Thu, 28 Apr 2005 11:09:20 -0500
I don't know about your examples, but in general the win from having small
integers not require any allocation or indirection is VERY large.
I wouldn't think that using a special-case big-times-small multiply would
buy much in MLton. The reason is that it is certainly the case that the
gmp multiply will test for this case any way, so the only speed up would
come from avoiding the test in that code. This is going to be nil except
perhaps in the case where the big number isn't very big (so that the whole
operation is fast and the percentage of time used in the test is non-trivial).