https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117421
Jeffrey A. Law <law at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |WAITING
Ever confirmed|0 |1
Last reconfirmed| |2025-08-21
Flags| |needinfo+
--- Comment #6 from Jeffrey A. Law <law at gcc dot gnu.org> ---
This issue was reviewed during the RISC-V patchwork meeting this week. The
general consensus was similar to c#5 from Alexander. This has become a bit
unfocused and it's not really clear where Wojciech wants this to go.
I believe Wojciech has been active on other BZs within the last month or so,
but hasn't responded to Alexander's request for clarification.
I'm going to leave this open, but in a waiting state on Wojciech.
Alexander's comments about performance characteristics are spot on as well. We
need to be careful here. Instruction counts can be useful, but they can also
be incredibly misleading if you're not careful. This is likely one of those
scenarios due to the likely impacts of branch prediction, OOO behavior, etc.
