RFR: 8325541: C2 SuperWord: refactor filter / split [v2]

Christian Hagedorn chagedorn at openjdk.org
Mon Feb 12 08:49:01 UTC 2024


On Fri, 9 Feb 2024 15:23:15 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

>> src/hotspot/share/opto/superword.cpp line 1655:
>> 
>>> 1653:         tty->cr();
>>> 1654:         tty->print_cr("WARNING: Removed pack: %s:", error_message);
>>> 1655:         print_pack(pack);
>> 
>> Could this be part of `remove_pack_at()`?
>
> Can I leave this for a future RFE?
> [JDK-8325252](https://bugs.openjdk.org/browse/JDK-8325252) C2 SuperWord: refactor the packset

Sure

>> src/hotspot/share/opto/superword.hpp line 548:
>> 
>>> 546:   Node* vector_opd(Node_List* p, int opd_idx);
>>> 547:   // Can code be generated for pack p?
>>> 548:   bool implemented(const Node_List* p);
>> 
>> Could be made `const` as well if you also make `is_marked_reduction()` const.
>
> I could. But I don't want to push this too far in this RFE. I could also make the whole function const.
> Maybe I can have a future RFE where I just make everything const that can be const in SuperWord?
> But also once I refactor out all the components (e.g. Reductions), then I can just make things const at that point.

Up to you, I just thought since you changed the argument to `const` that you could also make the entire method `const`. But could also be done later.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/17785#discussion_r1485871480
PR Review Comment: https://git.openjdk.org/jdk/pull/17785#discussion_r1485872545


More information about the hotspot-compiler-dev mailing list