[PATCH] D33230: [AArch64] Make instruction fusion more aggressive.

[Florian Hahn via Phabricator via llvm-commits]

fhahn added inline comments.


================
Comment at: lib/CodeGen/MachineScheduler.cpp:3236
 
+  // Keep clustered nodes together.
+  if (tryGreater(TryCand.SU == DAG->getNextClusterSucc(),
----------------
javed.absar wrote:
> This seems to change core PostRA-Scheduler logic and that may impact other targets. On the other hand, I see that GenericScheduler tries ClusteredNodes candidates and this was missing from here (perhaps for no good reason). 
> 
Yes that change may impact other targets, but I think it makes sense to try keep clustered instruction together during PostRA scheduling as well, as GenericScheduler does. 


https://reviews.llvm.org/D33230