Scor­pio News


April–June 1987 – Volume 1. Issue 2.

Page 18 of 51

much ‘better’ results? In the case of ComPas, this would be mostly due to its use of 6 bytes for a real number, instead of 4 almost everywhere else. This explains why David Parkinson didn’t bother fixing ComPas to use the HSA-88B, as there would be quite a lot of work involved a that kind of patch! The time taken by ComPas was 330 seconds, seven times as long as HP4 and eleven times as long as HP5. Very much a case of horses for courses, I suppose...

The best esult I have seen for the benchmark, in case you are still interested after all that, was 2.50000000000117e+05 given by a DEC VAX under VMS. Alas, we don’t know how long it took, as the job put its results in a spooled file, and printed them later. My contact is hoping to get the program to send the “Started” and “Finished” to his terminal next time he runs it! Soon I will try it on the Open University DEC 20 mainframe...

Gratuitous Attacks on Other Contributors!

Not really, just a few comments on the content of issue 1. Quite a lot of it wasn’ really directly concerned with the hardware listed on the front cover of the magazine. In particular, there were a lot of pages about Fortran, a language that should have died out years ago. And Dave Hunt’s dictionary was rather incomplete. Items of this nature should be up to date, and include definitions relating to important languages like Pascal. You know the sort of thing I mean, “Recursion, see recursion”

I was interested by the article about Gemini’s Multinet, as I think there is far more sense in having at least one processor for each user than having lots of terminals tied onto one micro. Also, it sounds as if Multinet works properly, unlike the awful Econet. That being so, I wonder why it isn’t mentioned in Network magazine? Back to Dave Hunt, whose reply to Bert Martin’s letter was four times as long as the letter itself. Methinks he doth protest too much!

Installing ED80 for use with the SVC.

ED80 is the program editor I use, which is supplied with just about every language Hisoft produce. It comes with an installation progam, to enable the user to set it up for a particular system’s display hardware, and very good it is, too. What it doesn’t do is word processing, with the word-wrapping and justification that implies. Programs tend not to need that sort of thing.

When I finished installing mine, following the instructions carefully (and the manual does explain the process clearly), I found I had a problem. The SVC was not showing any cursor. After a few minutes trying to work by pressing a key and watching where a character appeared, I decided I had to do something. Going back to the manual, I found out how to get ED80 to send an initialisation string to the SVC, and set it up to enable the cursor, in the form of a flashing block. I happen to prefer that sort to the more usual underline character. On running ED80, there was my flashing cursor blob. However, as soon as I touched a key it vanished. Further investigation suggested that CP/M wee resetting it to the original invisible type for inscrutable reasons of its own. So I dug out the CP/H BIOS listings, and found the routine that was doing it I made a CP/M with that routine removed, and ran ED80 again. The cursor still vanished as soon as in was moved from the top left of the screen. I went to the pub...

In the end, I had a look at the code of ED80, using Hisoft’s MON.COM utility, and found that being jolly good chaps, and tidy programmers, they had routed all calls to the CP/M screen writing routine to a single routine of their own. Even better, there were useful areas of zeros in the program, which looked as if a patch inserted in them would do no harm. So I cobbled together a patch, and put it in. It may be ugly, it may be longer than it needs to be, it may even break some obscure law of Dijkstra or Codd, for all I know, but it also happens to work. Here it is, in case anybody needs it:

Page 18 of 51