Subject: Re: standards/5150: 1.3 i386 fails "paranoia" IEEE 754 math test
To: John Franklin <john@iptimaging.com>
From: Tim Rightnour <root@garbled.net>
List: netbsd-bugs
Date: 03/12/1998 16:22:45
On 12-Mar-98 John Franklin spoke unto us all:
# >not the architecture at all..  If you know anyone with a commercial x86
# >compiler, have them run it and see if they get the double rounding bit.
# 
# Actually, I do at work.  I've got a copy of Symantec C++ 7.5 and several
# versions of MS Visual C++.  Only 1.5 will make a DOS executable, tho.
# 
# I tried Symantec C++ 7.5 producing a DOS (16-bit) executable and it came up
# with the same results all four times (for debug/release, and Fast Floating
# Point/inline x87).  They're in symantec.txt.
# 
# MSVC 1.5 (they're last 16-bit DOS compiler) got the same "FLOATING-POINT
# ERROR 8"s as for the Alpha, but also got a _POW out of range (see msvc.txt)
# 

I've been told that Microsoft C compilers are known to have problems with this
and other IEEE 754 tests.  I find it quite odd though that the sparc works
flawlessly.  I will dig around and compile it on turbo C if I get an
opportunity..  Pity we can only claim we are a little better than M$ in this
case.. ;)

---
Tim Rightnour    -  root@garbled.net
http://www.zynetwc.com/~garbled/garbled.html