Date: Mon, 5 May 1997 18:05:44 +0200 (MET DST) From: Wolfgang Helbig <helbig@MX.BA-Stuttgart.De> To: griepent@wias-berlin.de Cc: bugs@FreeBSD.ORG Subject: Re: Underflow floating point exceptions under FreeBSD 2.1.5 Message-ID: <199705051605.SAA20925@helbig.informatik.ba-stuttgart.de> In-Reply-To: <9705051344.AA16083@pascal.wias-berlin.de> from "griepent@wias-berlin.de" at "May 5, 97 03:44:49 pm"
index | next in thread | previous in thread | raw e-mail
> Hello, FreeBSD experts! > > I am going to use numerical applications on a X11 Window desktop and > I want to work with the nice and stable FreeBSD system at home. > To stay compatible with an institute I want to use a numerical > package written in ANSI-C which works flawlessly on DEC and SGI machines. > I was able to compile (gcc-2.6.3) the package on several Intel > platforms (from 486 DX to PentiumPro) under FreeBSD-2.1.5. > > Trying to run it there always occured strange underflow floating > point exceptions in code fragments computing mathematical > functions like > > sqrt, cos, log > > with harmless arguments of type double in the range of 1.0e-20 to 1.0e+20, > which should be normally no big deal in the IEEE 754 world! > > My first idea was to recompile the code with gnu debugger support > enabled and no optimization options. So I have got the same > underflow floating point exceptions, but now more localized, namely > in the above mathematical functions contained in the shared library > > /usr/lib/libm.so > > My next idea was a little bit more successful: > > recompiling the libm sources with gdb support enabled. > > Building the numerical package with that new math library and setting > appropriate breakpoints in the sqrt, log and cos function calls of the > library I was able to localize the underflows in the loops which > iteratively compute the values of that functions. > > Remembering a similar looping problem (many years ago on other platforms) > I have decided to > > recompile the libm sources with no optimization! > > Again building the package with that math library the code is running now > and we get the same numerical results and performance as on the DEC or SGI > machines! > > Now my questions: > > (1) Let me assume that the original libm shared library was compiled > also with the gcc-2.6.3 compiler. > Is there a (known) optimization bug in the gcc-2.6.3 compiler? Your assumption is correct. And there are known (and unknown :-) ) optimization bugs in gcc-2.6.3. That's why virtual all system software in FreeBSD is compiled with the lowest "optimization" flag -O and not with the even buggier -O2 etc. The bug you described is unknown (at least to me). > (2) What about the other system libraries? Perhaps I have to be aware > to run in trouble in a similar manner! > (3) Is there any hope to get rid of that problems in one of the next > versions of FreeBSD or newer gcc versions? I think so. FreeBSD 2.2.X and -current is compiled with gcc-2.7.2, but the bug might have survived the version switch, so I would ask you to send me a minimal C program that exposes the error you were experiencing. Thanks in advance, Wolfganghome | help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199705051605.SAA20925>