RFR: 8349764: RISC-V: C1: Improve Class.isInstance intrinsic [v2]

Hamlin Li mli at openjdk.org
Thu Feb 13 09:09:15 UTC 2025


On Thu, 13 Feb 2025 04:20:40 GMT, Gui Cao <gcao at openjdk.org> wrote:

>> src/hotspot/cpu/riscv/c1_Runtime1_riscv.cpp line 953:
>> 
>>> 951:         __ ld(x17, Address(x17));
>>> 952:         __ beq(klass, x17, success);
>>> 953:         __ j(fail);
>> 
>> Hmm, I think this jump could be saved if we put a direct return here instead. Like:
>> 
>>         __ beq(klass, x17, success);
>>         __ mv(result, 0);
>>         __ ret();
>> 
>> What do you think? @Hamlin-Li @zifeihan
>
> Yes, I think we can put a direct return here.

Yes, seems better.
I was thinking to move bind(fail) here, but seems current lookup_secondary_supers_table_var only accept a success label.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/23551#discussion_r1954102465


More information about the hotspot-compiler-dev mailing list