[Mono-list] IEEE conversions

Morten Sylvest Olsen mortenolsen@mail.com
Mon, 11 Mar 2002 12:18:06 +0100 (CET)


Does anybody know whether the CIL conv.r(4,8) instructions issued on the
stack value of an arithmetic operation imply that the floating point
result is converted from internal precision to IEEE precision? It doesn't
explicitly say so in part III.

I guess x86 is about the only architecture where this matter, to implement
strict IEEE semantics would then necessite a move from the FPU stack to
memory after each operation, and since the CLR does not explicitly support
this, conv seems the only way to force this?

- Morten

One way to stop a runaway horse is to bet on him.
email: morten@bergsoe.k-net.dk,c958496@student.dtu.dk
addresse: Bergsøe Kollegiet v. 1105 2850 Nærum