[MLton-devel] upcoming public release

Stephen Weeks MLton@mlton.org
Thu, 30 Jan 2003 23:31:17 -0800

> If the cost is really the allocation time (not GC time) then it seems like
> an argument for loop unrolling.  I.e., if I unroll a list-reverse loop once
> then the cost for the check if GC is needed goes down by a factor of 2.
> Of course the heap entries still have to be initialized, but I'm just wondering
> if some better optimization might not solve the problem with no change to
> user code.

There may be some optimization (like unrolling) that cuts the total
time down by half.  But whatever it is, there is still another half to
be had by eliminating the extra reverse.  I really don't see any
compiler optimization as relevant here, except for the one that we
have mentioned before that does the reverse in place using mutation.
But even that will still be slower than doing nothing.

This SF.NET email is sponsored by:
SourceForge Enterprise Edition + IBM + LinuxWorld = Something 2 See!
MLton-devel mailing list