From: asstud3@sci.kun.nl (Frits Daalmans) Subject: [Q] fpu-emulator precision (where is float?) Date: Thu, 20 May 1993 12:42:17 GMT
Greetings!
I have a question concerning the floating point emulator in the kernel
(my machine is a 386DX40 without numeric coprocessor and I am using linux
0.99.8, gcc 2.3.3, libc4.3.3)
When compiling the demo programs of the pbmplus package, there are some
routines (for example a FFT routine in ppmforge i believe) that expect
single precision arithmetic (float instead of double) and are *NOT* easily
converted to double (e.g by defining float double ).
The program compiles OK but gives a floating point error (read: crashes) on
execution.
Although this problem came to light in this program which I don't really
need, I can imagine programs where single precision is something I can't
do without, so I have the following questions:
1) Is there a version of the FPU-emu that supports single precision? ("float" in
C), or
2) Is someone working on this, or
3) should I shut up and buy a coprocessor ;-)
You can respond by E-mail to asstud3@sci.kun.nl or
frits@rulglj.leidenuniv.nl
and I will summarize to the net once I find out why the systems where I work
now don't support NetNews :-(
Thanks in advance,
Frits Daalmans (frits@rulglj.leidenuniv.nl)
PhD student Conformational Analysis
Leiden University (but gratuitously still using the Internet services of
the Uni of Nijmegen for the rest of this month)
PS I ***LOVE*** Linux (really!)