[PowerPC] Fix PPCLoopPreIncPrep for depth > 1 loops

This pass had the same problem as the data-prefetching pass: it was only
checking for depth == 1 loops in practice. Fix that, add some debugging
statements, and make sure that, when we grab an AddRec, it is for the loop we
expect.

llvm-svn: 234670

lib/Target/PowerPC/PPCLoopPreIncPrep.cpp

@@ -22,6 +22,7 @@
 #define DEBUG_TYPE "ppc-loop-preinc-prep"
 #include "PPC.h"
 #include "PPCTargetMachine.h"
+#include "llvm/ADT/DepthFirstIterator.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/Statistic.h"
@@ -143,8 +144,10 @@ bool PPCLoopPreIncPrep::runOnFunction(Function &F) {
 
   bool MadeChange = false;
 
-  for (LoopInfo::iterator I = LI->begin(), E = LI->end();
-       I != E; ++I) {
+  if (LI->empty())
+    return MadeChange;
+
+  for (auto I = df_begin(*LI->begin()), E = df_end(*LI->begin()); I != E; ++I) {
     Loop *L = *I;
     MadeChange |= runOnLoop(L);
   }
@@ -159,16 +162,15 @@ bool PPCLoopPreIncPrep::runOnLoop(Loop *L) {
   if (!L->empty())
     return MadeChange;
 
+  DEBUG(dbgs() << "PIP: Examining: " << *L << "\n");
+
   BasicBlock *Header = L->getHeader();
 
   const PPCSubtarget *ST =
     TM ? TM->getSubtargetImpl(*Header->getParent()) : nullptr;
 
-  unsigned HeaderLoopPredCount = 0;
-  for (pred_iterator PI = pred_begin(Header), PE = pred_end(Header);
-       PI != PE; ++PI) {
-    ++HeaderLoopPredCount;
-  }
+  unsigned HeaderLoopPredCount =
+    std::distance(pred_begin(Header), pred_end(Header));
 
   // Collect buckets of comparable addresses used by loads and stores.
   typedef std::multimap<const SCEV *, Instruction *, SCEVLess> Bucket;
@@ -205,9 +207,13 @@ bool PPCLoopPreIncPrep::runOnLoop(Loop *L) {
       if (L->isLoopInvariant(PtrValue))
         continue;
 
-      const SCEV *LSCEV = SE->getSCEV(PtrValue);
-      if (!isa<SCEVAddRecExpr>(LSCEV))
+      const SCEV *LSCEV = SE->getSCEVAtScope(PtrValue, L);
+      if (const SCEVAddRecExpr *LARSCEV = dyn_cast<SCEVAddRecExpr>(LSCEV)) {
+        if (LARSCEV->getLoop() != L)
+          continue;
+      } else {
         continue;
+      }
 
       bool FoundBucket = false;
       for (unsigned i = 0, e = Buckets.size(); i != e; ++i)
@@ -236,11 +242,16 @@ bool PPCLoopPreIncPrep::runOnLoop(Loop *L) {
   // returns a value (which might contribute to determining the loop's
   // iteration space), insert a new preheader for the loop.
   if (!LoopPredecessor ||
-      !LoopPredecessor->getTerminator()->getType()->isVoidTy())
+      !LoopPredecessor->getTerminator()->getType()->isVoidTy()) {
     LoopPredecessor = InsertPreheaderForLoop(L, this);
+    if (LoopPredecessor)
+      MadeChange = true;
+  }
   if (!LoopPredecessor)
     return MadeChange;
 
+  DEBUG(dbgs() << "PIP: Found " << Buckets.size() << " buckets\n");
+
   SmallSet<BasicBlock *, 16> BBChanged;
   for (unsigned i = 0, e = Buckets.size(); i != e; ++i) {
     // The base address of each bucket is transformed into a phi and the others
@@ -251,6 +262,10 @@ bool PPCLoopPreIncPrep::runOnLoop(Loop *L) {
     if (!BasePtrSCEV->isAffine())
       continue;
 
+    DEBUG(dbgs() << "PIP: Transforming: " << *BasePtrSCEV << "\n");
+    assert(BasePtrSCEV->getLoop() == L &&
+           "AddRec for the wrong loop?");
+
     Instruction *MemI = Buckets[i].begin()->second;
     Value *BasePtr = GetPointerOperand(MemI);
     assert(BasePtr && "No pointer operand");
@@ -271,6 +286,8 @@ bool PPCLoopPreIncPrep::runOnLoop(Loop *L) {
     if (!isSafeToExpand(BasePtrStartSCEV, *SE))
       continue;
 
+    DEBUG(dbgs() << "PIP: New start is: " << *BasePtrStartSCEV << "\n");
+
     PHINode *NewPHI = PHINode::Create(I8PtrTy, HeaderLoopPredCount,
       MemI->hasName() ? MemI->getName() + ".phi" : "",
       Header->getFirstNonPHI());

test/CodeGen/PowerPC/pip-inner.ll

@@ -0,0 +1,52 @@
+; RUN: llc < %s | FileCheck %s
+target datalayout = "E-m:e-i64:64-n32:64"
+target triple = "powerpc64-unknown-linux-gnu"
+
+; Function Attrs: nounwind
+define void @foo(double* %x, double* nocapture readonly %y) #0 {
+entry:
+  br label %for.cond1.preheader
+
+for.cond1.preheader:                              ; preds = %for.end, %entry
+  %i.015 = phi i32 [ 0, %entry ], [ %inc7, %for.end ]
+  br label %for.body3
+
+for.body3:                                        ; preds = %for.body3, %for.cond1.preheader
+  %indvars.iv = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next, %for.body3 ]
+  %arrayidx = getelementptr inbounds double, double* %y, i64 %indvars.iv
+  %0 = load double, double* %arrayidx, align 8
+  %add = fadd double %0, 1.000000e+00
+  %arrayidx5 = getelementptr inbounds double, double* %x, i64 %indvars.iv
+  store double %add, double* %arrayidx5, align 8
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, 16000
+  br i1 %exitcond, label %for.end, label %for.body3
+
+for.end:                                          ; preds = %for.body3
+  tail call void @bar(double* %x) #2
+  %inc7 = add nuw nsw i32 %i.015, 1
+  %exitcond16 = icmp eq i32 %inc7, 1000
+  br i1 %exitcond16, label %for.end8, label %for.cond1.preheader
+
+for.end8:                                         ; preds = %for.end
+  ret void
+
+; CHECK-LABEL: @foo
+
+; CHECK: lfdu [[REG1:[0-9]+]], 8({{[0-9]+}})
+; CHECK: fadd [[REG2:[0-9]+]], [[REG1]], {{[0-9]+}}
+; CHECK: stfdu [[REG2]], 8({{[0-9]+}})
+; CHECK: bdnz
+
+; CHECK: bl bar
+; CHECK-NEXT: nop
+
+; CHECK: blr
+}
+
+declare void @bar(double*) #1
+
+attributes #0 = { nounwind "target-cpu"="a2" }
+attributes #1 = { "target-cpu"="a2" }
+attributes #2 = { nounwind }
+