Applying integration of NastyMPI CI build #189
Conversation
| if (nullptr != m_globmem) { | ||
| m_globmem->flush_all(); | ||
| } | ||
| if (nullptr != m_team && *m_team != dash::Team::Null()) { |
There was a problem hiding this comment.
Can m_team ever be a nullptr? It seems to be initialized from a dash::Team reference in all c'tors.
There was a problem hiding this comment.
Currently yes, but it's a pointer, so I check it. Members can be specified by lots of operations and checking for nullptr is more robust than validating every possible member assignment.
Also, it's in Array::barrier() so it's not like the test will have a say in performance.
There was a problem hiding this comment.
... also, most lint tools would otherwise nag about this.
|
@devreal @fmoessbauer @rkowalewski See my comments in #184 Please review and merge. |
|
No objections from my side. |
| local_range[l] = ((dash::myid() + 1) * 1000) + l; | ||
| } | ||
| array.barrier(); | ||
|
|
||
| // Block- and global offset of target range: | ||
| auto block_offset = _dash_size - 1 - dash::myid(); |
|
Looks fine to me! |
No description provided.