| STANFORD UNIVERSITY MEDICAL CENTER
Rh STANFORD, CALIFORNIA 94305 ¢ (415) 321-1200

 

January 23, 1974

STANFORD UNIVERSITY SCHOOL OF MEDICINE
Department of Genetics

Dr. Paul B. Schneck

National Aeronautics and Space Administration
Institute for Space Studies

2880 Broadway

New York, New York 10025

Dear Dr. Schneck,

Thank you for your letter of January 17th and for sending me
the reprint of your paper on the Fortran Optimising Compiler. I will
be discussing this with my colleagues with respect to determining the
approriate time to attempt to implement your system, and I very much
appreciate your offer of cooperation. I may find that there are some
similar activities of which I was not previously aware that would
be easier to maintain in the local environment. But if I find them
I will certainly pass that back onto you.

It strikes me that innovations of this kind would be particularly
useful for exchange on the ARPA-net communication network. Do you know
1f your institute is now or is going to be affiliated with it? I am
enclosing a brief blurb about some role that we may be playing, which
may be a counter example to Fortran, and will send you some additional
remarks in a few days when we have gotten them together.

I found the general ideas in your paper most intriguing. What you
only hint at, but I am sure have given much thought to, is the impact
that having this additional stage of language translation would have on
programming practices. If one could be reasonably confident that all of
these dreary details were going to be taken care of later on, I am sure
that we would write many programs much more briskly and with more
attention to the basic flow of the logic rather than fussing about
optimum flow. So the fair comparison to evaluate your proposals is not as
between present programs and their optimized versions but the kind of
programs that would be most efficiently written at the highest source
level if these facilities were generally available and perfected.

I understand and sympathize with the objectives of using standard
programing texts as input but I am sure you will also have given some
thought to a higher level translator that was subject to some explicit

“control. Otherwise one will be finding the use of side-effect tricks in
order to confuse the optimizing compiler from performing treacherous re-
coding of sections that are intended to be left as is and where we do not
want to bother to edit the output text. One could think of constructions
like listing the names of functions that are intended to produce constants
whose values are to be obtained once but not evaluated again. There are
bound to be circumstances where this advantage is not foreseen by the
compiler.

LT. J. P. KENNEDY, JR. LABORATORIES FOR MOLECULAR MEDICINE, DEDICATED TO RESEARCH IN MENTAL RETARDATION

MOLECULAR BIOLOGY HEREDITY NEUROBIOLOGY DEVELOPMENTAL MEDICINE
wy gaP oar

Dr. Paul B. Schneck -2- 1/23/74

More generally, a point of perhaps dubious value: in the days when
I was more intrigued by bit-pushing, I often hoped for the equivalent of
a macro-expression expander in higher level languages. There are some
programs where the housekeeping overhead connected with procedure calls
takes up a significant proportion of the execution time of an inner
loop but where the programmer just does not want to bother to rewrite the
procedure body with new arguments. Programs are certainly best documented
and tested with conventional procedure calls but again one could see the
advantage of being able to list those procedure blocks which are intended
to be explicitly expanded in the optimized version. This concept made less
sense in the days when assembly level programing was more reasonable than
it is on many of today's machines, so there may be some new virtue to it.

I do foresee that one of the very great virtues of your approach is
in the field of self-documentation since it will now be possible in many
instances to define the values of variables at a point that corresponds
to their logical connotation rather than to their optimal position from
the standpoint of program flow)

So, I will write you again when I have some further advice about
the best time to take advantage of your offer to provide tape copies of
your program. This deferral is based on the idea that you may be continuing
to perfect it. If the converse is true and there is some hazard of its
or your going away, please let me know.

Sincerely yours,

Joshua Lederberg
., . Professor of Genetics

JL/rr