https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113440
Bug ID: 113440
Summary: Missed optimization for redundancy computation
elimination because of missed judgment for unsigned
overflow
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: 652023330028 at smail dot nju.edu.cn
Target Milestone: ---
Hello, we noticed that maybe there is a missed optimization for redundancy
computation elimination because of missed judgment for unsigned overflow.
https://godbolt.org/z/r54ezx9ea
unsigned n;
void func(unsigned a){
if((a+a<a&&a>0) || a==0) return;
n=(a+a)/a;
}
GCC -O3:
func(unsigned int):
lea eax, [rdi+rdi]
lea edx, [rdi-1]
cmp edx, eax
jb .L4
ret
.L4:
xor edx, edx
div edi
mov DWORD PTR n[rip], eax
ret
Expected code (Clang):
func(unsigned int): # @func(unsigned int)
test edi, edi
jle .LBB0_2
mov dword ptr [rip + n], 2
.LBB0_2: # %return
ret
Compare this to:
void func2(unsigned a){
if(a>2147483647 || a==0) return;
n=(a+a)/a;
}
evrp (tree) of func2:
Folding statement: _3 = _2 / a_5(D);
Applying pattern match.pd:954, gimple-match-3.cc:2072
gimple_simplified to _3 = 2;
Global Exported: _3 = [irange] unsigned int [0, 4294967294]
Folded into: _3 = 2;
Therefore, it may be because gcc misses a judgment on unsigned overflow or
value ranges.
Thank you very much for your time and effort! We look forward to hearing from
you.
