nrnmpi_spike_exchange: pass spikein as a pointer#668
Conversation
The function free and reallocate it.
Codecov Report
@@ Coverage Diff @@
## master #668 +/- ##
==========================================
- Coverage 55.30% 55.29% -0.01%
==========================================
Files 107 107
Lines 8703 8706 +3
==========================================
+ Hits 4813 4814 +1
- Misses 3890 3892 +2
Continue to review full report at Codecov.
|
|
Logfiles from GitLab pipeline #20780 (:no_entry:) have been uploaded here! Status and direct links:
|
| free(spikein); | ||
| spikein = (NRNMPI_Spike*) emalloc(icapacity * sizeof(NRNMPI_Spike)); | ||
| free(*spikein); | ||
| *spikein = (NRNMPI_Spike*) emalloc(icapacity * sizeof(NRNMPI_Spike)); |
There was a problem hiding this comment.
I remember having the code review for this bug but couldn’t find the discussion.
I am questioning the SimulationStack CI and how this was not captured there. May be we need to adjust BlueConfig so that there are sufficient spikes to trigger bug like this.
|
Logfiles from GitLab pipeline #20809 (:white_check_mark:) have been uploaded here! Status and direct links:
|
iomaganaris
left a comment
iomaganaris
left a comment
There was a problem hiding this comment.
LGTM
The simulation is now working on CPU and GPU with 1 or multiple ranks
* Pass spikein as a double pointer * Protect spikein with nrn_assert in case it is nullptr CoreNEURON Repo SHA: BlueBrain/CoreNeuron@e265f9d
6 participants
The function free and reallocate it.
This bug has been found by #667