[MLton-devel] MLton and profiling

[Joe Hurd]

On Mon, 10 Feb 2003, Stephen Weeks wrote:

> > I certainly prefer it the way it is now.
>
> Sorry, I couldn't tell whether this meant you prefer 20030130 or
> 20030209.

Sorry: I meant I prefer 20030209, the latest version.

> Are you saying that with 20030209 you are seeing only one node for the
> 'o' function?  That certainly wasn't the intent, and I don't see how
> it could happen.

Actually, it turned out to be a fluke. I use the following sequence of
commands:

mlton -basis 1997 -profile time -profile-stack true benchmark.sml
./benchmark
mlprof -show-line true -thresh 1 benchmark mlmon.out

and only one occurrence of 'o' made it above the threshold. If I lower
the threshold to 0, then I get several occurrences (but then dot goes
off to lunch creating the postscript call graph).

> Keep track of the lexical nesting of function names, and print inner
> names as a sequence of function names separated by "."s.  For example,
> your f above would appear as sort.f, and your anonymous function would
> appear as obj_order.<anon> (and of course the source position would
> still be there if you want it).

This would be great: even better than what I suggested. Of course, I
might still want my original suggestion if these kind of function
names are so long that it makes nodes in the call graph look ugly :-)
But if all the information is present I can at least hack the dot
file myself to prettify it.

Cheers,

Joe




-------------------------------------------------------
This SF.NET email is sponsored by:
SourceForge Enterprise Edition + IBM + LinuxWorld = Something 2 See!
http://www.vasoftware.com
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel