[MLton] Experimental 64-bit binary package (& SVN sources)
Matthew Fluet
fluet at tti-c.org
Wed Mar 7 08:23:32 PST 2007
skaller wrote:
> testing real
> 666,667c666,667
> < 11.59195423
> < 1.144729972
> ---
>> 11.59195518
>> 1.144729853
...
> difference with -type-check true
These are benign.
We generally ignore such differences on x86, because it all depends upon
how much of the floating-point computation gets done at 80bits, which in
turn depends upon too many things outside our control (e.g., how libm is
implemented, spilling of floating-point registers).
As I understand it, x86_64 SSE2 floating-point instructions compute at
32-bit and 64-bit floating-point precision, so perhaps we should take
the x86_64 results as "correct" and ignore differences on x86.
More information about the MLton
mailing list