Bimorphic inlining not applied at a call site that was initially monomorphic

[Filipp Zhinkin]

Hi Roland,

Thanks for taking a look at the problem.

I filed https://bugs.openjdk.org/browse/JDK-8326369 to track it.

Best regards,
Filipp.

On Tue, 20 Feb 2024 at 12:10, Roland Westrelin <rwestrel at redhat.com> wrote:

>
> > The question is if such a behavior is intentional (I didn't find an issue
> > suggesting the opposite)
> > and if it is, what differs it from scenarios where a call site is always
> > bimorphic and where the C2
> > successfully performs inlining (or where the C1 simply can't inline a
> > monomorphic vcall initially)?
>
> Thanks for all the details.
> I don't think it is intentional. It looks like a bug to me. This:
>
> diff --git a/src/hotspot/share/c1/c1_LIR.hpp
> b/src/hotspot/share/c1/c1_LIR.hpp
> index 6f527135fbe..4ea5b9ebb73 100644
> --- a/src/hotspot/share/c1/c1_LIR.hpp
> +++ b/src/hotspot/share/c1/c1_LIR.hpp
> @@ -2031,7 +2031,7 @@ class LIR_OpProfileCall : public LIR_Op {
>    bool should_profile_receiver_type() const {
>      bool callee_is_static = _profiled_callee->is_loaded() &&
> _profiled_callee->is_static();
>      Bytecodes::Code bc =
> _profiled_method->java_code_at_bci(_profiled_bci);
> -    bool call_is_virtual = (bc == Bytecodes::_invokevirtual &&
> !_profiled_callee->can_be_statically_bound()) || bc ==
> Bytecodes::_invokeinterface;
> +    bool call_is_virtual = (bc == Bytecodes::_invokevirtual/* &&
> !_profiled_callee->can_be_statically_bound()*/) || bc ==
> Bytecodes::_invokeinterface;
>      return C1ProfileVirtualCalls && call_is_virtual && !callee_is_static;
>    }
>  };
>
> makes both cases perform the same (that change is likely not the actual
> fix).
>
> Can you file a bug?
>
> Roland.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-compiler-dev/attachments/20240220/0de75514/attachment-0001.htm>