by Geng Wan » Sat, 15 May 1999 04:00:00
I am using egcs-2.91.66 and found exp(-1.0/0.) gives "nan" if any
optimization is turned on. With no optimization, it gives 0.0000. On an IBM
RISC/6000 I have with gcc2.8.1, there is no problem with any level of
optimization. Could anybody tell me how I can let the compiler conform to
IEEE floating point standard but still have other optimizations done?
Where is the best place to find references for this kind of problems?
I am new here. Thanks in advance.
Geng
by Andrew Chu » Sat, 15 May 1999 04:00:00
Try the docs that come with egcsQuote:>Where is the best place to find references for this kind of problems?
--
See http://anderoo.dhs.org/~anderoo/pgp.html for PGP key
It's a sin only if you dwell on the what ifs and the but ifs
1. ANSI/IEEE double-precision floating-point arithmetic?
Where do I find ANSI/IEEE Std 754-1985 double-precision
floating-point arithmetics in Linux?
Carsten.
---
2. finding & removing PEX fonts
5. Linux floating point/IEEE Std.745-1985 compliancy
6. FreeBSD for IBM/Cyrix 8x86PR166MMX
7. IEEE Floating Point significant digits
8. XFree 3.3.2 and ATI @Work AGP card
9. IEEE 64bit floating point libraries for PC
10. 370 Floating Point to IEEE conversion
11. IEEE floating-point arithmetic exceptions
12. significant digits IEEE floating point