next up previous contents index
Next: Rebuilding components of Up: Installing CERNLIB software Previous: Porting the CERN

Porting CERNLIB to FACOM VPX series

The FACOM VPX series run a Unix System V system. However, the floating point representation is that of IBM mainframes.

We start with the Sun Solaris versions of the libraries, e.g. KERNSUN with the flag SOLARIS.

In ZEBRA, we must ensure that data is correctly converted on input and output.

For IBM mainframes, the required definitions are in the deck IBM of patch FQ in the Zebra pam file. The conversion of data on input and output is performed in the routines FZICV and FZOCV respectively. The conversion is performed by sequences as shown below (plus the corresponding sequences for input). For the FACOM, the following is probably sufficient:

Selecting correction input/output conversion

+USE, FQIE3FSC.    use default CALL IE3FOS for output single prec.
+USE, FQIE3FDC.    use default CALL IE3FOD for output double prec.
+USE, FQIE3TSC.    use default CALL IE3FOS for input single prec.
+USE, FQIE3TDC.    use default CALL IE3FOD for input double prec.

which will call the KERNLIB conversion routines (which must of course be provide) for floating point data and copy as is for all other data. (The VPX series uses the ASCII character set and is big endian).

In this respect, the PATCH IBX, which is for AIX on IBM mainframes, and KERNIBX CAR, which contains Fortran versions of the floating point conversion routines, may work directly on the FACOM.

FZOCVFB
Output conversion of bit strings
FZOCVFI
Output conversion of integer data
FZOCVFF
Output conversion of single precision data
FZOCVFD
Output conversion of double precision data
FZOCVFH
Output conversion of hollerith data

Janne Saarela
Wed Mar 29 15:50:22 METDST 1995