On Mon, Apr 12, 2021 at 11:53 PM H.J. Lu <hjl.to...@gmail.com> wrote:
>
> On Mon, Apr 12, 2021 at 2:21 AM Richard Biener
> <richard.guent...@gmail.com> wrote:
> >
> > On Sat, Apr 10, 2021 at 5:11 PM H.J. Lu via Gcc-patches
> > <gcc-patches@gcc.gnu.org> wrote:
> > >
> > > Add inline_ignore_target function attribute to inform the compiler that
> > > target specific option mismatch on functions with the always_inline
> > > attribute may be ignored. On x86 targets, this attribute can be used on
> > > integer functions to ignore target non-integer option mismatch.
> >
> > I'm not sure I like such attribute but please adjust
> > default_target_can_inline_p
> > accordingly (only few targets override this hook).
> >
> > Richard.
> >
>
> Like this?
Hmm, so
bool
default_target_can_inline_p (tree caller, tree callee)
{
- tree callee_opts = DECL_FUNCTION_SPECIFIC_TARGET (callee);
- tree caller_opts = DECL_FUNCTION_SPECIFIC_TARGET (caller);
- if (! callee_opts)
+ tree callee_opts;
+ if (lookup_attribute ("inline_ignore_target",
+ DECL_ATTRIBUTES (callee)))
callee_opts = target_option_default_node;
+ else
+ {
+ callee_opts = DECL_FUNCTION_SPECIFIC_TARGET (callee);
+ if (! callee_opts)
+ callee_opts = target_option_default_node;
+ }
this means the attribute will "not work" for
statici inline void __attribute__((always_inline,inline_ignore_target))
foo() {}
void __attribute__((target("-mfoo")))
bar() { foo(); }
because while it ignores the target attribute on 'foo',
target_option_default_node
will still conflict with the target attribute of 'bar'.
Is that the intended behavior? Ignore 'foo's target attribute for the
inlining decision, but not 'bar's?
What about __attribute__((optimization)) mismatches?
It might be more sensible to have a separate always-inline
attribute, say __attribute__((force_inline)) that does what I think
always-inline should do (inline, no matter what)?
Now, I still believe always-inline should behave and this new attribute
is totally unnecessary.
Richard.
> Thanks.
>
> --
> H.J.
