Re: Q: access to subprogram

[kenner@lab.ultra.nyu.edu (Richard Kenner)]

In article <m2g26ufuq2.fsf@devo.ridgecrest.ca.us> Ronald Cole <ronald@ridgecrest.ca.us> writes:
>You probably remember well the wailing and gnashing of teeth
>from the CISC-owners when gcc went from version 1 (handled CISC
>well/RISC not-so-well) to version 2 (handles RISC much better/CISC
>not-as-well-as-version-1/holds "inbetween architectures" like the
>Pentium in contempt).

I remember no such thing since it never happened!  GCC version 2 added
new mechanisms to support RISC processors but had *absolutely no
effect whatsoever* on CISC processors.  None of the config files for
CISC processors had to be changed, nor was there any change in the
generated code for these machines.

As to the Pentium issue, this is more complex and there were two
factors.  First, some of the optimizations required for the
best-quality code on that CPU are both very unique to it and are very
large pieces of code to add to the compiler.  Secondly, the group that
did add these, and other, Pentium-specific optimizations to GCC did
such a bad job that it was not possible to fix their code without
rewriting it and nobody wanted to do that without knowing which of the
optimizations were the most important (otherwise, you'd risk spending
a lot of work on complex optimizations that had a small effect).  By
the time this was done, the P6 became available.  It does not benefit
from most of the Pentium-specific optimizations (which I knew a while
before it came out), so the issue rapidly became moot.