https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121604
Bug ID: 121604
Summary: [15/16 Regression] Wrong folding of svbrkb intrinsic
since r15-5957-g5289540ed58e42
Product: gcc
Version: 16.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: acoplan at gcc dot gnu.org
Target Milestone: ---
The following testcase is miscompiled since
r15-5957-g5289540ed58e42ae66255e31f22afe4ca0a6e15e:
$ cat brkb.c
#include <arm_sve.h>
svbool_t foo() {
return svbrkb_b_m(svpfalse(), svptrue_b8(), svptrue_b8());
}
$ gcc/xgcc -B gcc -c brkb.c -O2 -march=armv8.2-a+sve -S -o -
.arch armv8.2-a+crc+sve
.file "brkb.c"
.text
.align 2
.p2align 5,,15
.global foo
.variant_pcs foo
.type foo, %function
foo:
.LFB2:
.cfi_startproc
ptrue p0.b, all
ret
.cfi_endproc
.LFE2:
.size foo, .-foo
.ident "GCC: (GNU) 16.0.0 20250819 (experimental) [master
7eb8b65780d]"
.section .note.GNU-stack,"",@progbits
it should instead fold to pfalse, see the definition of the intrinsic here:
https://developer.arm.com/architectures/instruction-sets/intrinsics/svbrkb[_b]_m
For a runtime testcase, take the following:
```
#include <arm_sve.h>
__attribute__((noipa))
svbool_t foo() {
return svbrkb_b_m(svpfalse(), svptrue_b8(), svptrue_b8());
}
int main(void)
{
/* brkb (ptrue) -> pfalse. */
if (svptest_any (svptrue_b8 (), foo ()))
__builtin_abort ();
}
```
It looks like BRKA is affected too, at first glance.
