Truncate GEP indexes larger than the pointer size down to pointer size

when doing this transform if the GEP is not inbounds.  No testcase because
it is very difficult to trigger this: instcombine already canonicalizes
GEP indices to pointer size, so it relies specific permutations of the
instcombine worklist.

Thanks to Duncan for pointing this possible problem out.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@92495 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 18 insertions, 9 deletions

diff --git a/lib/Transforms/InstCombine/InstCombineCompares.cpp b/lib/Transforms/InstCombine/InstCombineCompares.cpp
index 005f1a0bc4..90ab4f43df 100644
--- a/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -197,6 +197,9 @@ static void ComputeUnsignedMinMaxValuesFromKnownBits(const APInt &KnownZero,
 Instruction *InstCombiner::
 FoldCmpLoadFromIndexedGlobal(GetElementPtrInst *GEP, GlobalVariable *GV,
                              CmpInst &ICI, ConstantInt *AndCst) {
+  // We need TD information to know the pointer size unless this is inbounds.
+  if (!GEP->isInBounds() && TD == 0) return 0;
+  
   ConstantArray *Init = dyn_cast<ConstantArray>(GV->getInitializer());
   if (Init == 0 || Init->getNumOperands() > 1024) return 0;
   
@@ -354,6 +357,12 @@ FoldCmpLoadFromIndexedGlobal(GetElementPtrInst *GEP, GlobalVariable *GV,
   // order the state machines in complexity of the generated code.
   Value *Idx = GEP->getOperand(2);
 
+  // If the index is larger than the pointer size of the target, truncate the
+  // index down like the GEP would do implicitly.  We don't have to do this for
+  // an inbounds GEP because the index can't be out of range.
+  if (!GEP->isInBounds() &&
+      Idx->getType()->getPrimitiveSizeInBits() > TD->getPointerSizeInBits())
+    Idx = Builder->CreateTrunc(Idx, TD->getIntPtrType(Idx->getContext()));
   
   // If the comparison is only true for one or two elements, emit direct
   // comparisons.
diff --git a/test/Transforms/InstCombine/load-cmp.ll b/test/Transforms/InstCombine/load-cmp.ll
index a3b68316aa..fe5df92843 100644
--- a/test/Transforms/InstCombine/load-cmp.ll
+++ b/test/Transforms/InstCombine/load-cmp.ll
@@ -6,7 +6,7 @@
    [double -10.0, double 1.0, double 4.0, double 2.0, double -20.0, double -40.0]
 
 define i1 @test1(i32 %X) {
-  %P = getelementptr [10 x i16]* @G16, i32 0, i32 %X
+  %P = getelementptr inbounds [10 x i16]* @G16, i32 0, i32 %X
   %Q = load i16* %P
   %R = icmp eq i16 %Q, 0
   ret i1 %R
@@ -16,7 +16,7 @@ define i1 @test1(i32 %X) {
 }
 
 define i1 @test2(i32 %X) {
-  %P = getelementptr [10 x i16]* @G16, i32 0, i32 %X
+  %P = getelementptr inbounds [10 x i16]* @G16, i32 0, i32 %X
   %Q = load i16* %P
   %R = icmp slt i16 %Q, 85
   ret i1 %R
@@ -26,7 +26,7 @@ define i1 @test2(i32 %X) {
 }
 
 define i1 @test3(i32 %X) {
-  %P = getelementptr [6 x double]* @GD, i32 0, i32 %X
+  %P = getelementptr inbounds [6 x double]* @GD, i32 0, i32 %X
   %Q = load double* %P
   %R = fcmp oeq double %Q, 1.0
   ret i1 %R
@@ -36,7 +36,7 @@ define i1 @test3(i32 %X) {
 }
 
 define i1 @test4(i32 %X) {
-  %P = getelementptr [10 x i16]* @G16, i32 0, i32 %X
+  %P = getelementptr inbounds [10 x i16]* @G16, i32 0, i32 %X
   %Q = load i16* %P
   %R = icmp sle i16 %Q, 73
   ret i1 %R
@@ -48,7 +48,7 @@ define i1 @test4(i32 %X) {
 }
 
 define i1 @test5(i32 %X) {
-  %P = getelementptr [10 x i16]* @G16, i32 0, i32 %X
+  %P = getelementptr inbounds [10 x i16]* @G16, i32 0, i32 %X
   %Q = load i16* %P
   %R = icmp eq i16 %Q, 69
   ret i1 %R
@@ -60,7 +60,7 @@ define i1 @test5(i32 %X) {
 }
 
 define i1 @test6(i32 %X) {
-  %P = getelementptr [6 x double]* @GD, i32 0, i32 %X
+  %P = getelementptr inbounds [6 x double]* @GD, i32 0, i32 %X
   %Q = load double* %P
   %R = fcmp ogt double %Q, 0.0
   ret i1 %R
@@ -71,7 +71,7 @@ define i1 @test6(i32 %X) {
 }
 
 define i1 @test7(i32 %X) {
-  %P = getelementptr [6 x double]* @GD, i32 0, i32 %X
+  %P = getelementptr inbounds [6 x double]* @GD, i32 0, i32 %X
   %Q = load double* %P
   %R = fcmp olt double %Q, 0.0
   ret i1 %R
@@ -82,7 +82,7 @@ define i1 @test7(i32 %X) {
 }
 
 define i1 @test8(i32 %X) {
-  %P = getelementptr [10 x i16]* @G16, i32 0, i32 %X
+  %P = getelementptr inbounds [10 x i16]* @G16, i32 0, i32 %X
   %Q = load i16* %P
   %R = and i16 %Q, 3
   %S = icmp eq i16 %R, 0
@@ -101,7 +101,7 @@ define i1 @test8(i32 %X) {
 ]
 
 define i1 @test9(i32 %X) {
-  %P = getelementptr [4 x { i32, i32 } ]* @GA, i32 0, i32 %X, i32 1
+  %P = getelementptr inbounds [4 x { i32, i32 } ]* @GA, i32 0, i32 %X, i32 1
   %Q = load i32* %P
   %R = icmp eq i32 %Q, 1
   ret i1 %R