https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89134
Bug ID: 89134
Summary: A missing optimization opportunity for a simple branch
in loop
Product: gcc
Version: 9.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: jiangning.liu at amperecomputing dot com
Target Milestone: ---
For this simple case,
__attribute__((pure)) __attribute__((noinline)) int inc(int i)
{
/* Do something else here */
return i+1;
}
extern int do_something(void);
extern int b;
void test(int n)
{
for (int i=0; i<n; i=inc(i))
{
if (b) {
b = do_something();
}
}
}
"GCC -O3" is generating a loop as below,
.L6:
cbz w1, .L5
bl do_something
mov w1, w0
mov w0, w19
str w1, [x21]
bl inc
mov w19, w0
cmp w20, w0
bgt .L6
.L3:
ldp x19, x20, [sp, 16]
ldr x21, [sp, 32]
ldp x29, x30, [sp], 48
ret
.L5:
mov w0, w19
bl inc
mov w19, w0
cmp w20, w0
ble .L3
mov w0, w19
bl inc
mov w19, w0
cmp w20, w0
bgt .L5
b .L3
.L13:
ret
But the loop with head at label .L5 can be completely optimized away, because
inc is a pure function and it is essentially an empty loop doing nothing.
Therefore the code can be optimized to be like,
.L6:
cbz w1, .L3
bl do_something
mov w1, w0
mov w0, w19
str w1, [x21]
bl inc
mov w19, w0
cmp w20, w0
bgt .L6
.L3:
ldp x19, x20, [sp, 16]
ldr x21, [sp, 32]
ldp x29, x30, [sp], 48
ret
