search

ARM-software / synchronization-benchmarks

Collection of synchronization micro-benchmarks and traces from infrastructure applications
Other
38 stars 36 forks source link

Should we warmup the thread before lock testing? #9

Closed zoybai closed 6 years ago

zoybai commented 6 years ago

The testing time for 1 ~ 8 threads is so small (0.1~0.5ms) that we saw a large skew (lock acquire average latency in ns) between 1 thread and max threads for lh_empty test.

lh_empty is overhead test which should yield stable lock acquire latency. We suspect no enough warmup time caused this.

geoffreyblake commented 6 years ago

Pull request has been closed.