hotspot
hotspot copied to clipboard
The Linux perf GUI for performance analysis.
now that we are getting support for trace points, it would be good if we could take this a step further: similar to sched switch in/out events, we might want...
add an "annotate source" view, similar to what other profilers have: - source code with cost aggregated by line, show both inclusive/exclusive cost if possible - assembly code with cost...
The GSYM format has the potential to speed up perfparser processing significantly. See [1] and [2]. Investigate how we can support it in perfparser, potentially by adding support for it...
parsing perf.data can often produce misleading results when anything is amiss. We need to tell the user about such situations by extending upstream error reporting: - elf not found -...
**Is your feature request related to a problem? Please describe.** I would like to analyze a problem on my ASUS Tinkerboard which runs Ubuntu 18.04. On my other system I...
Currently it seems impossible to compare results from two profiling runs. I think the only way to do this is to start two separate hotspot instances and alt-tab. It would...
when dwarf unwinding overflows the stack copy, we should try to stitch the backtrace to a previous more-complete stack
it would be nice to get the ability to attribute the self cost of a given function, or even a whole library, to its callers. Basically this would hide the...
Often backtraces are broken, for example because of missing debug symbols or because the call stack size that was recorded is too small. In the top-down and in the flame...
we can now show off-CPU time, but it would also be useful to deduce the on-CPU time and then finally the on-stack time too