profBug

[Henry Cejtin]

Excellent.  This looks very good.

I  agree that if you don't specify -s it can get it wrong, but the problem is
that if you clump all the static routines together in `<static>',  it  really
isn't telling you anything useful.

It's  interesting that you prefer writing the stuff bottom-up.  I have always
really liked top down.  Way way back I remember being pained by the bottom-up
order  in  Pascal.   I  can  see why ML uses bottom-up, since everything is a
`let*', but it still hurts.  I use structures to fix some of it, but lots  of
times I would just like something like Haskell's `where'.  Oh well.

One  thing I'm still not completely clear on with the profiling.  What is the
notion of the -d stuff.  Is the default (`-d 0') supposed to be  at  the  CPS
function  level and everything above that only mapping to the assembler code?
Have we abandoned trying to relate the profiling to the CPS stuff at all?  It
seems that currently the profiling stuff only relates to the assembly code we
generate.  (Of course this could just be my out-of-date version of MLton.)
What exactly are the higher -d values supposed to show?  (I know it is more
detail, but I'm trying to ask if there is any concept of what more I get.)