0bd5bbb31e
This patch adds a test case showing how a single extra .loc can cause binary differences when using -x86-pad-for-align=true. The issue has been discussed in D94542, PR42138, PR48742.
44 lines
624 B
ArmAsm
44 lines
624 B
ArmAsm
# See PR48742.
|
|
.text
|
|
.p2align 4
|
|
foo:
|
|
movq %rdi, %rax
|
|
.p2align 4,,10
|
|
.p2align 3
|
|
L1:
|
|
movzbl (%rax), %edx
|
|
cmpb $10, %dl
|
|
je L4
|
|
L2:
|
|
cmpb $100, %dl
|
|
je L5
|
|
cmpb $200, %dl
|
|
je L5
|
|
cmpb $300, %dl
|
|
jne L5
|
|
.p2align 4,,10
|
|
.p2align 3
|
|
L3:
|
|
movq %rax, %rdx
|
|
incq %rax
|
|
cmpb $30, (%rax)
|
|
jne L3
|
|
leaq 2(%rdx), %rax
|
|
movzbl (%rax), %edx
|
|
cmpb $90, %dl
|
|
jne L2
|
|
.p2align 4,,10
|
|
.p2align 3
|
|
L4:
|
|
cmpb $99, 4(%rax)
|
|
je L7
|
|
L5:
|
|
incq %rax
|
|
jmp L1
|
|
.p2align 4,,10
|
|
.p2align 3
|
|
L6:
|
|
ret
|
|
L7:
|
|
ret
|