use different handle ranges for
old-style array and grid reductions"
https://bitbucket.org/cactuscode/cactus/commits/32def5eb2030151b15753e040470c956ef757c29
this is no longer the case and the numeric handle values
intentionally differ.
This may make existing code that incorrectly used the handles to fail
with errors reported about invalid operator handles or unknown
operators (depending on the level return code checking done).
The common error case is that CCTK_ReductionHandle was used with
CCTK_ReduceLocalScalar which must be fixed by using
CCTK_ReductionArrayHandle.
Please see ticket #1001:
https://bitbucket.org/einsteintoolkit/tickets/issues/1001/distinguish-between-local-and-global
for details.
Yours,
Roland
--
My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from http://pgp.mit.edu .
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
Url : http://lists.einsteintoolkit.org/pipermail/users/attachments/20210202/59b9b5c7/attachment.bin
More information about the Users
mailing list