[daip] AIPS text version woes

[Eric Greisen]

On 05/18/2015 01:03 AM, Michael Bietenholz wrote:
> Hi Eric:
>
> So - just to confirm, you've found that
>
> gcc/gfortran 4.8.2 through 4.9.2 all see to fail.
>
> gcc/gfortran 4.4.7 seems to compile AIPS and produce reasonably fast
>   executables, but renders gdb unusable
>
> g77 (3.4.6?) also compiles ok, slower executables but gdb works?
>
> I did get it to compile seemingly ok with gcc 4.4.5 but haven't
> had a chance to test yet.
>
>
>                   thanks,   michael
>
>
>
>> On 05/08/2015 03:07 PM, Michael Bietenholz wrote:
>>
>> I gave up on gfortran even the 4.4.7 and went back to g77.
>> The gfortran ran Y2K in 121 sec while g77 takes 174 but gdb claimed
>> that the task just disappeared a lot when that was not part of the
>> bugs.  I use my gnu-complied version for debugging so the gdb issue is
>> paramount.  I can access the Intel version with a logical inside aips
>> so I can get performance if I need it.
>>
>> Sigh...
>>
>> Eric
>>

I have found things to try - after a MNJ tomorrow



                 AIPS considerations to use
                gfortran versions 4.8 and up


1. In the $HOME/.AIPSRC file or interactively in install.pl
    select the version (full path may be needed) of gfortran
    plus a matching gcc.

    Alternatively, you can edit $SYSLOCAL files FDEFAULT.SH and
    LDOPTS.SH for gfortran (COMPILER) and CCOPTS.SH for gcc.

2. Edit $TST/INC/PAPC.INC to change

       PARAMETER (PKPWRD = 2)
C     PARAMETER (PKPWRD = 200000000)

    to

C     PARAMETER (PKPWRD = 2)
       PARAMETER (PKPWRD = 200000000)

    Note - 200 million is about 1.6 Gbytes of ram.  You can set
    a smaller number if you want.

3. Edit $SYSUNIX/OPTIMIZE.LIS to change

#  APCLN              LNX64            1           0

    to

    APCLN              LNX64            1           0

4. Then let INSTEP2 and INSTEP4 run.


Eric