On Thu, Oct 31, 2019 at 09:16:00AM +0000, Richard Sandiford wrote:
> > Yes, it is indeed not clear, subject to ongoing discussions.
> > My reading of the spec was that all the *-name-list are comma
> > separated lists of identifiers, some others in the committee
> > want now (yesterday's discussions) string literals instead
> > when I and others pointed out that isa(core-avx512) can't be valid,
> > but strangely only for isa/arch/extension but not e.g. for
> > kind or vendor which would still take identifier lists etc.
>
> Might be completely wrong, but wouldn't the identifiers be subject to
> macro expansion? That would make it harder to use the pragmas safely
> in system headers.
Well, all the identifiers in the OpenMP/OpenACC pragmas after the omp
are subject to macro expansion, including clause names etc., so even
#define device foobar
#pragma omp declare variant (foo) match (device={kind(host)})
is a problem, or macro redefinition of declare, variant, match,
kind or host. That said, for the arch and isa it is perhaps a bigger
problem, as can be seen in the c-c++-common/gomp/declare-variant-9.c
testcase where I had to #undef i386 because in non-strict modes
that is defined...
Jakub
