+254
−94
File changed.
Preview size limit exceeded, changes collapsed.
+45
−0
+58
−0
File added.
Preview size limit exceeded, changes collapsed.
Loading
Current thread-level analysis can hide a lot of data races in OpenMP tasking codes, when tasks are randomly scheduled to the same thread, but even worse when unrelated task dependencies introduce random synchronization between different threads. This patch introduces a new analysis mode for Archer activated at runtime by exporting ARCHER_OPTION=tasking=1, where OpenMP tasks are explicitly mapped to TSan fibers. Under review as #74646
File changed.
Preview size limit exceeded, changes collapsed.
File added.
Preview size limit exceeded, changes collapsed.