cwr@pnet01.cts.com (Will Rose) (03/04/91)
Reference the recent request for Minix floating point code, Peter Housel's
fp code compiles under PC 1.5.10 with no problems that I can find - I used
the following Makefile:
=====================================================================
#
# Makefile for fpp and floating point library - vn 0.3 of Apr 90
#
CFLAGS = -f -F -LIB -DFLOATLIB -D_MINIX -D_POSIX_SOURCE
FPPOBJS = fpp.s atof.s strtod.s strtod_aux.s ldexp.s cff.s cmf8.s \
zrf8.s norm8.s norm4.s
fpp: $(FPPOBJS)
cc -i -o fpp $(FPPOBJS)
fpp.s:
cc -c -F -D_MINIX -D_POSIX_SOURCE fpp.c
CSRCS = fabs.c pow.c exp.c sqrt.c log10.c log.c _poly.c _mult.c \
ceil.c floor.c atof.c strtod.c printf.c fprintf.c sprintf.c \
vprintf.c vsprintf.c vfprintf.c
COBJS = fabs.s pow.s exp.s sqrt.s log10.s log.s _poly.s _mult.s \
ceil.s floor.s atof.s strtod.s printf.s fprintf.s sprintf.s \
vprintf.s vsprintf.s vfprintf.s
GENSRCS = frexp.x modf.x strtod_aux.x ldexp.x
GENOBJS = frexp.s modf.s strtod_aux.s ldexp.s
AUXSRCS = cff.x cfu.x cuf.x cif.x addsub.x cfi.x fat.x trp.x dvf8.x \
ngf8.x mlf8.x cmf8.x zrf4.x zrf8.x norm4.x norm8.x
AUXOBJS = cff.s cfu.s cuf.s cif.s addsub.s cfi.s fat.s trp.s dvf8.s \
ngf8.s mlf8.s cmf8.s zrf4.s zrf8.s norm4.s norm8.s
MINIX1_1SRCS = ret8.x return.x
# MINIX1_1OBJS = ret8.s return.s
LIBOBJS = ${COBJS} ${GENOBJS} ${AUXOBJS} ${MINIX1_1OBJS}
libfp.a: ${LIBOBJS}
ar cr libfp.a `lorder ${LIBOBJS} | /usr/local/bin/tsort`
# This produces zero-length files with the 1.5 make
# The 1.5.5 make doesn't have enough room
# -libpack < $*.x | sed '/^$$/d' > $*.s
.SUFFIXES: .x
.x.s:
-libpack <$*.x >$*.s
clean:
rm -f libfp.a fpp $(LIBOBJS) core
depend:
mkdep fpp.c $(CSRCS)
# DO NOT DELETE THIS LINE -- mkdep uses it.
# DO NOT PUT ANYTHING AFTER THIS LINE, IT WILL GO AWAY.
fpp.s: /usr/include/stdio.h
fpp.s: fpp.c
fabs.s: /usr/include/math.h
fabs.s: fabs.c
pow.s: /usr/include/errno.h
pow.s: /usr/include/limits.h
pow.s: /usr/include/math.h
pow.s: pow.c
exp.s: /usr/include/errno.h
exp.s: /usr/include/limits.h
exp.s: /usr/include/math.h
exp.s: exp.c
sqrt.s: /usr/include/errno.h
sqrt.s: /usr/include/math.h
sqrt.s: sqrt.c
log10.s: /usr/include/math.h
log10.s: log10.c
log.s: /usr/include/errno.h
log.s: /usr/include/math.h
log.s: log.c
_poly.s: /usr/include/math.h
_poly.s: _poly.c
_mult.s: /usr/include/math.h
_mult.s: _mult.c
ceil.s: /usr/include/math.h
ceil.s: ceil.c
floor.s: /usr/include/math.h
floor.s: floor.c
atof.s: atof.c
strtod.s: /usr/include/ctype.h
strtod.s: strtod.c
printf.s: /usr/include/stdarg.h
printf.s: /usr/include/stdio.h
printf.s: printf.c
fprintf.s: /usr/include/stdarg.h
fprintf.s: /usr/include/stdio.h
fprintf.s: fprintf.c
sprintf.s: /usr/include/stdarg.h
sprintf.s: /usr/include/stdio.h
sprintf.s: sprintf.c
vprintf.s: /usr/include/stdarg.h
vprintf.s: /usr/include/stdio.h
vprintf.s: vprintf.c
vsprintf.s: /usr/include/stdarg.h
vsprintf.s: /usr/include/stdio.h
vsprintf.s: vsprintf.c
vfprintf.s: /usr/include/ctype.h
vfprintf.s: /usr/include/stdarg.h
vfprintf.s: /usr/include/stdio.h
vfprintf.s: /usr/include/sys/types.h
vfprintf.s: vfprintf.c
# IF YOU PUT ANYTHING HERE IT WILL GO AWAY
================================================================
I think I used a separate pass for lorder and tsort, which often
cause problems with libraries; but the list of files is so small,
I'm pretty certain it all went smoothly.
Fred Fish's Portable Maths Library is also no problem, tho' it
needs couple of minor patches. In fact, the only useful function
missing from Peter Housel's library is atan, which often occurs as a
basis for trig functions. To avoid searching a whole additional library,
I just added the following to libfp.a, from one of Peter's references:
================================================================
/* atan routine taken unchanged from JACM 23, 2, April 1976 pp. 242-251 */
#include <float.h>
#include <math.h>
double atan(x)
double x ;
{
int j, k, n ;
double p, q, s, v, w ;
n = 64 ; /* precision - 32 ? */
p = 1 / pow(2, n) ;
s = 1 / pow(2, n/2) ;
v = x / (1 + sqrt(1 + x * x)) ;
q = 1 ;
while ((1 - s) > p) {
q = 2 * q / (1 + s) ;
w = 2 * s * v / (1 + v * v) ;
w = w / (1 + sqrt(1 - w * w)) ;
w = (v + w)/(1 - v * w) ;
v = w/(1 + sqrt(1 + w * w)) ;
s = (2 * sqrt(s))/(1 + s) ;
}
q = q * log((1 + v)/(1 - v)) ;
return(q);
}
double atan2(x, y)
double x; double y;
{
return atan(y / x) ; /* yes, y/x is the standard */
}
================================================================
Hope this helps - Will
-----------------------------------------------------------------------
"If heaven too had passions | Will Rose
even heaven would | UUCP: {nosc ucsd hplabs!hp-sdd}!crash!pnet01!cw
grow old." - Li Ho. | ARPA: crash!pnet01!cwr@nosc.mil
| INET: cwr@pnet01.cts.com
UUCP: {nosc ucsd hplabs!hp-sdd}!crash!pnet01!cwr
ARPA: crash!pnet01!cwr@nosc.mil
INET: cwr@pnet01.cts.com