>
>
> On 3/11/24 4:38 PM, Eric Botcazou wrote:
> > Hi,
> >
> > this is a regression present on all active branches: the attached Ada
> > testcase
> > triggers an assertion failure when compiled with -O2 -gnatp -flto:
> >
> > /* Initialize the static chain. */
> > p = DECL_STRUCT_FUNCTION (fn)->static_chain_decl;
> > gcc_assert (fn != current_function_decl);
> > if (p)
> > {
> > /* No static chain? Seems like a bug in tree-nested.cc. */
> > gcc_assert (static_chain); <--- here
> >
> > setup_one_parameter (id, p, static_chain, fn, bb, &vars);
> > }
> >
> > The problem is that the ICF pass identifies two functions, one of which has
> > a
> > static chain but the other does not. The proposed fix is just to prevent
> > this
> > identification from occurring.
> >
> > Tested on x86-64/Linux, OK for all active branches?
> >
> >
> > 2024-03-11 Eric Botcazou <ebotca...@adacore.com>
> >
> > PR ipa/113996
> > * ipa-icf.h (sem_function): Add static_chain_present member.
> > * ipa-icf.cc (sem_function::get_hash): Hash it.
> > (sem_function::equals_wpa): Compare it.
> > (sem_function::equals_private): Likewise.
> > (sem_function::init): Initialize it.
> >
> >
> > 2024-03-11 Eric Botcazou <ebotca...@adacore.com>
> >
> > * gnat.dg/lto27.adb: New test.
> OK.
Patch is still OK, but ipa-ICF will only identify the functions if
static chain is unused. Perhaps just picking the winning candidate to be
version without static chain and making ipa-inline to not ICE when calls
with static chain lands to function with no static chain would help us
to optimize better.
Also IPA-SRA can optimize out unused static chains and ipa-prop can
propagate across them.
Honza
> jeff
>
