• Aucun résultat trouvé

Ron talked about a somewhat different metacompiler, or System Synthesizer Something which you would approach with a problem. Now I wfth what

Dans le document the Air (Page 175-179)

language would you approach Ron's Synthesizer. You certainly wouldn't approach it in CMS-2. You wouldn't approach it in SPL. Those languages problems. The software people exist so that people with military problems can use the hardware to solve problems. Finally, I'd like you to remember what Nimensky said about why compilers weren't used more. He gave three reasons. He said, first, the languages were not appropriate.

Second, that the compilers were too expensive and, third, that what the compilers produced was too inefficient .•• Now, there are some hardware people around who say that you software people have had ten years to deliver machine independent, transferable programs and you haven't done it and, maybe, now it's their tum. Suppose they deliver to you a FORTRAN machine. We would have to extend the language a little to explicitly declare the data types, to explicitly declare the semantics, but it would be a FORTRAN machine. You wouldn't know the addressing structure, you wouldn't know the internal word length, no compiler would be nec-essary, there would be no machine language; would this machine answer Nimensky's objections? Can we build it?

If

we can build

it,

what would

1?J.

-you system programmers do for a living? .... With that statement of prej-udices, I will tum to the panel, ask them to introduce themselves, and

speak for five minutes, if they wish, on their view of the problem, state their qualifications, and then maybe we'll get an argument going.

Samtmann: I'm Bob Samtmann from the Naval Air Development Center.

I'm an engineer. I've been associated with the computer subsystem devel-opment for the F-lll over the past five years. While my work has been associated primarily with hardware development, I've picked up an interest in software by osmosis. Just recently I've been assigned to the F-14

program computer subsystem development. Also, recently, live been getting active on the AADC program where I've been working with Systems Consultants Corporation on the instruction repertoire development. I will be taking an active lead in the compiler development -- whatever com-patibility of languages, in comcom-patibility of compilers running on host machines. Problems in efficiency .•• Right now, we'll probably have to go back and do some additional thinking ..• probe the problem some more ... Thats' about all I can say right now.

Henderson: I'm Vi Henderson of Logicon. I've been working in the aero-space field for approximately fifteen years; mainly with Air Force and programmers because they're expert programmers. I tend to think that

they've become system programmers because they have forgotten, or have not addressed themselves to becoming expert in the application

area. They haven't learned the guidance and control problem. They haven't learned the dynamics of a system, and that sort of thing. Therefore I

they leave that field and work in an area they can handle technically.

We've talked a great I and yet I it's ironical to me, the problem of coding

-

175--for an avionic system is relatively low cost, if we just talk about the coding particular mission requirement or set of missions. I believe that the synthesizer can only evolve •.. It will be a long, long time before it be-comes an on-line tool. The development cycle has been, for software, a long tortuo,:, .. s road; particularly, the check-out side of it. For those applications, such as the Apollo ..• some very high cost system where

you can't affort to buy the system to begin with ..• you get the tail-wagging-the-dog situation, where the amount of check out, redundancy, resources required to really test the software becomes enormous. We don't pay

enough attention to that part of the problem, in building languages, building the application programs, and disciplining the whole management process.

Deerfield: I'm Alan Deerfield from the Raytheon Company. I'm a

-to throw one at me during the course of the conversations. I won 't get angry. I probably won 't be able to answer your questions either. lid like to go on record with a couple of my views. Most of what Bruce Wald mentioned I agree with, although 1'm sure if we got down to inter-preting, our formal agreement would probably be different. Yes, I be-lieve I can make a computer, for example, without very many problems,

strictly mathematical. As an engineer, as a person with mathematical problems, lid be very happy if I could write my problems down and let two other reasons why I think compilers are necessary in the military, or, at least, problems with them that I think should be attacked. The first not software, tradeoff of your using ten words to program that square root for me. Even if you Ire as fast. Consequently, if I build a com'puter

-your algebra directly. The thing I object to, then, is the fact that you spend an

a~fully

lot of time developing algebraic handling. Let me just briefly elaborate. Suppose, for example, that (going step by step) I go from one accumulator to two accumulators. One reason for going to two accumulators is to eliminate unnecessary load and store orders.

If

I build such a machine, I expect the compiler designer to go in and

Dans le document the Air (Page 175-179)