From e2ee98ab169fe8d1d4bd39fe0ecb89274eceb438 Mon Sep 17 00:00:00 2001
From: Hal Finkel <hfinkel@anl.gov>
Date: Thu, 27 Mar 2014 23:12:31 +0000
Subject: [PATCH] [PowerPC] Use a small cleanup pass to remove VSX self copies

As explained in r204976, because of how the allocation of VSX registers
interacts with the call-lowering code, we sometimes end up generating self VSX
copies. Specifically, things like this:
  %VSL2<def> = COPY %F2, %VSL2<imp-use,kill>
(where %F2 is really a sub-register of %VSL2, and so this copy is a nop)

This adds a small cleanup pass to remove these prior to post-RA scheduling.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@204980 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/PowerPC/PPC.h                |  1 +
 lib/Target/PowerPC/PPCInstrInfo.cpp     | 74 +++++++++++++++++++++++++
 lib/Target/PowerPC/PPCTargetMachine.cpp |  3 +
 test/CodeGen/PowerPC/vsx-self-copy.ll   | 27 +++++++++
 4 files changed, 105 insertions(+)
 create mode 100644 test/CodeGen/PowerPC/vsx-self-copy.ll

diff --git a/lib/Target/PowerPC/PPC.h b/lib/Target/PowerPC/PPC.h
index df8dd8f28ba..c42c5be14be 100644
--- a/lib/Target/PowerPC/PPC.h
+++ b/lib/Target/PowerPC/PPC.h
@@ -37,6 +37,7 @@ namespace llvm {
 #endif
   FunctionPass *createPPCEarlyReturnPass();
   FunctionPass *createPPCVSXCopyPass();
+  FunctionPass *createPPCVSXCopyCleanupPass();
   FunctionPass *createPPCVSXFMAMutatePass();
   FunctionPass *createPPCBranchSelectionPass();
   FunctionPass *createPPCISelDag(PPCTargetMachine &TM);
diff --git a/lib/Target/PowerPC/PPCInstrInfo.cpp b/lib/Target/PowerPC/PPCInstrInfo.cpp
index be4dafa0e30..0f799017916 100644
--- a/lib/Target/PowerPC/PPCInstrInfo.cpp
+++ b/lib/Target/PowerPC/PPCInstrInfo.cpp
@@ -1969,6 +1969,80 @@ char PPCVSXCopy::ID = 0;
 FunctionPass*
 llvm::createPPCVSXCopyPass() { return new PPCVSXCopy(); }
 
+#undef DEBUG_TYPE
+#define DEBUG_TYPE "ppc-vsx-copy-cleanup"
+
+namespace llvm {
+  void initializePPCVSXCopyCleanupPass(PassRegistry&);
+}
+
+namespace {
+  // PPCVSXCopyCleanup pass - We sometimes end up generating self copies of VSX
+  // registers (mostly because the ABI code still places all values into the
+  // "traditional" floating-point and vector registers). Remove them here.
+  struct PPCVSXCopyCleanup : public MachineFunctionPass {
+    static char ID;
+    PPCVSXCopyCleanup() : MachineFunctionPass(ID) {
+      initializePPCVSXCopyCleanupPass(*PassRegistry::getPassRegistry());
+    }
+
+    const PPCTargetMachine *TM;
+    const PPCInstrInfo *TII;
+
+protected:
+    bool processBlock(MachineBasicBlock &MBB) {
+      bool Changed = false;
+
+      SmallVector<MachineInstr *, 4> ToDelete;
+      for (MachineBasicBlock::iterator I = MBB.begin(), IE = MBB.end();
+           I != IE; ++I) {
+        MachineInstr *MI = I;
+        if (MI->getOpcode() == PPC::XXLOR &&
+            MI->getOperand(0).getReg() == MI->getOperand(1).getReg() &&
+            MI->getOperand(0).getReg() == MI->getOperand(2).getReg())
+          ToDelete.push_back(MI);
+      }
+
+      if (!ToDelete.empty())
+        Changed = true;
+
+      for (unsigned i = 0, ie = ToDelete.size(); i != ie; ++i) {
+        DEBUG(dbgs() << "Removing VSX self-copy: " << *ToDelete[i]);
+        ToDelete[i]->eraseFromParent();
+      }
+
+      return Changed;
+    }
+
+public:
+    virtual bool runOnMachineFunction(MachineFunction &MF) {
+      TM = static_cast<const PPCTargetMachine *>(&MF.getTarget());
+      TII = TM->getInstrInfo();
+
+      bool Changed = false;
+
+      for (MachineFunction::iterator I = MF.begin(); I != MF.end();) {
+        MachineBasicBlock &B = *I++;
+        if (processBlock(B))
+          Changed = true;
+      }
+
+      return Changed;
+    }
+
+    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+      MachineFunctionPass::getAnalysisUsage(AU);
+    }
+  };
+}
+
+INITIALIZE_PASS(PPCVSXCopyCleanup, DEBUG_TYPE,
+                "PowerPC VSX Copy Cleanup", false, false)
+
+char PPCVSXCopyCleanup::ID = 0;
+FunctionPass*
+llvm::createPPCVSXCopyCleanupPass() { return new PPCVSXCopyCleanup(); }
+
 #undef DEBUG_TYPE
 #define DEBUG_TYPE "ppc-early-ret"
 STATISTIC(NumBCLR, "Number of early conditional returns");
diff --git a/lib/Target/PowerPC/PPCTargetMachine.cpp b/lib/Target/PowerPC/PPCTargetMachine.cpp
index cb869bd91eb..e7438f394cc 100644
--- a/lib/Target/PowerPC/PPCTargetMachine.cpp
+++ b/lib/Target/PowerPC/PPCTargetMachine.cpp
@@ -182,6 +182,9 @@ bool PPCPassConfig::addPreRegAlloc() {
 }
 
 bool PPCPassConfig::addPreSched2() {
+  if (getPPCSubtarget().hasVSX())
+    addPass(createPPCVSXCopyCleanupPass());
+
   if (getOptLevel() != CodeGenOpt::None)
     addPass(&IfConverterID);
 
diff --git a/test/CodeGen/PowerPC/vsx-self-copy.ll b/test/CodeGen/PowerPC/vsx-self-copy.ll
new file mode 100644
index 00000000000..23615ca10c1
--- /dev/null
+++ b/test/CodeGen/PowerPC/vsx-self-copy.ll
@@ -0,0 +1,27 @@
+; RUN: llc -mcpu=pwr7 -mattr=+vsx < %s | FileCheck %s
+target datalayout = "E-m:e-i64:64-n32:64"
+target triple = "powerpc64-unknown-linux-gnu"
+
+define double @takFP(double %x, double %y, double %z) #0 {
+entry:
+  br i1 undef, label %if.then, label %return
+
+if.then:                                          ; preds = %if.then, %entry
+  %x.tr16 = phi double [ %call, %if.then ], [ %x, %entry ]
+  %call = tail call double @takFP(double undef, double undef, double undef)
+  %call4 = tail call double @takFP(double undef, double %x.tr16, double undef)
+  %cmp = fcmp olt double undef, %call
+  br i1 %cmp, label %if.then, label %return
+
+return:                                           ; preds = %if.then, %entry
+  %z.tr.lcssa = phi double [ %z, %entry ], [ %call4, %if.then ]
+  ret double %z.tr.lcssa
+
+; CHECK: @takFP
+; CHECK-NOT: xxlor 0, 0, 0
+; CHECK: blr
+}
+
+attributes #0 = { nounwind readnone }
+attributes #1 = { nounwind }
+
-- 
2.34.1