Re: GU: compiling problems

[Pedro A M Vazquez <vazquez@iqm.unicamp.br>]

Jack Connoly said:
> 
> Following this I ran compall, which ran with some warnings but no errors.
> However the lked script hangs up on morokm.o, which has 'unidentied iand and
> ifirst' variables in it.
> I tried to modify the morokmo source file without success.
> I would be grateful for any assistance on this.
> Thanks, Jack Connolly
> 
> 
Hello
	I've compiled Gamess (Nov/95) under FreeBSD 2.1.0 and 2.0.5
using f2c/gcc-2.6.3 without problems. Looking the activated code morokm.f
it seems to me IAND should never be exposed for Unix systems but the AND() 
function instead:

*CRY        I =      SHIFTR(LABELS, 48)
*CRY        J =  AND(SHIFTR(LABELS, 32),65535)
*CRY        K =  AND(SHIFTR(LABELS, 16),65535)
*CRY        L =  AND(       LABELS,     65535)
*IBM        I =       ISHFT(LABELS,-24)
*IBM        J = IAND( ISHFT(LABELS,-16),255)
*IBM        K = IAND( ISHFT(LABELS, -8),255)
*IBM        L = IAND(       LABELS     ,255)
            I =  AND(RSHIFT(LABELS, 24),255)
            J =  AND(RSHIFT(LABELS, 16),255)
            K =  AND(RSHIFT(LABELS,  8),255)
            L =  AND(       LABELS,     255)

this is translated to C as:

            i = labels >> 24 & 255; 
            j = labels >> 16 & 255;
            k = labels >> 8 & 255;
            l = labels & 255;

	IFIRST is an integer variable (grep -i IFIRST morokm.f):

      SAVE IFIRST                                                       
      DATA IFIRST/1/                                                    
      IF (IFIRST.EQ.0) GO TO 100                                        
      IFIRST = 0                              

	and should not impose problems too. I can't understand this 
unresolved symbol.

	It seems to me you have some comp/compall environment variables
problem. Please double check TARGET and other OS specific variables in
these scripts.
	I hope this help you.

	While we are on this subject, GNU f77 project, g77, shows very nice
results for games and it is worth to try. Bellow are the results for the 
benchmarks in the Gamess manual:

                bench10 bench04 bench13 bench07 (user time in seconds)

f2c/gcc    -O2    72.3   92.0    310.1   451.5
g77-0.5.18 -O2    61.0   78.7    281.6   335.5

this is for a Pentium166 (Soyo motherboard) with 64M RAM, 256k cache,
Nec 2.1 SCSI2 HD and an Adaptec 2940 PCI adaptor under FreeBSD2.1.0, the
average speedup is about 11-12% both for Gamess and Gaussian94.


Regards

Pedro
--
Pedro A M Vazquez - Instituto de Quimica - UNICAMP 
FreeBSD - Turning PC's into workstations. See http://www.FreeBSD.ORG/ for info.