https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128035
Bug ID: 128035
Summary: failure to inline some memcmp when -mno-sse is passed
Product: gcc
Version: 16.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
Assignee: unassigned at gcc dot gnu.org
Reporter: mjguzik at gmail dot com
Target Milestone: ---
Consider:
#define BUFSIZE 40
struct foo {
char buf[BUFSIZE];
};
int test(struct foo *p1, struct foo *p2) {
return __builtin_memcmp(p1, p2, sizeof(struct foo)) == 0;
}
Compiling with -O2 -mno-sse results in a call to memcmp. However, compiling
with mere -O2 gives inlined asm without simd. Thus I would expect such a result
for the -mno-sse case as well.
For reference:
-O2:
sub rsp, 8
mov edx, 40
call "memcmp"
test eax, eax
sete al
add rsp, 8
movzx eax, al
ret
-O2 -mno-sse:
mov rax, QWORD PTR [rdi]
mov rdx, QWORD PTR [rdi+8]
xor rax, QWORD PTR [rsi]
xor rdx, QWORD PTR [rsi+8]
or rax, rdx
je .L5
.L2:
mov eax, 1
.L3:
xor eax, 1
ret
.L5:
mov rax, QWORD PTR [rdi+16]
mov rdx, QWORD PTR [rdi+24]
xor rax, QWORD PTR [rsi+16]
xor rdx, QWORD PTR [rsi+24]
or rax, rdx
jne .L2
mov rax, QWORD PTR [rsi+32]
cmp QWORD PTR [rdi+32], rax
jne .L2
xor eax, eax
jmp .L3
