Re: PATCH: Fix 20000724-1.c

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Richard" == Richard Henderson <rth@redhat.com> writes:

    Richard> The only other thing would be the scheduler, but volatile
    Richard> asms are scheduling barriers.

    Richard> Worse come to worse, it might make sense to not do stack
    Richard> comhining across volatile asms either.

I guess my basic point is that the RTL should express what's going on
here.  If we can't deduce what's going on from what the user wrote,
then the user needed to right more.  If the user wrote enough, then we
have to be conservative.

If we just don't combine across asms (volatile or otherwise) then
we're saying that the user didn't write enough (i.e., we can't figure
out from the asm what's going on).  In other words, we're adding
detail to the semantics of an extended asm, which is OK, but we need
to document that.

But, by your argument that my original analysis was bogus, it seems
like the user *did* write enough.  So, we should be able to key off of
something about the instruction, not just its asm-ness.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com