final step needed to resolve PR6627, which allows us to flatten the code down to

a nice and tidy:
  %x1 = load i32* %0, align 4
  %1 = icmp eq i32 %x1, 1179403647
  br i1 %1, label %if.then, label %if.end

instead of doing lots of loads and branches.  May the FreeBSD bootloader
long fit in its allocated space.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130416 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Transforms/Scalar/GVN.cpp

@@ -432,14 +432,14 @@ namespace {
     /// addToLeaderTable - Push a new Value to the LeaderTable onto the list for
     /// its value number.
     void addToLeaderTable(uint32_t N, Value *V, BasicBlock *BB) {
-      LeaderTableEntry& Curr = LeaderTable[N];
+      LeaderTableEntry &Curr = LeaderTable[N];
       if (!Curr.Val) {
         Curr.Val = V;
         Curr.BB = BB;
         return;
       }
       
-      LeaderTableEntry* Node = TableAllocator.Allocate<LeaderTableEntry>();
+      LeaderTableEntry *Node = TableAllocator.Allocate<LeaderTableEntry>();
       Node->Val = V;
       Node->BB = BB;
       Node->Next = Curr.Next;
@@ -944,7 +944,10 @@ static Value *GetLoadValueForLoad(LoadInst *SrcVal, unsigned Offset,
 
     Value *PtrVal = SrcVal->getPointerOperand();
     
-    IRBuilder<> Builder(SrcVal->getParent(), SrcVal);
+    // Insert the new load after the old load.  This ensures that subsequent
+    // memdep queries will find the new load.  We can't easily remove the old
+    // load completely because it is already in the value numbering table.
+    IRBuilder<> Builder(SrcVal->getParent(), ++BasicBlock::iterator(SrcVal));
     const Type *DestPTy = 
       IntegerType::get(LoadTy->getContext(), NewLoadSize*8);
     DestPTy = PointerType::get(DestPTy, 
@@ -967,6 +970,7 @@ static Value *GetLoadValueForLoad(LoadInst *SrcVal, unsigned Offset,
     RV = Builder.CreateTrunc(RV, SrcVal->getType());
     SrcVal->replaceAllUsesWith(RV);
     gvn.getMemDep().removeInstruction(SrcVal);
+    //gvn.markInstructionForDeletion(SrcVal);
     SrcVal = NewLoad;
   }
   

test/Transforms/GVN/rle.ll

@@ -1,4 +1,4 @@
-; RUN: opt < %s -basicaa -gvn -S | FileCheck %s
+; RUN: opt < %s -basicaa -gvn -S -die | FileCheck %s
 
 ; 32-bit little endian target.
 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32"
@@ -597,11 +597,11 @@ if.end:
 ;; Load Widening
 ;;===----------------------------------------------------------------------===;;
 
-%widening1 = type { i32, i8, i8 }
+%widening1 = type { i32, i8, i8, i8, i8 }
 
 @f = global %widening1 zeroinitializer, align 4
 
-define i32 @test_widening1() nounwind ssp noredzone {
+define i32 @test_widening1(i8* %P) nounwind ssp noredzone {
 entry:
   %tmp = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 1), align 4
   %conv = zext i8 %tmp to i32
@@ -616,4 +616,27 @@ entry:
 ; CHECK-ret i32
 }
 
+define i32 @test_widening2() nounwind ssp noredzone {
+entry:
+  %tmp = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 1), align 4
+  %conv = zext i8 %tmp to i32
+  %tmp1 = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 2), align 1
+  %conv2 = zext i8 %tmp1 to i32
+  %add = add nsw i32 %conv, %conv2
+
+  %tmp2 = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 3), align 2
+  %conv3 = zext i8 %tmp2 to i32
+  %add2 = add nsw i32 %add, %conv3
+
+  %tmp3 = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 4), align 1
+  %conv4 = zext i8 %tmp3 to i32
+  %add3 = add nsw i32 %add2, %conv3
+
+  ret i32 %add3
+; CHECK: @test_widening2
+; CHECK-NOT: load
+; CHECK: load i32*
+; CHECK-NOT: load
+; CHECK-ret i32
+}
 

test/Transforms/PhaseOrdering/basic.ll

@@ -60,7 +60,7 @@ if.then:                                          ; preds = %land.lhs.true17
   %call25 = call i32 (...)* @doo()
   br label %if.end
 
-if.end:                                           ; preds = %if.then, %land.lhs.true17, %land.lhs.true9, %land.lhs.true, %entry
+if.end:
   ret void
 
 ; CHECK: @test2
@@ -71,3 +71,47 @@ if.end:                                           ; preds = %if.then, %land.lhs.
 
 declare i32 @doo(...)
 
+; PR6627 - This should all be flattened down to one compare.  This is the same
+; as test2, except that the initial load is done as an i8 instead of i32, thus
+; requiring widening.
+define void @test2a(i8* %arrayidx) nounwind ssp {
+entry:
+  %x1 = load i8* %arrayidx, align 4
+  %conv = zext i8 %x1 to i32
+  %cmp = icmp eq i32 %conv, 127
+  br i1 %cmp, label %land.lhs.true, label %if.end
+
+land.lhs.true:                                    ; preds = %entry
+  %arrayidx4 = getelementptr inbounds i8* %arrayidx, i64 1
+  %tmp5 = load i8* %arrayidx4, align 1
+  %conv6 = zext i8 %tmp5 to i32
+  %cmp7 = icmp eq i32 %conv6, 69
+  br i1 %cmp7, label %land.lhs.true9, label %if.end
+
+land.lhs.true9:                                   ; preds = %land.lhs.true
+  %arrayidx12 = getelementptr inbounds i8* %arrayidx, i64 2
+  %tmp13 = load i8* %arrayidx12, align 1
+  %conv14 = zext i8 %tmp13 to i32
+  %cmp15 = icmp eq i32 %conv14, 76
+  br i1 %cmp15, label %land.lhs.true17, label %if.end
+
+land.lhs.true17:                                  ; preds = %land.lhs.true9
+  %arrayidx20 = getelementptr inbounds i8* %arrayidx, i64 3
+  %tmp21 = load i8* %arrayidx20, align 1
+  %conv22 = zext i8 %tmp21 to i32
+  %cmp23 = icmp eq i32 %conv22, 70
+  br i1 %cmp23, label %if.then, label %if.end
+
+if.then:                                          ; preds = %land.lhs.true17
+  %call25 = call i32 (...)* @doo()
+  br label %if.end
+
+if.end:
+  ret void
+
+; CHECK: @test2a
+; CHECK: %x1 = load i32* {{.*}}, align 4
+; CHECK-NEXT: icmp eq i32 %x1, 1179403647
+; CHECK-NEXT: br i1 {{.*}}, label %if.then, label %if.end 
+}
+