[PATCH][Tree-optimization/PR89730]grant always_inline when -flive-patching=inline-only-static

[Richard Biener]

On Tue, Mar 26, 2019 at 5:25 PM Qing Zhao <qing.zhao@oracle.com> wrote:
>
> Hi, Richard,
>
> thanks for the suggestion.
>
> I tried it yesterday, but it did not work.
>
> the reason is:
>
> inside “can_inline_edge_by_limits_p”,  the “allowance for always_inline” is guarded by the following condition:
>
> else if “caller_tree != callee_tree”
>
> this condition is ONLY true when when callee has different optimization level than the caller.
>
> So, I think that the correct spot for the checking of live-patching=inline-only-static should still be inside “can_inline_edge_p”.
>
> so my previous patch for this bug should be fine.
>
> what’s your opinion?

You could still move it after the following?

  /* Check if caller growth allows the inlining.  */
  if (!DECL_DISREGARD_INLINE_LIMITS (callee->decl)
      && !disregard_limits
      && !lookup_attribute ("flatten",
                 DECL_ATTRIBUTES (caller->decl))
      && !caller_growth_limits (e))
    inlinable = false;

that said, looking at DECL_DISREGARD_INLINE_LIMITS in
can_inline_edge_p looks wrong.

Richard.

> thanks.
>
> Qing
>
> > On Mar 25, 2019, at 7:23 AM, Richard Biener <richard.guenther@gmail.com> wrote:
> >
> > On Wed, Mar 20, 2019 at 4:16 PM Qing Zhao <qing.zhao@oracle.com> wrote:
> >>
> >> Hi,
> >>
> >> there is a bug in the current support for -flive-patching=inline-only-static:
> >>
> >> it rejects inlining of external always_inline routine, therefore triggers a compilation time error.
> >>
> >> we should always inline “always_inline” routines.
> >>
> >> please review the following simple patch, it has been bootstrapped and regression tested on aarch64.
> >>
> >> Okay for committing?
> >
> > To me this seems like can_inline_edge_p was the wrong spot to disable inlining
> > for -flive-patching=inline-only-static and instead it should have been in
> > can_inline_edge_by_limits_p which already has proper allowance for
> > always-inline.
> >
> > So can you move the check there, like after
> >
> >      /* gcc.dg/pr43564.c.  Apply user-forced inline even at -O0.  */
> >      else if (always_inline)
> >        ;
> >
> > ?
> >
> >> thanks.
> >>
> >> Qing.
> >>
> >> gcc/ChangeLog:
> >>
> >> 2019-03-20  qing zhao  <qing.zhao@oracle.com>
> >>
> >>        PR tree-optimization/89730
> >>       * ipa-inline.c (can_inline_edge_p): Grant always_inline even when
> >>       -flive-patching=inline-only-static.
> >>
> >> gcc/testsuite/ChangeLog:
> >>
> >> 2019-03-20  qing zhao  <qing.zhao@oracle.com>
> >>
> >>        * gcc.dg/live-patching-4.c: New test.
> >>
>