Non-Determinism and Overcount on Modern Hardware Performance Counter Implementations
0 comments · hardware, pdf, performance
- Comments2
0 comments · hardware, pdf, performance
Abstract: Ideal hardware performance counters provide exact deterministic results. Real-world performance monitoring unit (PMU) implementations do not always live up to this ideal. Events that should be exact and deterministic (such as retired instructions) show run-to-run variation and overcount on x86_64 machines, even when run in strictly controlled environments. These effects are non-intuitive to casual users and cause difficulties when strict determinism is desirable, such as when…
Story details from Lobsters ↗ · linked article on web.eece.maine.edu.
Non-Determinism and Overcount on Modern Hardware Performance Counter Implementations was first seen on Lobsters · Hottest on Nov 1, 2026. Since then it has appeared in 42 checks on 2 different days. Its best position was #2 of 25, reached on Nov 2, 09:51 UTC. It dropped off the list after Nov 2, 12:51 UTC.
Position on Lobsters · Hottest at every check (lower is better).
Best rank per day (UTC).
Platform metric as reported by Lobsters.
A few top-level comments from Lobsters, shortened. Read the full thread on Lobsters.
Funny, I was just writing a blog post on this and how it uses ex_ret_cond on AMD for estimating instructions that have passed in Rewind [1]. This lines up with everything that Rewind applies for AMD/Intel. [1] http://rewindvm.dev/
wrote it: https://fzakaria.com/2026/11/01/a-clock-made-of-branches
Archived daily, weekly and monthly lists that include this entry.
Not at the moment. It was last on the list on Nov 2, 12:51 UTC, after 2 days of trending.
#2 on Lobsters · Hottest, on Nov 2, 09:51 UTC. Trendoday checks this list every 30 minutes.
It was first seen on Nov 1, 2026 and has been on the list on 2 days (42 checks) since.