How does his KUTrace Linux kernel patches differ from the existing tracing infrastructure? This talk [1] seems to suggest that the reason his KUTrace is so fast is because the other tracers collect more information.
How does the tracing capabilities here compared to what the Windows WPA subsystem can offer? Does it have less overhead for collecting such traces?
Sampling profilers like Superluminal claims to be able to recognize very small delays. For how many cases do you need something like KUTrace and when is Superluminal enough?
What more besides the ideas in Richards ACM Queue article are present? https://queue.acm.org/detail.cfm?id=3291278
How does his KUTrace Linux kernel patches differ from the existing tracing infrastructure? This talk [1] seems to suggest that the reason his KUTrace is so fast is because the other tracers collect more information.
How does the tracing capabilities here compared to what the Windows WPA subsystem can offer? Does it have less overhead for collecting such traces?
Sampling profilers like Superluminal claims to be able to recognize very small delays. For how many cases do you need something like KUTrace and when is Superluminal enough?
[1] https://www.youtube.com/watch?v=UYwWollxzAk I felt a little uneasy watching it due to Richard being so disagreeable to anyoone asking him questions