Closed blackgeorge-boom closed 1 year ago
The ccode
is defined differently in both architectures:
AArch64: def ccode : Operand<i32> {
X86: def ccode : Operand<i8> {
When code is lowered to instructions like:
AArch64: %44:gpr32 = CSINCWr $wzr, $wzr, 13, implicit $nzcv
X86: %33:gr8 = SETCCr 15, implicit $eflags
in X86, the result is stored as one byte and aligned at one-byte boundary (not sure why our -align-bytes-to-four
option does not cover this case.