[PATCH] D19924: Representing bundle locked groups as fragments in MCAssembler

[Colin LeMahieu via llvm-commits]

colinl added a comment.

I'll add a test for multiple bundle alignment directives.

The semantics are slightly different than a plain alignment.  Plain align will always pad to an alignment boundary though bundle locking only does this if the instruction crosses the boundary.

{ sub } ; 4-byte
.align 16
{ add } ; 8-byte
{ sub } ; 4-byte

Would assemble as
0: { sub
4:   nop
8:   nop
12: nop }
16: { add }
24: { sub }

Whereas bundle locking would do.

{ sub } ; 4-byte
.bundle_align_mode 4
{ add; add } ; 8-byte
.bundle_align_mode 0
{ sub } ; 4-byte

Would assemble as
0: { sub }
4: { add }
12: { sub }


Repository:
  rL LLVM

http://reviews.llvm.org/D19924