https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126163
Bug ID: 126163
Summary: __builtin_mul_overflow wrong code
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: gjl at gcc dot gnu.org
Target Milestone: ---
Created attachment 64962
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64962&action=edit
y.c.274r.expand: RTL dump
__uint24 abu24;
bool mulu24_ovf (__uint24 a, __uint24 b)
{
return __builtin_mul_overflow (b, a, &abu24);
}
$ avr-gcc y.c -S -Os -fdump-rtl-expand-details
The generates asm always returns 0, i.e. there will never be an overflow
detected:
mulu24_ovf:
rcall __mulpsi3
sts abu24,r22
sts abu24+1,r23
sts abu24+2,r24
ldi r24,0
ret
The dumped tree in .expand looks reasonable:
_Bool mulu24_ovf (__int24 unsigned a, __int24 unsigned b)
{
__int24 unsigned _1;
__int24 unsigned _2;
__complex__ __int24 unsigned _5;
_Bool _8;
;; basic block 2, loop depth 0
;; pred: ENTRY
_5 = .MUL_OVERFLOW (a_3(D), b_4(D));
_1 = REALPART_EXPR <_5>;
abu24 = _1;
_2 = IMAGPART_EXPR <_5>;
_8 = (_Bool) _2;
return _8;
;; succ: EXIT
}
but the RTL that follows has an unconditional jump to the good case.
Target: avr
Configured with: ../../source/gcc-master/configure --target=avr
--enable-languages=c,c++
Thread model: single
Supported LTO compression algorithms: zlib
gcc version 17.0.0 20260614 (experimental) (GCC)
