https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120242
Jeffrey A. Law <law at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC|rsandifo at gcc dot gnu.org |
--- Comment #2 from Jeffrey A. Law <law at gcc dot gnu.org> ---
So after 2 years, it's finally come back to bite us... The real regression is
from this:
Vineet/Robin, y'all will certainly remember this gem:
commit 8eb9cdd142182aaa3ee39750924bc0a0491236c3
Author: Vineet Gupta <vine...@rivosinc.com>
Date: Mon Oct 16 21:59:09 2023 -0600
expr: don't clear SUBREG_PROMOTED_VAR_P flag for a promoted subreg
[target/111466]
RISC-V suffers from extraneous sign extensions, despite/given the ABI
guarantee that 32-bit quantities are sign-extended into 64-bit registers,
meaning incoming SI function args need not be explicitly sign extended
(so do SI return values as most ALU insns implicitly sign-extend too.)
[ ... ]
We've got a bogus promoted state which is interacting badly with combine. We
weren't ever to find a test for the old code back when we dug into this
problem, but we've got one now....
