[MLton] IntInf implementation questions

[Henry Cejtin]

Re IntInf multiply, as I recall, the time it takes for a 32*32->64 multiply
is the same as for a 32*32->32 multiply on the x86, so if it could be done
inline (no call to a C routine), it might still be faster to do the big digit
multiply and then check for the right kind of overflow.  Probably not a big
win unless you are frequently multiplying smalls with a product that doesn't
fit.

Re dontInline, is this exported to any thing any where?  There are definitely
times when it would be a useful hack for speed.  The particular case I know
where I did something like this by `hand' was in a kind of TextIO version
where I did not want the buffer under/over-flow case to be inlined because
then it would inline all the input1 calls.