On Wed, 04 Sep 2024 15:20:45 PDT (-0700), jeffreya...@gmail.com wrote:
On 9/4/24 4:07 PM, Palmer Dabbelt wrote:
These tests were checking that the output of the setCC instruction was bit
flipped, but it looks like they're really designed to test that
redundant sign extension elimination fires on conditionals from function
inputs. Jeff just posed a patch to clean this code up with trips up on
the arbitrary xori/snez instruction selection decision changing, so
let's just robustify the tests.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/sge.c: Adjust regex to match the input.
* gcc.target/riscv/sgeu.c: Likewise.
* gcc.target/riscv/sle.c: Likewise.
* gcc.target/riscv/sleu.c: Likewise.
Works for me. Didn't see worth much effort here. Based on the history
of those tests their main purpose is to ensure we don't have an
extension after the sCC.
Ah, OK, I guess I misunderstood then -- we'd had a bunch of issues
overly sign extending inputs, so I figured that's what this was trying
to do. I think the output side is still covered by the
scan-assembler-not for W ops, so we should be safe there.
---
I haven't tested this, but it should be indepndent from Jeff's.
Independent, but it would eliminate the need for my twiddle to these
tests. I don't much care either way other than making sure they stay
in a PASS state ;-)
Ya, sorry, I guess "independent" is really the wrong word there. I was
trying to say we could merge this and it'd pass before/after your
change.
jeff
