I propose the name "Falk Diagram" for this type of cycle-by-cycle diagram with perf counter values on the y-axis. I first proposed it here, buried in a footnote: https://travisdowns.github.io/blog/2020/01/17/avxfreq1.html#fn:falk … Even a very primitive form of falk diagrams (in userspace, without re-run) was very useful.
Sushi Roll: A CPU research kernel with minimal noise for cycle-by-cycle micro-architectural introspectionhttps://gamozolabs.github.io/metrology/2019/08/19/sushi_roll.html …
-
-
-
It's hard to overstate how cool these are--they looked like black magic at first. I hope Sushi Roll gets released eventually, cuz I'd really like to do some experiments (eg investigate OoO schedulers) but I'm not quite prepared to write a kernel
- Još 7 drugih odgovora
Novi razgovor -
-
-
Cool stuff! Did you look into Intel PT too?
-
I've used it before for code coverage, but I'm not the biggest fan. The cost/complexity of parsing the traces often seems prohibitive to prototyping or fuzzing.
Kraj razgovora
Novi razgovor -
-
-
That was a great read. Stupid question but, how do you get the data out of the machine, to analyze/plot it? (I assume it does not have a filesystem or a network stack?).
-
Serial or network (udp and tcp stack for the NIC I use)
Kraj razgovora
Novi razgovor -
-
-
Amazing blog post! Small nit: it’s voilà, not walla
-
I like walla. Let's make it walla.
- Još 1 odgovor
Novi razgovor -
-
-
Fantastic blogpost Brandon!
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
Tweet je nedostupan.
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.