Implement AlignedVector::shrink_to_fit() #15974
Conversation
|
/rebuild |
Sure, see the last commit. @bangerth But wouldn't this logic apply to every non- |
| /** | ||
| * Flag indicating if replicate_across_communicator() has been called. | ||
| */ | ||
| bool replicated_across_communicator; |
There was a problem hiding this comment.
Should we guard this with ifdef. Similar as we do in FEEval?
There was a problem hiding this comment.
In FEEvaluationData we don't guard the declaration of the variables but all occurrences in definitions (also in asserts, initialization in constructors). Done.
There was a problem hiding this comment.
Odd that the compiler does not warn that the variable is not initialized.
There was a problem hiding this comment.
I would leave it like this if the compiler does not complain, otherwise, the size of the vector is different for release and debug because of the additional member.
bergbauer
force-pushed
the
shrink_to_fit
branch
from
dbe653f
to
bb17bfb
Compare
| DeclExceptionMsg(ExcAlignedVectorChangeAfterReplication, | ||
| "Changing the vector after a call to " | ||
| "replicate_across_communicator() is not allowed."); |
There was a problem hiding this comment.
I have introduced this exception message which we can use for all non-const functions in a follow-up.
| # ifdef DEBUG | ||
| Assert(replicated_across_communicator == false, | ||
| ExcAlignedVectorChangeAfterReplication()); | ||
| # endif |
There was a problem hiding this comment.
I guess we don't need to guard this. Assert is only performed in debug mode.
There was a problem hiding this comment.
In FEEvaluationData the Asserts are also guarded by #ifdef DEBUG. I can remove it if it is not necessary.
There was a problem hiding this comment.
OK. Let's keep it this way to be consistent.
|
@bangerth Any final comments? The requested assert has been added. |
|
Ping I want to use this to release memory after mapping compression in |
No description provided.