[llvm] [AMDGPU] Avoid bundling memops in a MBB if the MBB contains a SCHED_BARRIER (PR #153533)

[Yoonseo Choi via llvm-commits]

================
@@ -114,7 +114,7 @@ body:             |
     $vgpr3 = DS_READ_B32_gfx9 $vgpr1, 0, 0, implicit $exec
 ...
 
-# Middle dbg_value should be bundled
+# Loads are not bundled due to middle deb_value
----------------
yoonseoch wrote:

Back to original. 

https://github.com/llvm/llvm-project/pull/153533