Prevent Constant Folding From Optimizing inrange GEP

This patch does the following things:

1. update SymbolicallyEvaluateGEP so that it bails out if it cannot preserve inrange arribute;
2. update llvm/test/Analysis/ConstantFolding/gep.ll to remove UB in it;
3. remove inaccurate comment above ConstantFoldInstOperandsImpl in llvm/lib/Analysis/ConstantFolding.cpp;
4. add a new regression test that makes sure that no optimizations change an inrange GEP in an unexpected way.

Patch by Zhaomo Yang!

Differential Revision: https://reviews.llvm.org/D51698

llvm-svn: 341888

diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index bbd812b..c73250a 100644 (file)
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -960,10 +960,8 @@ Constant *SymbolicallyEvaluateGEP(const GEPOperator *GEP,
         NewIdxs.size() > *LastIRIndex) {
       InRangeIndex = LastIRIndex;
       for (unsigned I = 0; I <= *LastIRIndex; ++I)
-        if (NewIdxs[I] != InnermostGEP->getOperand(I + 1)) {
-          InRangeIndex = None;
-          break;
-        }
+        if (NewIdxs[I] != InnermostGEP->getOperand(I + 1))
+          return nullptr;
     }
 
   // Create a GEP.
@@ -985,11 +983,6 @@ Constant *SymbolicallyEvaluateGEP(const GEPOperator *GEP,
 /// returned, if not, null is returned.  Note that this function can fail when
 /// attempting to fold instructions like loads and stores, which have no
 /// constant expression form.
-///
-/// TODO: This function neither utilizes nor preserves nsw/nuw/inbounds/inrange
-/// etc information, due to only being passed an opcode and operands. Constant
-/// folding using this function strips this information.
-///
 Constant *ConstantFoldInstOperandsImpl(const Value *InstOrCE, unsigned Opcode,
                                        ArrayRef<Constant *> Ops,
                                        const DataLayout &DL,

diff --git a/llvm/test/Analysis/ConstantFolding/gep.ll b/llvm/test/Analysis/ConstantFolding/gep.ll
index caa7f15..2596186 100644 (file)
--- a/llvm/test/Analysis/ConstantFolding/gep.ll
+++ b/llvm/test/Analysis/ConstantFolding/gep.ll
@@ -8,23 +8,20 @@ target triple = "x86_64-unknown-linux-gnu"
 
 @vt = external global [3 x i8*]
 
-; CHECK: define i32 (...)* @f0()
-define i32 (...)* @f0() {
-  ; CHECK-NEXT: load i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, inrange i64 0, i64 2) to i32 (...)**)
-  %load = load i32 (...)*, i32 (...)** getelementptr (i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, inrange i64 0, i64 1) to i32 (...)**), i64 1)
-  ret i32 (...)* %load
+; CHECK: define i32 (...)** @f0()
+define i32 (...)** @f0() {
+  ; CHECK-NEXT: ret i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, inrange i64 0, i64 2) to i32 (...)**
+  ret i32 (...)** getelementptr (i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, inrange i64 0, i64 1) to i32 (...)**), i64 1)
 }
 
-; CHECK: define i32 (...)* @f1()
-define i32 (...)* @f1() {
-  ; CHECK-NEXT: load i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, i64 0, i64 2) to i32 (...)**)
-  %load = load i32 (...)*, i32 (...)** getelementptr (i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, i64 0, inrange i64 1) to i32 (...)**), i64 1)
-  ret i32 (...)* %load
+; CHECK: define i32 (...)** @f1()
+define i32 (...)** @f1() {
+  ; CHECK-NEXT: ret i32 (...)** getelementptr (i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, i64 0, inrange i64 1) to i32 (...)**), i64 1)
+  ret i32 (...)** getelementptr (i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, i64 0, inrange i64 1) to i32 (...)**), i64 1)
 }
 
-; CHECK: define i32 (...)* @f2()
-define i32 (...)* @f2() {
-  ; CHECK-NEXT: load i32 (...)*, i32 (...)** bitcast (i8** getelementptr ([3 x i8*], [3 x i8*]* @vt, i64 1, i64 1) to i32 (...)**)
-  %load = load i32 (...)*, i32 (...)** getelementptr (i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, i64 0, inrange i64 1) to i32 (...)**), i64 3)
-  ret i32 (...)* %load
+; CHECK: define i32 (...)** @f2()
+define i32 (...)** @f2() {
+  ; CHECK-NEXT: ret i32 (...)** getelementptr (i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, i64 0, inrange i64 1) to i32 (...)**), i64 3)
+  ret i32 (...)** getelementptr (i32 (...)*, i32 (...)** bitcast (i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @vt, i64 0, inrange i64 1) to i32 (...)**), i64 3)
 }

diff --git a/llvm/test/Other/optimize-inrange-gep.ll b/llvm/test/Other/optimize-inrange-gep.ll
new file mode 100644 (file)
index 0000000..9e86889
--- /dev/null
+++ b/llvm/test/Other/optimize-inrange-gep.ll
@@ -0,0 +1,18 @@
+; RUN: opt -O0 -S < %s | FileCheck %s
+; RUN: opt -O1 -S < %s | FileCheck %s
+; RUN: opt -O2 -S < %s | FileCheck %s
+; RUN: opt -O3 -S < %s | FileCheck %s
+; RUN: opt -Os -S < %s | FileCheck %s
+; RUN: opt -Oz -S < %s | FileCheck %s
+
+target datalayout = "e-p:64:64"
+
+; Make sure that optimizations do not optimize inrange GEP.
+
+@vtable = constant { [3 x i8*] } { [3 x i8*] [i8* null, i8* null, i8* null] }
+
+define void @foo(i8*** %p) {
+  ;CHECK: store i8** getelementptr {{.*}} ({ [3 x i8*] }, { [3 x i8*] }* @vtable, i{{.*}} 0, inrange i32 0, i{{.*}} 3), i8*** %p
+  store i8** getelementptr ({ [3 x i8*] }, { [3 x i8*] }* @vtable, i32 0, inrange i32 0, i32 3), i8*** %p
+  ret void
+}