On Mon, Oct 12, 1998 at 11:24:06AM +0100, Markus Kuhn wrote:

That's very interesting. Under Linux/Pentium_II (cpu=686, model=3, vendor_id=GenuineIntel, stepping=4, gcc 2.7.2.3) the effect does not show up immediately. I had to enforce using the software FP-library with gcc -msoft-float to reproduce this effect, because without this option with the hardware FPU I get the correct result (and this is apparently not thanks to Intel's internal 80-bit arithmetic, because I did use gcc -ffloat-store to enforce 64-bit rounding).

Have you looked at the actual assemby output? x86 is the worst platform to "prove" something about double arithmetic. Unless you set the cpu to double rounding mode (say, using __setfpucw), therefore invalidating long double, nothing will help. -ffloat-store does not do what you want in every case.

There seem to be interesting differences between IEEE and IEEE.

No, but between IEEE and x86 ;(

A somewhat related question: Given a floating-point value X, is there in C9X a portable way to get the next smaller and the

You could (in C90 as well) try to mess around with frexp/ldexp and/or scalb. The draft c9x standard (as well as xopen-extended) seems to contain a "nextafter" function: nextafter (x, y): Return X + epsilon if X < Y, X - epsilon if X > Y. -----==- | ----==-- _ | ---==---(_)__ __ ____ __ Marc Lehmann +-- --==---/ / _ \/ // /\ \/ / pcg@goof.com |e| -=====/_/_//_/\_,_/ /_/\_\ --+ The choice of a GNU generation | |