head	1.2;
access;
symbols;
locks
	wmc:1.2; strict;
comment	@# @;


1.2
date	2003.03.21.11.42.51;	author wmc;	state Exp;
branches;
next	1.1;

1.1
date	2003.03.21.10.14.45;	author wmc;	state Exp;
branches;
next	;


desc
@Hmm. This version was the one I used for some time, and which
gave every appearence of working. Nonetheless it is regrettable
that it ends up declaring real_FAILURE as an INTEGER... though
this doesn't seem to matter, it needs fixing.
@


1.2
log
@Fixed version (real_FAILURE really is real) and add notes.
,
@
text
@*IDENT QTPOSFIX
*/
*/ This modset was needed for quite a while to make qtpos1a bit compare
*/ properly (under my setup: fujitsu. Whether it is needed for other 
*/ setups I don't know). The problem was that GC_ISUM didn't work properly.
*/
*/ It now (2003/03) looks like the problem may well be with something
*/ ?gcom? assuming that MPI_INTEGER was always 32 bit. Or somesuch (Paul
*/ Burton found this). This applies to the gcom that came with vn4.5, and to
*/ vn2.8.
*/
*/ There is a new version of gcom (vn2.9b5) that appears to fix the problem.
*/
*/ It is not clear whether this problem applies at 32 bit or not.
*/
*DECLARE QTPOS1A
*D APB6F401.84
C WMC. Make sure this *is* a real.
      REAL
     &  real_FAILURE(Q_LEVELS)    ! version of LOGICAL FAILURE array,    
*D APB6F401.249,APB6F401.251
C WMC
            real_FAILURE(K)=1.0
          ELSE
            real_FAILURE(K)=0.0
*D APB6F401.255
           real_FAILURE(K)=0.0
*D APB6F401.272
      CALL GC_RSUM(Q_LEVELS,N_PROCS,info,real_FAILURE)                        
*D APB6F401.276
C WMC
C        IF (int_FAILURE(K) .EQ. 0) THEN                                 
        IF (real_FAILURE(K) .lt. 0.1) THEN                                 
@


1.1
log
@Initial revision
@
text
@d2 14
d18 3
a20 2
C WMC
     &, real_FAILURE(Q_LEVELS)    ! version of LOGICAL FAILURE array,    
@