bradley@utgpu.UUCP (11/13/87)
A few weeks ago I posted a note asking for suggestions on how to
get better performance from a program we are converting from
Turbo Pascal to Logitech Modula-II. The following is a summary
of the responses. I tried to send personal mail to all people
who took the time to make comments, but from the response I got
back, few or none of them received my notes. Hence, I'd like to
thank all the respondents here, and apologize for not thanking
them personally.
The commonest suggestion was to order (and wait for) Logitech
3.0. Several people commented on the optimizer it will contain,
but no one who had responded had had enough experience to know
how much of an improvement it would be.
A few people focused on the quality of the conversion. We have
been using Logitech's converter, and then going over the code by
hand to make it a little more ``Modula-II'' like. We had to do
this -- Logitech's converter doesn't properly convert string
operations, and our program has lots of those.
One or two people suggested profilers. This is probably a good
idea, and we are looking into the suggestion.
Interestingly, only one person suggested a different compiler --
M2SDS from Interface Technologies. We have received an
announcement about Stony Brook's compiler. They claim it can
produce code that out performs Microsoft C. Does anyone have any
experience with it?
For the record, one person suggested leaving the program in
Turbo. Does this say something about the current quality of code
generation for the PC from Modula-II compilers??
For our project, we will wait for our copy of Logitech 3.0, and
try out the optimizer. If performance is still bad we may be in
the market for a new compiler.
Thanks again to all.
... John Bradley:
University of Toronto Computing Services
uucp: ...!utgpu!bradley
bitnet: bradley@utoronto