[MLton-devel] cvs commit: (almost) full support for Real32

Matthew Fluet fluet@cs.cornell.edu
Sat, 26 Jul 2003 08:02:21 -0400 (EDT)


> I remember many problems caused by the failure to handle the conversion to
> infinity in things like Mathematica.  I guess as long as Real32.isFinite
> and friends do the right thing then it is ok.  It is vital though that the
> conversion to 32 bits not get optimized away here.

Then would you suggest that the right thing for the x86-codegen to do when
it sees a conversion from 64 bits to 32 bits is to force the memory write,
thereby forcing the conversion?

> Also I am a bit worried that I will end up with objects that are not
> Real.== to themselves.  I guess that even the 80 vs. 64 bits hack can
> cause that.

It's worse than that.  You could have x and y that are very close (but
unequal) at 80 bits, for which Real.==(x, y) returns false and then later,
after they have both been written to memory (into 64 bits), see
Real.==(x,y) return true, because they are equal at 64bit precision.



-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel