https://bugs.kde.org/show_bug.cgi?id=465465

--- Comment #15 from Nick Nethercote <n.netherc...@gmail.com> ---
The existing cache and branch simulations are very simplistic, and about 20
years out of date. I think hardware counters are a much better way of getting
cache statistics. I recently made `--cache-sim=no` the default for this reason.

I understand that the statistics you've added aren't available via hardware
counters. But still I worry that using an unrealistic simulation as the
foundation could lead to inaccurate results.

Another thing: in the example output file you have lines like these:

> 4,496,605,023 (51.80%) 47,791,341,277 (5545.9%) 127,816 (2819.7%) 147,917 
> (23112.0%) 73,046 (1014.2%) 154,413 (203.7%)   8,192 (51.35%)  
> binary_search_test


The percentages in the all the new columns (everything other than `Ir`) greatly
exceed 100%, which seems wrong.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to