implement the first part of PR8882: when lowering an inbounds

gep to explicit addressing, we know that none of the intermediate
computation overflows.

This could use review: it seems that the shifts certainly wouldn't
overflow, but could the intermediate adds overflow if there is a 
negative index?

Previously the testcase would instcombine to:

define i1 @test(i64 %i) {
  %p1.idx.mask = and i64 %i, 4611686018427387903
  %cmp = icmp eq i64 %p1.idx.mask, 1000
  ret i1 %cmp
}

now we get:

define i1 @test(i64 %i) {
  %cmp = icmp eq i64 %i, 1000
  ret i1 %cmp
}



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125271 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 33 insertions, 10 deletions

diff --git a/lib/Transforms/InstCombine/InstCombineAddSub.cpp b/lib/Transforms/InstCombine/InstCombineAddSub.cpp
index 91116ca..b04a05b 100644
--- a/lib/Transforms/InstCombine/InstCombineAddSub.cpp
+++ b/lib/Transforms/InstCombine/InstCombineAddSub.cpp
@@ -399,6 +399,10 @@ Value *InstCombiner::EmitGEPOffset(User *GEP) {
   const Type *IntPtrTy = TD.getIntPtrType(GEP->getContext());
   Value *Result = Constant::getNullValue(IntPtrTy);
 
+  // If the GEP is inbounds, we know that none of the addressing operations will
+  // overflow in an unsigned sense.
+  bool isInBounds = cast<GEPOperator>(GEP)->isInBounds();
+  
   // Build a mask for high order bits.
   unsigned IntPtrWidth = TD.getPointerSizeInBits();
   uint64_t PtrSizeMask = ~0ULL >> (64-IntPtrWidth);
@@ -414,31 +418,34 @@ Value *InstCombiner::EmitGEPOffset(User *GEP) {
       if (const StructType *STy = dyn_cast<StructType>(*GTI)) {
         Size = TD.getStructLayout(STy)->getElementOffset(OpC->getZExtValue());
         
-        Result = Builder->CreateAdd(Result,
-                                    ConstantInt::get(IntPtrTy, Size),
-                                    GEP->getName()+".offs");
+        if (Size)
+          Result = Builder->CreateAdd(Result, ConstantInt::get(IntPtrTy, Size),
+                                      GEP->getName()+".offs",
+                                      isInBounds /*NUW*/);
         continue;
       }
       
       Constant *Scale = ConstantInt::get(IntPtrTy, Size);
       Constant *OC =
               ConstantExpr::getIntegerCast(OpC, IntPtrTy, true /*SExt*/);
-      Scale = ConstantExpr::getMul(OC, Scale);
+      Scale = ConstantExpr::getMul(OC, Scale, isInBounds/*NUW*/);
       // Emit an add instruction.
-      Result = Builder->CreateAdd(Result, Scale, GEP->getName()+".offs");
+      Result = Builder->CreateAdd(Result, Scale, GEP->getName()+".offs",
+                                  isInBounds /*NUW*/);
       continue;
     }
     // Convert to correct type.
     if (Op->getType() != IntPtrTy)
       Op = Builder->CreateIntCast(Op, IntPtrTy, true, Op->getName()+".c");
     if (Size != 1) {
-      Constant *Scale = ConstantInt::get(IntPtrTy, Size);
       // We'll let instcombine(mul) convert this to a shl if possible.
-      Op = Builder->CreateMul(Op, Scale, GEP->getName()+".idx");
+      Op = Builder->CreateMul(Op, ConstantInt::get(IntPtrTy, Size),
+                              GEP->getName()+".idx", isInBounds /*NUW*/);
     }
 
     // Emit an add instruction.
-    Result = Builder->CreateAdd(Op, Result, GEP->getName()+".offs");
+    Result = Builder->CreateAdd(Op, Result, GEP->getName()+".offs",
+                                isInBounds /*NUW*/);
   }
   return Result;
 }
diff --git a/test/Transforms/InstCombine/icmp.ll b/test/Transforms/InstCombine/icmp.ll
index baff2ca..72ada24 100644
--- a/test/Transforms/InstCombine/icmp.ll
+++ b/test/Transforms/InstCombine/icmp.ll
@@ -1,5 +1,8 @@
 ; RUN: opt < %s -instcombine -S | FileCheck %s
 
+target datalayout =
+"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+
 define i32 @test1(i32 %X) {
 entry:
         icmp slt i32 %X, 0              ; <i1>:0 [#uses=1]
@@ -218,3 +221,16 @@ define i1 @test23(i32 %x) nounwind {
 	%i4 = icmp eq i32 %i3, -1
 	ret i1 %i4
 }
+
+@X = global [1000 x i32] zeroinitializer
+
+; PR8882
+; CHECK: @test24
+; CHECK:    %cmp = icmp eq i64 %i, 1000
+; CHECK:   ret i1 %cmp
+define i1 @test24(i64 %i) {
+  %p1 = getelementptr inbounds i32* getelementptr inbounds ([1000 x i32]* @X, i64 0, i64 0), i64 %i
+  %cmp = icmp eq i32* %p1, getelementptr inbounds ([1000 x i32]* @X, i64 1, i64 0)
+  ret i1 %cmp
+}
+
diff --git a/test/Transforms/InstCombine/sub.ll b/test/Transforms/InstCombine/sub.ll
index 29d96a4..778a1b9 100644
--- a/test/Transforms/InstCombine/sub.ll
+++ b/test/Transforms/InstCombine/sub.ll
@@ -256,7 +256,7 @@ define i64 @test24b(i8* %P, i64 %A){
   %G = sub i64 %C, ptrtoint ([42 x i16]* @Arr to i64)
   ret i64 %G
 ; CHECK: @test24b
-; CHECK-NEXT: shl i64 %A, 1
+; CHECK-NEXT: shl nuw i64 %A, 1
 ; CHECK-NEXT: ret i64 
 }
 
@@ -267,7 +267,7 @@ define i64 @test25(i8* %P, i64 %A){
   %G = sub i64 %C, ptrtoint (i16* getelementptr ([42 x i16]* @Arr, i64 1, i64 0) to i64)
   ret i64 %G
 ; CHECK: @test25
-; CHECK-NEXT: shl i64 %A, 1
+; CHECK-NEXT: shl nuw i64 %A, 1
 ; CHECK-NEXT: add i64 {{.*}}, -84
 ; CHECK-NEXT: ret i64 
 }