🚀 go-pugleaf

In article <123@garth.UUCP> smryan@garth.UUCP (Steven Ryan) writes:

| Beat me! Whip me! Make me use an optimiser!
| 
| Isn't this kind of silly? Are there really compilers that force people to
| use the optimiser or is it optional? If it's optional and you don't want,
| don't use. If you can buy a compiler without an optimiser (which is easy),
| do so. If nobody wants an optimiser, nobody buys them. If somebody wants
| one, why not let them buy it?

I once worked on a language (DG/L at Data General), in which it had NO
option to turn off the optimizer.  It was 9 passes plus a control
program that ran in a 64K address space (and no segments to extend the
address space), and had some optimizations that the later DG compilers
never did, such as doing interprocedural analysis with code motion
from internal subroutines to their caller's.

I've also found a few cases in finishing GCC ports for the 88k and
MIPS, where optimizing the code produced correct code, and not
optimizing produced compiler error messages.  This is because
constraints in the machine description were not quite correct, but the
optimizer smoothed things over by putting variables into registers.

I personally never compile any code, except test suites, without
enabling both debug and 'full' optimization these days, unless I'm
using a vendor's PCC derived compiler.
--
Michael Meissner	email: meissner@osf.org		phone: 617-621-8861
Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142

Considering the flames and intolerance, shouldn't USENET be spelled ABUSENET?