





DOS executables

     Warren Ferguson, ferguson@seas.smu.edu, provides DOS executables in the
ucbDOS directory, along with source codes and scripts for the Borland C++ for
Windows IDE.  These may prove helpful with other compilers for MSDOS or Win-
dows.

     I tried to annotate all changes with the symbol DOS - so a search for all
occurrences of DOS should point to all the changes I made.

     I've also included the fdlibm files copysign.c, rint.c, scalbn.c, and
fdlibm.h.  These "must" be compiled in ANSI mode under Borland C. The only
trouble I recall was converting int's to long's, or unsigned int's to unsigned
long's, and adding L or UL to long hex constants.

     I was sure to turn ON floating point and turn OFF fast floating point.
The float routines should be compiled with the defines DOS and FLOAT.  The
double routines should be compiled with the defines DOS and DOUBLE.  the dou-
ble extended routines should be compiled with the defines DOS and LONGDOUBLE.

     The changes to the ucb* routines are similar. Here the routines "must"
not be compiled not in ANSI; else the _control87 routines are not accessible.
I compiled these ucb* routines using Borland extensions.

     Borland C naturally runs in a Windows environment, so I used its windows
project manager. These projects are described, in binary format, in the *.ide
files. The ide files s*.ide are for float, d*.ide for double, and w*.ide for
double extended precisions.

     The executables are also larger than normal because I've included all
possible debugging information.

     In ucbmulte.c I changed 'int count' to 'long count' and, in the routine
nearhalf(), I changed 'int n' to 'long n'. Without these changes you cannot
run ucbmulte with ntests > 32K.

     I also have a batch file RUNALL.BAT that runs all the tests, dumps the
output from each test into a file called RESULTS.

     All files are ascii except those files *.ide and *.exe which are binary
files.

f*.exe   ... float executables
d*.exe   ... double executables
w*.exe   ... longdouble executables































































































































































































































































































































































































































































































































































