592400 - performance microbenchmark tests should have metric defined as iterations/second

Status: VERIFIED FIXED

(Tamarin Graveyard :: Tools, defect)

Description

[Chris Peyer]

The microbenchmark tests define their metric as "v8".  This is incorrect as the metric is actually iterations per second.  A new metric should be defined for these tests.

Comment 1

[Chris Peyer]

I propose using "iter/s" or "iter/sec" as the metric name.  When run, a desc string will also be printed that can have the full explanation.  Thoughts?

Comment 2

[Lars T Hansen]

"iterations/second".  The abbreviated forms save screen and disk space but require me to think harder.  I think that's the wrong tradeoff.

Comment 3

[Chris Peyer]

Attachment: change microbenchmark metric to k_iterations/second

Decided to go with k_iterations/second as that is what the value truly is.

When running, metric will print out:


Metric: k_iterations/second  : thousands of iterations per second
Dir: asmicro/
...

Comment 4

[Lars T Hansen]

Comment on attachment 471252 [details] [diff] [review]
change microbenchmark metric to k_iterations/second

It really is iterations/second, not thousands of iterations per second.  The reason is that time is measured in milliseconds, so the division by 1000.0 is to convert the time to seconds.

If you fix that you can assume R+ without re-review.

Comment 5

[Chris Peyer]

Doh - changed back to iterations/second and pushed to redux:

http://hg.mozilla.org/tamarin-redux/rev/726446364981