AMDGPU: Fix not adding exec to defs of cmpx instruction pseudos

[oota-llvm.git] / lib / Target / AMDGPU / SIFixSGPRLiveRanges.cpp

diff --git a/lib/Target/AMDGPU/SIFixSGPRLiveRanges.cpp b/lib/Target/AMDGPU/SIFixSGPRLiveRanges.cpp
index 3dda827b671ac1fd7129ff20d07d3dd9c0d77b4d..abe331477ad6474dbaa52ade225f0084997879a4 100644 (file)
--- a/lib/Target/AMDGPU/SIFixSGPRLiveRanges.cpp
+++ b/lib/Target/AMDGPU/SIFixSGPRLiveRanges.cpp
@@ -47,6 +47,7 @@
 #include "AMDGPU.h"
 #include "SIInstrInfo.h"
 #include "SIRegisterInfo.h"
+#include "llvm/ADT/DepthFirstIterator.h"
 #include "llvm/CodeGen/LiveIntervalAnalysis.h"
 #include "llvm/CodeGen/LiveVariables.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
@@ -113,16 +114,20 @@ bool SIFixSGPRLiveRanges::runOnMachineFunction(MachineFunction &MF) {
   const TargetInstrInfo *TII = MF.getSubtarget().getInstrInfo();
   const SIRegisterInfo *TRI = static_cast<const SIRegisterInfo *>(
       MF.getSubtarget().getRegisterInfo());
+  bool MadeChange = false;
 
   MachinePostDominatorTree *PDT = &getAnalysis<MachinePostDominatorTree>();
   std::vector<std::pair<unsigned, LiveRange *>> SGPRLiveRanges;
 
   LiveIntervals *LIS = &getAnalysis<LiveIntervals>();
   LiveVariables *LV = getAnalysisIfAvailable<LiveVariables>();
+  MachineBasicBlock *Entry = MF.begin();
 
-  // First pass, collect all live intervals for SGPRs
-  for (const MachineBasicBlock &MBB : MF) {
-    for (const MachineInstr &MI : MBB) {
+  // Use a depth first order so that in SSA, we encounter all defs before
+  // uses. Once the defs of the block have been found, attempt to insert
+  // SGPR_USE instructions in successor blocks if required.
+  for (MachineBasicBlock *MBB : depth_first(Entry)) {
+    for (const MachineInstr &MI : *MBB) {
       for (const MachineOperand &MO : MI.defs()) {
         if (MO.isImplicit())
           continue;
@@ -132,29 +137,23 @@ bool SIFixSGPRLiveRanges::runOnMachineFunction(MachineFunction &MF) {
             // Only consider defs that are live outs. We don't care about def /
             // use within the same block.
             LiveRange &LR = LIS->getInterval(Def);
-            if (LIS->isLiveOutOfMBB(LR, &MBB))
+            if (LIS->isLiveOutOfMBB(LR, MBB))
               SGPRLiveRanges.push_back(std::make_pair(Def, &LR));
           }
         } else if (TRI->isSGPRClass(TRI->getPhysRegClass(Def))) {
-            SGPRLiveRanges.push_back(
-                std::make_pair(Def, &LIS->getRegUnit(Def)));
+          SGPRLiveRanges.push_back(std::make_pair(Def, &LIS->getRegUnit(Def)));
         }
       }
     }
-  }
 
-  // Second pass fix the intervals
-  for (MachineFunction::iterator BI = MF.begin(), BE = MF.end();
-                                                  BI != BE; ++BI) {
-    MachineBasicBlock &MBB = *BI;
-    if (MBB.succ_size() < 2)
+    if (MBB->succ_size() < 2)
       continue;
 
     // We have structured control flow, so the number of successors should be
     // two.
-    assert(MBB.succ_size() == 2);
-    MachineBasicBlock *SuccA = *MBB.succ_begin();
-    MachineBasicBlock *SuccB = *(++MBB.succ_begin());
+    assert(MBB->succ_size() == 2);
+    MachineBasicBlock *SuccA = *MBB->succ_begin();
+    MachineBasicBlock *SuccB = *(++MBB->succ_begin());
     MachineBasicBlock *NCD = PDT->findNearestCommonDominator(SuccA, SuccB);
 
     if (!NCD)
@@ -181,16 +180,25 @@ bool SIFixSGPRLiveRanges::runOnMachineFunction(MachineFunction &MF) {
       bool LiveInToA = LIS->isLiveInToMBB(*LR, SuccA);
       bool LiveInToB = LIS->isLiveInToMBB(*LR, SuccB);
 
-      if ((!LiveInToA && !LiveInToB) ||
-          (LiveInToA && LiveInToB))
+      if (!LiveInToA && !LiveInToB) {
+        DEBUG(dbgs() << PrintReg(Reg, TRI, 0)
+              << " is live into neither successor\n");
+        continue;
+      }
+
+      if (LiveInToA && LiveInToB) {
+        DEBUG(dbgs() << PrintReg(Reg, TRI, 0)
+              << " is live into both successors\n");
         continue;
+      }
 
       // This interval is live in to one successor, but not the other, so
       // we need to update its range so it is live in to both.
-      DEBUG(dbgs() << "Possible SGPR conflict detected " <<  " in " << *LR <<
-                      " BB#" << SuccA->getNumber() << ", BB#" <<
-                      SuccB->getNumber() <<
-                      " with NCD = " << NCD->getNumber() << '\n');
+      DEBUG(dbgs() << "Possible SGPR conflict detected for "
+            << PrintReg(Reg, TRI, 0) <<  " in " << *LR
+            << " BB#" << SuccA->getNumber() << ", BB#"
+            << SuccB->getNumber()
+            << " with NCD = BB#" << NCD->getNumber() << '\n');
 
       assert(TargetRegisterInfo::isVirtualRegister(Reg) &&
              "Not expecting to extend live range of physreg");
@@ -202,6 +210,8 @@ bool SIFixSGPRLiveRanges::runOnMachineFunction(MachineFunction &MF) {
                 TII->get(AMDGPU::SGPR_USE))
         .addReg(Reg, RegState::Implicit);
 
+      MadeChange = true;
+
       SlotIndex SI = LIS->InsertMachineInstrInMaps(NCDSGPRUse);
       LIS->extendToIndices(*LR, SI.getRegSlot());
 
@@ -214,5 +224,5 @@ bool SIFixSGPRLiveRanges::runOnMachineFunction(MachineFunction &MF) {
     }
   }
 
-  return false;
+  return MadeChange;
 }