7cee60346e
Reading the system clock on every call to opal_progress() is an expensive operation on most architectures, and it can negatively affect the performance, for example of message rate benchmarks. We change opal_progress() to read the clock once per 8 calls, unless there are active users of the event mechanism. Signed-off-by: Yossi Itigin <yosefe@mellanox.com>