x86 optimization: logical AND and OR in if-conditionals can turn to multiple branch instructions #160612
Description
extern void subroutine_foo(void);
extern void subroutine_bar(void);
void func_a(int x, int y) {
if (x == 0 || y == 0)
subroutine_foo();
else
subroutine_bar();
}
void func_b(int x, int y) {
if (x == 0)
subroutine_foo();
else if (y == 0)
subroutine_foo();
else
subroutine_bar();
}x86-64 clang-trunk-20250924 with -Os option produces:
func_a:
testl %edi, %edi
setne %al
testl %esi, %esi
setne %cl
testb %cl, %al
jne subroutine_bar@PLT
jmp subroutine_foo@PLT
func_b:
testl %edi, %edi
je subroutine_foo@PLT
testl %esi, %esi
je subroutine_foo@PLT
jmp subroutine_bar@PLThttps://godbolt.org/z/ns1YxeKr5
Clang misses that there's no need to make temporary boolean variables for the conditional. It can use multiple branch instructions instead.
This issue looks x86 specific. Clang for AArch64 target generates optimal code.