The pictures in the article don't work (at least for me).
Here is the wayback machine snapshot where everything displays correctly.
https://web.archive.org/web/20160718172225/http://gernotklin...
Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
11–20 of 29 posts
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#12Also worth considering is Oprofile: https://en.m.wikipedia.org/wiki/OProfile
I used to use oprofile, but I don't think it works on current kernels.
So I use "perf stat" now. :)
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#13No love for VTune?
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#14Julia Evans has this really nice ezine regarding Linux debugging tools: https://wizardzines.com/zines/debugging/
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#15No love for VTune?
It was great on physical hosts but could not get it working in EC2 VMs
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#16gcc has an address and thread sanitizer built in these days.
gcc -fsanitize=address ...
gcc -fsanitize=thread ...
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#17Also worth mentioning here is perf[1], which is great for low overhead profiling. Also, perf profiles can be turned into profiles compatible with GCC and LLVM PGO to build optimized binaries based on production runs, using autofdo[2]. In my use case, the instrumentation overhead was too high to use regular profiling on production workloads. [1]: https://en.m.wikipedia.org/wiki/Perf_%28Linux%29 [2]: https://github.com…
For anyone looking to use perf, Brendan Gregg's page on it is the best resource I know of: http://www.brendangregg.com/perf.html
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#18A problem with google perftools is the `SIGPROF` signal used in sampling will interrupt polling such as used in ZeroMQ. Otherwise, it is a good tool in the toolbox.
that is/was a problem with zeroMQ, not gperftools:
27 SIGPROF terminate process profiling timer alarm (see setitimer(2))
Additionally, it seems fixed in zeroMQ for some things already in 2016... so i doubt thats still valid issue.
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#19Also worth mentioning here is perf[1], which is great for low overhead profiling. Also, perf profiles can be turned into profiles compatible with GCC and LLVM PGO to build optimized binaries based on production runs, using autofdo[2]. In my use case, the instrumentation overhead was too high to use regular profiling on production workloads. [1]: https://en.m.wikipedia.org/wiki/Perf_%28Linux%29 [2]: https://github.com…
Fun fact: the pprof tool that comes in gperftools can read perf data files, so you can use them together if you prefer pprof’s reporting.
Re: Evaluation of CPU profiling tools: gperftools, Valgrind and gprof (2015)
#20gcc has an address and thread sanitizer built in these days. gcc -fsanitize=address ... gcc -fsanitize=thread ... They perform better than valgrind.
This article refers to valgrind's profiling features (callgrind) and not its more common/popular 'memcheck' feature.
Sanitizers and memcheck are unrelated to the profiling discussion here.