> > Do you know why it doesn't work? I tried a free bsd machine the other
> > day and had the same problems.
[quoted text clipped - 3 lines]
> came up with a solution, which leads me to conclude that gprof has big
> problems on Mac OS X 10.4.
I assume that gprof measure cpu time based on who was at bat when the
100-Hz (or 1000-Hz) interrupt ticked, probably mediated by a signal into
the process being monitored. It sounds like this signal isn't
happening, for whatever reason. Probably a setup issue.
Joe Gwinn
bdbkun@gmail.com - 27 Jul 2006 19:20 GMT
Gprof does not interrupt the process being monitored to determine
execution time. GCC introduces hooks at compile time to count how many
times the PC goes into each function, and how long it stays in each.
Gprof is just used to summarize and display the data.
GCC profiling operates in user-space with no specific requirement in
terms of kernel modules, signaling, etc.
Either the function call data is properly recorded as per the hooks put
in place by GCC, but then it is not properly displayed by gprof; or the
time data is not properly recorded by the GCC hooks.
Either way, this is a major pain for developers used to Unix systems.
Shark may be nice, but when you don't know it, when you already know
gprof and need to work in a console environment... It's a major
compatibility issue.
> > > Do you know why it doesn't work? I tried a free bsd machine the other
> > > day and had the same problems.
[quoted text clipped - 10 lines]
>
> Joe Gwinn