On 7/5/2026 11:27 PM, Jim Lin wrote:
With -funroll-loops the loop body is duplicated, emitting more than one
minu and breaking the scan-assembler-times {\mminu} 1 expectation. Skip
the test when -funroll-loops is present, matching zbb-min-max-04.c.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/zbb-min-max-05.c: Skip under -funroll-loops.
I was a bit curious that I hadn't seen this failure, so I did a bit of
digging. My tester focuses on regressions, this is a relatively new
test that's probably been failing since integration and thus wasn't ever
flagged by my tester.
It's not being flagged in the pre/post commit CI system run by RISE. I
suspect, but didn't confirm that the --with-cpu option that system uses
twiddles the cost model in a way that we don't see the unrolling
effect. Unfortunately that system doesn't just provide the .sum files
in a way that can be trivially downloaded and examined :(
Anyway, thanks for taking care of this. I've pushed it to the trunk.
However, can you verify this and other submissions you've made are
covered by the DCO?
https://gcc.gnu.org/dco.html
If you're going to regularly contribute to GCC it may make sense to get
that noted in the MAINTAINERS file.
Thanks again,
Jeff
