profiling a self compile

[Stephen Weeks]

> The  <unknown> times (13.72% of 522.19 CPU seconds, or 71.64 seconds) is time
> spent in the shared libraries.  This either means time actually spent  there,
> or  time  spent in the system as a result of a system call (with the C system
> call in the shared library).  Was your system time actually this large?  This
> strikes me as pretty huge.  Also a lot of time in GC-related areas.  How high
> was the percentage of time spent in the GC?

I just ran a self compile.  Here are the timings (according to both MLton and
time).

MLton finished in 358.03 + 223.03 (38% GC)
573.08user 8.02system 10:23.46elapsed 93%CPU 

So there is very little system time, but the usual 40% GC time.  Could the
unknown be attributable to GC?

> Is it my imagination, or is it true that with the current native back end and
> profiler the output of mlprof seems harder to trace back to source code?

It seems the same to me, especially with -d 0.