product form of inverse / elementary inverse matrix based simplex skeleton is working and it’s time to get it running on ibsys and tops10, so ibm7094 and dec10 respectively. to set the stage and make it real, here’s some visuals. matters are straightforward for tops10 with telnet, kermit, and a familiar runtime env. ibsys, on the other hand, is a completely different story. it means simh emulated tape, possibly cards, and batch jobs in general instead of interactive terminals.

thanks to richard cornwell’s ibsys pages linked above, it took only a few hours of puttering to boot ibsys tape stage11.tp and run a job via tape. the focus quickly turned to running a job to compile and execute the simplex.for source code linked above, learning on the fly from ibm fortran iv ibftc and rich’s example jobs.

getting even a minimal hello world fortran compile and execute turned out to be relatively painful. there seems to be a variety of issues at play, from hardware config to ibsys job deck details, but can at least now ‘print’ integers and holleriths and see the proper result on the ‘job line printer’. ‘write (6,xyz)’ is not yet working. in any case, compiling and running fortran code on 7094 is go.

along the way, realized something completely unexpected. the way this happened led to a kind of accidental insight. the fortran code, with a few additional lines representing ibsys control cards, is put into a tape file. the simh 7094 boots from an ibsys tape file, then does the job in the ‘job tape file’, compiling and running the fortran code. then have it set to simply shut down, returning to the linux command line. so the whole sequence starts and ends at the linux command line.

this is an ‘ephemeral 7094’. and can imagine a similar ‘ephemeral dec10’. for exploring early numerical algorithms in fortran, this could be liberating. the focus isn’t on babysitting 7094s and dec10s. it’s on the algorithms and fortran code. maybe even having time to stretch further back into assembly.