https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118353
Jeffrey A. Law <law at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |law at gcc dot gnu.org
--- Comment #1 from Jeffrey A. Law <law at gcc dot gnu.org> ---
Dunno if it's relevant. But after Mark W's adjustment to the switch lowering
code bootstrap & regression testing times for most targets in my tester jumped
significantly (though presumably his risc-v times improved significantly).
These are bootstraps & regression tests in QEMU emulated environments, so
they're not a great measure of real performance. But they likely do point out
that we probably have more sensitivity to these issues than is perhaps obvious
and riscv is quite possibly an outlier behavior wise.
mips64el-linux-gnu 10hrs -> 23hrs
hppa-linux-gnu 10hrs -> 20hrs
s390x-linux-gnu 11hrs -> 19hrs
And so-on.
I don't have a breakdown of where the time is spent (ie, bootstrapping vs
testsuite).
Finally while the doubing in times sucks, it just brings is back to where we
were in Nov 2024 before your work went in. So it's not really a regression in
my mind.
