https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116815
--- Comment #6 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Uroš Bizjak from comment #5)
> Created attachment 62666 [details]
> x86 patch in testing
The pr116815-6.c testcase from Dhruv's patch ("[PATCH v4] [aarch64] Make better
use of overflowing operations in max/min(a, add/sub(a, b)) [PR116815]") at [1]
compiles on x86_64 (gcc -O2) to:
umaxadd:
addl %edi, %esi # 18 [c=4 l=2] addsi3_cc_overflow_1/0
movl %edi, %eax # 21 [c=4 l=2] *movsi_internal/0
cmovnc %esi, %eax # 19 [c=4 l=3] *movsicc_noc/0
ret # 25 [c=0 l=1] simple_return_internal
umaxsub:
movl %edi, %eax # 21 [c=4 l=2] *movsi_internal/0
subl %esi, %eax # 18 [c=4 l=2] *subsi_3/0
cmovnb %edi, %eax # 19 [c=4 l=3] *movsicc_noc/1
ret # 25 [c=0 l=1] simple_return_internal
uminadd:
addl %edi, %esi # 18 [c=4 l=2] addsi3_cc_overflow_1/0
movl %edi, %eax # 21 [c=4 l=2] *movsi_internal/0
cmovc %esi, %eax # 19 [c=4 l=3] *movsicc_noc/0
ret # 25 [c=0 l=1] simple_return_internal
uminsub:
movl %edi, %eax # 21 [c=4 l=2] *movsi_internal/0
subl %esi, %eax # 18 [c=4 l=2] *subsi_3/0
cmovb %edi, %eax # 19 [c=4 l=3] *movsicc_noc/1
ret # 25 [c=0 l=1] simple_return_internal
[1] https://gcc.gnu.org/pipermail/gcc-patches/2025-August/692789.html
