[PATCH] D34142: [ARM] Add macro fusion for AES instructions.

Florian Hahn via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Jun 19 07:00:48 PDT 2017


fhahn added inline comments.


================
Comment at: lib/Target/ARM/ARMTargetMachine.cpp:399
+    const ARMSubtarget &ST = C->MF->getSubtarget<ARMSubtarget>();
+    if (ST.hasFuseAES())
+      DAG->addMutation(createARMMacroFusionDAGMutation());
----------------
javed.absar wrote:
> evandro wrote:
> > Granted that `fuse-aes` is currently the only fusion, but I wonder if it´d make sense to create a more encompassing `hasFusion()` method that would combine all and any instruction fusion.
> In light of that, should 'shouldScheduleAdjacent' be renamed 'shouldScheduleAdjacentAES' ?
> I understand you have an equivalent patch for AArch64 of this,  so any ask here would need parallel changes there and depending on how things converge on https://reviews.llvm.org/D34144 
I've added a `hasFusion()` function. I think it might be worth adding a similar function on AArch64, which I will do once this change is committed.


https://reviews.llvm.org/D34142





More information about the llvm-commits mailing list