https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119509
--- Comment #10 from Nick Hudson <skrll at netbsd dot org> ---
(In reply to Jeffrey A. Law from comment #9)
> So I was looking at this for riscv. While I think the code can clearly be
> improved, the path for doing so is far from obvious.
Thanks for looking.
Is my candidate fix not usable? It changes the output of the test program for
riscv as follows
-- riscv64.foo.s 2025-12-08 08:25:17.259616639 +0000
+++ riscv64.foo.s.new 2025-12-08 08:47:53.751282204 +0000
@@ -10,11 +10,9 @@
foo:
.LFB0:
.cfi_startproc
- li a5,16777216
- addi a5,a5,-256
- and a0,a0,a5
- li a5,255
- sgtu a0,a0,a5
+ slli a0,a0,40
+ srli a0,a0,48
+ snez a0,a0
ret
.cfi_endproc
.LFE0:
