[AArch64] Make instruction fusion more aggressive.

Summary:
This patch makes instruction fusion more aggressive by
* adding artificial edges between the successors of FirstSU and
  SecondSU, similar to BaseMemOpClusterMutation::clusterNeighboringMemOps.
* updating PostGenericScheduler::tryCandidate to keep clusters together,
   similar to GenericScheduler::tryCandidate.

This change increases the number of AES instruction pairs generated on
 Cortex-A57 and Cortex-A72. This doesn't change code at all in
 most benchmarks or general code, but we've seen improvement on kernels
 using AESE/AESMC and AESD/AESIMC. 

Reviewers: evandro, kristof.beyls, t.p.northover, silviu.baranga, atrick, rengolin, MatzeB

Reviewed By: evandro

Subscribers: aemerson, rengolin, MatzeB, javed.absar, llvm-commits

Differential Revision: https://reviews.llvm.org/D33230

llvm-svn: 303618

3 files changed, 20 insertions, 1 deletions

diff --git a/llvm/lib/CodeGen/MachineScheduler.cpp b/llvm/lib/CodeGen/MachineScheduler.cpp
index 41e161f71e5..72b7ad47e09 100644
--- a/llvm/lib/CodeGen/MachineScheduler.cpp
+++ b/llvm/lib/CodeGen/MachineScheduler.cpp
@@ -3233,6 +3233,12 @@ void PostGenericScheduler::tryCandidate(SchedCandidate &Cand,
               Top.getLatencyStallCycles(Cand.SU), TryCand, Cand, Stall))
     return;
 
+  // Keep clustered nodes together.
+  if (tryGreater(TryCand.SU == DAG->getNextClusterSucc(),
+                 Cand.SU == DAG->getNextClusterSucc(),
+                 TryCand, Cand, Cluster))
+    return;
+
   // Avoid critical resource consumption and balance the schedule.
   if (tryLess(TryCand.ResDelta.CritResources, Cand.ResDelta.CritResources,
               TryCand, Cand, ResourceReduce))
diff --git a/llvm/lib/Target/AArch64/AArch64MacroFusion.cpp b/llvm/lib/Target/AArch64/AArch64MacroFusion.cpp
index a6926a6700e..3b71d529db5 100644
--- a/llvm/lib/Target/AArch64/AArch64MacroFusion.cpp
+++ b/llvm/lib/Target/AArch64/AArch64MacroFusion.cpp
@@ -232,6 +232,19 @@ static bool scheduleAdjacentImpl(ScheduleDAGMI *DAG, SUnit &AnchorSU) {
           dbgs() << DAG->TII->getName(FirstMI->getOpcode()) << " - " <<
                     DAG->TII->getName(SecondMI->getOpcode()) << '\n'; );
 
+    if (&SecondSU != &DAG->ExitSU)
+      // Make instructions dependent on FirstSU also dependent on SecondSU to
+      // prevent them from being scheduled between FirstSU and and SecondSU.
+      for (SUnit::const_succ_iterator
+             SI = FirstSU.Succs.begin(), SE = FirstSU.Succs.end();
+           SI != SE; ++SI) {
+        if (!SI->getSUnit() || SI->getSUnit() == &SecondSU)
+          continue;
+        DEBUG(dbgs() << "  Copy Succ ";
+              SI->getSUnit()->print(dbgs(), DAG); dbgs() << '\n';);
+        DAG->addEdge(SI->getSUnit(), SDep(&SecondSU, SDep::Artificial));
+      }
+
     ++NumFused;
     return true;
   }
diff --git a/llvm/lib/Target/AArch64/AArch64TargetMachine.cpp b/llvm/lib/Target/AArch64/AArch64TargetMachine.cpp
index 132f192f2a9..82b1ff7b1fa 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetMachine.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetMachine.cpp
@@ -277,7 +277,7 @@ public:
   ScheduleDAGInstrs *
   createPostMachineScheduler(MachineSchedContext *C) const override {
     const AArch64Subtarget &ST = C->MF->getSubtarget<AArch64Subtarget>();
-    if (ST.hasFuseLiterals()) {
+    if (ST.hasFuseAES() || ST.hasFuseLiterals()) {
       // Run the Macro Fusion after RA again since literals are expanded from
       // pseudos then (v. addPreSched2()).
       ScheduleDAGMI *DAG = createGenericSchedPostRA(C);