[MLton] cvs commit: sped up output1 a lot

Henry Cejtin henry@sourcelight.com
Fri, 2 Jan 2004 12:23:51 -0600


I am really amazed that things are so fast now.  I will build up a MLton from
the CVS, but can you quickly tell me if the common case on input1 and output1
are now being inlined?  It would be amazing if it was this fast and they
were not.  Also, in that case, having them be inlined would almost certainly
speed things up by quite a bit more.  In my test I forced them to be inlined
by simply calling having the buffer fill code be done by a separate function,
and then calling that function 100 times in top-level code, in a branch of an
if which would never be taken.  The idea was that that was enough to keep the
inliner from inlining that function.