https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111562
Bug ID: 111562
Summary: Missed optimization when the value of another variable
can be used directly
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 found some optimizations that GCC may have missed. We would greatly
appreicate if you can take a look and let us know what you think.
Example 1:
https://godbolt.org/z/nPW7j1xvq
Given the following code:
```c++
int var22=0;
int var23=0;
void test(int var7, int var9, int var11, int var12) {
var22 = var12 + var9 + var11;
var23 = var12 + (var9 + var7 + var11);
}
```
In fact, `var23` can be directly optimized to `var22+var7`, but gcc-trunk -O3
does not:
```asm
test(int, int, int, int):
lea eax, [rcx+rsi]
add esi, edi
add esi, edx
add eax, edx
add esi, ecx
mov DWORD PTR var22[rip], eax
mov DWORD PTR var23[rip], esi
ret
var23:
.zero 4
var22:
.zero 4
```
Example 2:
https://godbolt.org/z/8vE69WPc7
Given the following code:
```c++
extern int var_15;
extern int var_16;
void test(int var_2, int var_3, int var_4, int var_6, int var_9) {
var_15 = var_9 - (var_2 + var_6 + var_3 + var_4);
var_16 = var_4 + var_6 + var_2;
}
```
By adjusting the calculation order, the `var_16` calculation can actually use
the intermediate result of the `var_15` calculation process directly, but
gcc-trunk -O3 does not:
```asm
test(int, int, int, int, int):
lea eax, [rdi+rcx]
add ecx, edx
add eax, esi
add ecx, edi
add eax, edx
mov DWORD PTR var_16[rip], ecx
sub r8d, eax
mov DWORD PTR var_15[rip], r8d
ret
```
Thank you very much for your time and effort! We look forward to hearing from
you.