Re: Thompson's Plan 9 C compiler

kurt@tc.fluke.COM (Kurt Guntheroth) writes:


      William Wulf, architect of the BLISS-11 (high optimizing) compiler argues
      with you ...
       ... I wish I had the exact quote, but somebody stole my copy
      of Wulf's book.


The exact quote is:


We would like to conclude by presenting our view on the
importance of efficiency. It is fashionable in some circles,
especially some academic ones, to deprecate the importance of
efficiency. The argument goes: "It is programming time, not
execution time, that matters. What difference does it make if
the program runs one minute or two?" This argument is often used,
for example, to encourage the use of some of the newer, often
less efficient, programming languages.


It is difficult to argue with this. In fact, we don't for our own
programs. Our time is _much_ more important than the machine's.
However, the other fellow's program (yours) is an entirely
different matter! Every second that his program executes, ours
can't. If his program is inefficient, we are the ones who suffer,
not him. The problem of efficiency is _not_ one of how long it
takes to run _a_ program, but rather one of obtaining the maximum
benefit from a finite resource.


The reason that compiler optimization is important is that
programmer efficiency and execution efficiency need not be a
choice we must make. Optimization is a technological device to
let us have our cake and eat it, too -- to have _both_ convenient
and well-structured programming and efficient programs.


oz
--

Return to the comp.compilers page.
Search the comp.compilers archives again.