Could someone out there run a non-trivial program compiled, linked,
and run with -pg, and see if you get something like this?

Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total
 time   seconds   seconds    calls  ms/call  ms/call  name
 98.63    737.80   737.80                             start
  0.90    744.55     6.74                             mcount
  0.13    745.54     0.99    16487     0.06     0.20  eval
  0.04    745.85     0.31    12197     0.03     0.07  erange
....

In practice, "eval" is running for 95% of runtime; what's this "start" thing
got to do with it?

-s