[MLton] cvs commit: Added support for files larger than 2G

[Stephen Weeks]

> Did you notice any change in speed for the wc-* benchmarks?  IO
> certainly isn't doing a 64bit op per character read, but it will for
> each underlying read.  I imagine it isn't really a problem.

No problems at all.  Here are the only benchmark run time ratios
comparing MLton before (MLton0) and after (MLton0) the change where
the MLton1 ratio was more than 0.05 different from 1.

benchmark         MLton0 MLton1
lexgen              1.00   1.24
mandelbrot          1.00   0.92
peek                1.00   0.75
psdes-random        1.00   0.94

So, the wc's didn't show up at all.