[FIX] Initialize incoming scalar memory locations for PHIs

llvm-svn: 252437

polly/lib/CodeGen/BlockGenerators.cpp

@@ -466,6 +466,8 @@ void BlockGenerator::generateScalarStores(ScopStmt &Stmt, LoopToScevMapT &LTS,
 
 void BlockGenerator::createScalarInitialization(Scop &S) {
   Region &R = S.getRegion();
+  BasicBlock *ExitBB = R.getExit();
+
   // The split block __just before__ the region and optimized region.
   BasicBlock *SplitBB = R.getEnteringBlock();
   BranchInst *SplitBBTerm = cast<BranchInst>(SplitBB->getTerminator());
@@ -509,10 +511,13 @@ void BlockGenerator::createScalarInitialization(Scop &S) {
     if (Inst && R.contains(Inst))
       continue;
 
-    // PHI nodes that are not marked as such in their SAI object are exit PHI
-    // nodes we model as common scalars but do not need to initialize.
-    if (Inst && isa<PHINode>(Inst))
-      continue;
+    // PHI nodes that are not marked as such in their SAI object are either exit
+    // PHI nodes we model as common scalars but without initialization, or
+    // incoming phi nodes that need to be initialized. Check if the first is the
+    // case for Inst and do not create and initialize memory if so.
+    if (auto *PHI = dyn_cast_or_null<PHINode>(Inst))
+      if (!S.hasSingleExitEdge() && PHI->getBasicBlockIndex(ExitBB) >= 0)
+        continue;
 
     Builder.CreateStore(Array->getBasePtr(),
                         getOrCreateScalarAlloca(Array->getBasePtr()));

polly/test/Isl/CodeGen/uninitialized_scalar_memory.ll

@@ -0,0 +1,88 @@
+; RUN: opt %loadPolly -S -polly-codegen < %s | FileCheck %s
+;
+; Verify we initialize the scalar locations reserved for the incoming phi
+; values.
+;
+; CHECK:      polly.start:
+; CHECK-NEXT:   store float %ebig.0, float* %ebig.0.s2a
+; CHECK-NEXT:   store i32 %iebig.0, i32* %iebig.0.s2a
+; CHECK-NEXT:   br label %polly.stmt.if.end.entry
+;
+;    int g(void);
+;    float M;
+;    int max(float *restrict xbig, int eres, int bres, float *restrict indx) {
+;      int i, iebig;
+;      float ebig;
+;      for (i = 0; i < 4 + eres; i++) {
+;        if (g())
+;          break;
+;
+;        if (xbig[i] > ebig) {
+;          ebig = xbig[i];
+;          iebig = (int)(indx[i] + bres);
+;        }
+;      }
+;      return (iebig);
+;    }
+;
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+@M = common global float 0.000000e+00, align 4
+
+define i32 @max(float* noalias %xbig, i32 %eres, i32 %bres, float* noalias %indx) {
+entry:
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.inc ], [ 0, %entry ]
+  %iebig.0 = phi i32 [ undef, %entry ], [ %iebig.1, %for.inc ]
+  %ebig.0 = phi float [ undef, %entry ], [ %ebig.1, %for.inc ]
+  %add = add nsw i32 %eres, 4
+  %tmp = sext i32 %add to i64
+  %cmp = icmp slt i64 %indvars.iv, %tmp
+  br i1 %cmp, label %for.body, label %for.end.loopexit
+
+for.body:                                         ; preds = %for.cond
+  %call = call i32 @g() #2
+  %tobool = icmp eq i32 %call, 0
+  br i1 %tobool, label %if.end, label %if.then
+
+if.then:                                          ; preds = %for.body
+  %iebig.0.lcssa1 = phi i32 [ %iebig.0, %for.body ]
+  br label %for.end
+
+if.end:                                           ; preds = %for.body
+  %arrayidx = getelementptr inbounds float, float* %xbig, i64 %indvars.iv
+  %tmp3 = load float, float* %arrayidx, align 4
+  %cmp1 = fcmp ogt float %tmp3, %ebig.0
+  br i1 %cmp1, label %if.then.2, label %if.end.9
+
+if.then.2:                                        ; preds = %if.end
+  %arrayidx4 = getelementptr inbounds float, float* %xbig, i64 %indvars.iv
+  %tmp4 = load float, float* %arrayidx4, align 4
+  %arrayidx6 = getelementptr inbounds float, float* %indx, i64 %indvars.iv
+  %tmp5 = load float, float* %arrayidx6, align 4
+  %conv = sitofp i32 %bres to float
+  %add7 = fadd float %tmp5, %conv
+  %conv8 = fptosi float %add7 to i32
+  br label %if.end.9
+
+if.end.9:                                         ; preds = %if.then.2, %if.end
+  %iebig.1 = phi i32 [ %conv8, %if.then.2 ], [ %iebig.0, %if.end ]
+  %ebig.1 = phi float [ %tmp4, %if.then.2 ], [ %ebig.0, %if.end ]
+  br label %for.inc
+
+for.inc:                                          ; preds = %if.end.9
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  br label %for.cond
+
+for.end.loopexit:                                 ; preds = %for.cond
+  %iebig.0.lcssa = phi i32 [ %iebig.0, %for.cond ]
+  br label %for.end
+
+for.end:                                          ; preds = %for.end.loopexit, %if.then
+  %iebig.02 = phi i32 [ %iebig.0.lcssa, %for.end.loopexit ], [ %iebig.0.lcssa1, %if.then ]
+  ret i32 %iebig.02
+}
+
+declare i32 @g() #1