[WIP][IR][Constants] Use ptr addrspace(N) zeroinitializer to represent a zero-value pointer

Author: shiltian

llvm/include/llvm/IR/Constant.h

@@ -189,6 +189,8 @@ class Constant : public User {
 
   LLVM_ABI static Constant *getNullValue(Type *Ty);
 
+  LLVM_ABI static Constant *getZeroValue(Type *Ty);
+
   /// @returns the value for an integer or vector of integer constant of the
   /// given type that has all its bits set to true.
   /// Get the all ones value

llvm/lib/AsmParser/LLParser.cpp

@@ -6575,7 +6575,7 @@ bool LLParser::convertValIDToValue(Type *Ty, ValID &ID, Value *&V,
     if (auto *TETy = dyn_cast<TargetExtType>(Ty))
       if (!TETy->hasProperty(TargetExtType::HasZeroInit))
         return error(ID.Loc, "invalid type for null constant");
-    V = Constant::getNullValue(Ty);
+    V = Constant::getZeroValue(Ty);
     return false;
   case ValID::t_None:
     if (!Ty->isTokenTy())

llvm/lib/IR/AsmWriter.cpp

@@ -1806,6 +1806,15 @@ static void writeConstantInternal(raw_ostream &Out, const Constant *CV,
       }
     }
 
+    // Use zeroinitializer for inttoptr(0) constant expression.
+    if (CE->getOpcode() == Instruction::IntToPtr) {
+      Constant *SrcCI = cast<Constant>(CE->getOperand(0));
+      if (SrcCI->isNullValue()) {
+        Out << "zeroinitializer";
+        return;
+      }
+    }
+
     Out << CE->getOpcodeName();
     writeOptimizationInfo(Out, CE);
     Out << " (";

llvm/lib/IR/ConstantFold.cpp

@@ -126,6 +126,16 @@ Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V,
   if (isa<PoisonValue>(V))
     return PoisonValue::get(DestTy);
 
+  // TODO: For the following two cases, we can fold the constant to a constant
+  // according to the data layout.
+
+  // We can't fold inttoptr(0) to ConstantNullPointer.
+  if (opc == Instruction::IntToPtr && V->isZeroValue())
+    return nullptr;
+  // We can't fold ptrtoint(nullptr) to null.
+  if (opc == Instruction::PtrToInt && V->isNullValue())
+    return nullptr;
+
   if (isa<UndefValue>(V)) {
     // zext(undef) = 0, because the top bits will be zero.
     // sext(undef) = 0, because the top bits will all be the same.

llvm/lib/IR/Constants.cpp

@@ -72,12 +72,21 @@ bool Constant::isNegativeZeroValue() const {
 }
 
 // Return true iff this constant is positive zero (floating point), negative
-// zero (floating point), or a null value.
+// zero (floating point), zero-value pointer, or a null value.
 bool Constant::isZeroValue() const {
   // Floating point values have an explicit -0.0 value.
   if (const ConstantFP *CFP = dyn_cast<ConstantFP>(this))
     return CFP->isZero();
 
+  // Zero value pointer is a constant expression of inttoptr(0).
+  if (const auto *CE = dyn_cast<ConstantExpr>(this)) {
+    if (CE->getOpcode() == Instruction::IntToPtr) {
+      Constant *SrcCI = cast<Constant>(CE->getOperand(0));
+      if (SrcCI->isZeroValue())
+        return true;
+    }
+  }
+
   // Check for constant splat vectors of 1 values.
   if (getType()->isVectorTy())
     if (const auto *SplatCFP = dyn_cast_or_null<ConstantFP>(getSplatValue()))
@@ -369,7 +378,10 @@ bool Constant::containsConstantExpression() const {
   return false;
 }
 
-/// Constructor to create a '0' constant of arbitrary type.
+/// Constructor that creates a null constant of any type. For most types, this
+/// means a constant with value '0', but for pointer types, it represents a
+/// nullptr constant. A nullptr isn't always a zero-value pointer in certain
+/// address spaces on some targets.
 Constant *Constant::getNullValue(Type *Ty) {
   switch (Ty->getTypeID()) {
   case Type::IntegerTyID:
@@ -400,6 +412,19 @@ Constant *Constant::getNullValue(Type *Ty) {
   }
 }
 
+/// Constructor that creates a zero constant of any type. For most types, this
+/// is equivalent to getNullValue. For pointer types, it creates an inttoptr
+/// constant expression.
+Constant *Constant::getZeroValue(Type *Ty) {
+  switch (Ty->getTypeID()) {
+  case Type::PointerTyID:
+    return ConstantExpr::getIntToPtr(
+        ConstantInt::get(Type::getInt8Ty(Ty->getContext()), 0), Ty);
+  default:
+    return Constant::getNullValue(Ty);
+  }
+}
+
 Constant *Constant::getIntegerValue(Type *Ty, const APInt &V) {
   Type *ScalarTy = Ty->getScalarType();
 
@@ -735,7 +760,7 @@ static bool constantIsDead(const Constant *C, bool RemoveDeadUsers) {
     ReplaceableMetadataImpl::SalvageDebugInfo(*C);
     const_cast<Constant *>(C)->destroyConstant();
   }
-  
+
   return true;
 }
 

llvm/test/CodeGen/AMDGPU/lower-buffer-fat-pointers-constants.ll

@@ -193,23 +193,23 @@ define i32 @ptrtoint_very_short() {
 define <2 x i160> @ptrtoint_vec() {
 ; CHECK-LABEL: define <2 x i160> @ptrtoint_vec
 ; CHECK-SAME: () #[[ATTR0]] {
-; CHECK-NEXT:    ret <2 x i160> zeroinitializer
+; CHECK-NEXT:    ret <2 x i160> ptrtoint (<2 x ptr addrspace(7)> zeroinitializer to <2 x i160>)
 ;
   ret <2 x i160> ptrtoint (<2 x ptr addrspace(7)> zeroinitializer to <2 x i160>)
 }
 
 define ptr addrspace(7) @inttoptr() {
 ; CHECK-LABEL: define { ptr addrspace(8), i32 } @inttoptr
 ; CHECK-SAME: () #[[ATTR0]] {
-; CHECK-NEXT:    ret { ptr addrspace(8), i32 } zeroinitializer
+; CHECK-NEXT:    ret { ptr addrspace(8), i32 } { ptr addrspace(8) zeroinitializer, i32 0 }
 ;
   ret ptr addrspace(7) inttoptr (i160 0 to ptr addrspace(7))
 }
 
 define <2 x ptr addrspace(7)> @inttoptr_vec() {
 ; CHECK-LABEL: define { <2 x ptr addrspace(8)>, <2 x i32> } @inttoptr_vec
 ; CHECK-SAME: () #[[ATTR0]] {
-; CHECK-NEXT:    ret { <2 x ptr addrspace(8)>, <2 x i32> } { <2 x ptr addrspace(8)> zeroinitializer, <2 x i32> <i32 1, i32 2> }
+; CHECK-NEXT:    ret { <2 x ptr addrspace(8)>, <2 x i32> } { <2 x ptr addrspace(8)> <ptr addrspace(8) zeroinitializer, ptr addrspace(8) zeroinitializer>, <2 x i32> <i32 1, i32 2> }
 ;
   ret <2 x ptr addrspace(7)> inttoptr (<2 x i160> <i160 1, i160 2> to <2 x ptr addrspace(7)>)
 }

llvm/test/CodeGen/AMDGPU/remove-incompatible-s-time.ll

@@ -33,11 +33,11 @@
 ]
 
 ; REALTIME:    @ConstantExpr0 = internal global i64 ptrtoint (ptr @needs_s_memrealtime to i64)
-; NOREALTIME:  @ConstantExpr0 = internal global i64 0
+; NOREALTIME:  @ConstantExpr0 = internal global i64 ptrtoint (ptr null to i64)
 @ConstantExpr0 = internal global i64 ptrtoint (ptr @needs_s_memrealtime to i64)
 
 ; MEMTIME:     @ConstantExpr1 = internal global i64 ptrtoint (ptr @needs_s_memtime to i64)
-; NOMEMTIME:   @ConstantExpr1 = internal global i64 0
+; NOMEMTIME:   @ConstantExpr1 = internal global i64 ptrtoint (ptr null to i64)
 @ConstantExpr1 = internal global i64 ptrtoint (ptr @needs_s_memtime to i64)
 
 ; REALTIME:         define i64 @needs_s_memrealtime

llvm/test/Instrumentation/AddressSanitizer/asan-scalable-vector.ll

@@ -7,13 +7,16 @@ define void @test() #1 {
 ; CHECK-NEXT:  [[ENTRY:.*:]]
 ; CHECK-NEXT:    [[CTX_PG:%.*]] = alloca <vscale x 16 x i1>, align 2
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0(ptr [[CTX_PG]])
-; CHECK-NEXT:    [[TMP0:%.*]] = load i8, ptr inttoptr (i64 17592186044416 to ptr), align 1
+; CHECK-NEXT:    [[TMP3:%.*]] = lshr i64 ptrtoint (ptr null to i64), 3
+; CHECK-NEXT:    [[TMP4:%.*]] = or i64 [[TMP3]], 17592186044416
+; CHECK-NEXT:    [[TMP2:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT:    [[TMP0:%.*]] = load i8, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne i8 [[TMP0]], 0
-; CHECK-NEXT:    br i1 [[TMP1]], label %[[BB2:.*]], label %[[BB3:.*]]
-; CHECK:       [[BB2]]:
-; CHECK-NEXT:    call void @__asan_report_store8(i64 0) #[[ATTR4:[0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP1]], label %[[BB5:.*]], label %[[BB6:.*]]
+; CHECK:       [[BB5]]:
+; CHECK-NEXT:    call void @__asan_report_store8(i64 ptrtoint (ptr null to i64)) #[[ATTR4:[0-9]+]]
 ; CHECK-NEXT:    unreachable
-; CHECK:       [[BB3]]:
+; CHECK:       [[BB6]]:
 ; CHECK-NEXT:    store ptr [[CTX_PG]], ptr null, align 8
 ; CHECK-NEXT:    ret void
 ;

llvm/test/Instrumentation/HWAddressSanitizer/RISCV/alloca.ll

@@ -70,7 +70,7 @@ define void @test_alloca() sanitize_hwaddress !dbg !15 {
 ; ZERO-BASED-SHADOW-LABEL: define void @test_alloca
 ; ZERO-BASED-SHADOW-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__hwasan_personality_thunk !dbg [[DBG8:![0-9]+]] {
 ; ZERO-BASED-SHADOW-NEXT:  entry:
-; ZERO-BASED-SHADOW-NEXT:    [[DOTHWASAN_SHADOW:%.*]] = call ptr asm "", "=r,0"(ptr null)
+; ZERO-BASED-SHADOW-NEXT:    [[DOTHWASAN_SHADOW:%.*]] = call ptr asm "", "=r,0"(ptr zeroinitializer)
 ; ZERO-BASED-SHADOW-NEXT:    [[TMP0:%.*]] = call ptr @llvm.frameaddress.p0(i32 0)
 ; ZERO-BASED-SHADOW-NEXT:    [[TMP1:%.*]] = ptrtoint ptr [[TMP0]] to i64
 ; ZERO-BASED-SHADOW-NEXT:    [[TMP2:%.*]] = lshr i64 [[TMP1]], 20
@@ -143,7 +143,7 @@ declare void @llvm.dbg.value(metadata, metadata, metadata)
 ;.
 ; DYNAMIC-SHADOW: [[META0]] = !{ptr @hwasan.note}
 ; DYNAMIC-SHADOW: [[META1:![0-9]+]] = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: [[META2:![0-9]+]], producer: "{{.*}}clang version {{.*}}", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: [[META3:![0-9]+]], splitDebugInlining: false, nameTableKind: None)
-; DYNAMIC-SHADOW: [[META2]] = !DIFile(filename: "alloca.cpp", directory: {{.*}})
+; DYNAMIC-SHADOW: [[META2]] = !DIFile(filename: "{{.*}}alloca.cpp", directory: {{.*}})
 ; DYNAMIC-SHADOW: [[META3]] = !{}
 ; DYNAMIC-SHADOW: [[META4:![0-9]+]] = !{i32 7, !"Dwarf Version", i32 4}
 ; DYNAMIC-SHADOW: [[META5:![0-9]+]] = !{i32 2, !"Debug Info Version", i32 3}
@@ -160,7 +160,7 @@ declare void @llvm.dbg.value(metadata, metadata, metadata)
 ;.
 ; ZERO-BASED-SHADOW: [[META0]] = !{ptr @hwasan.note}
 ; ZERO-BASED-SHADOW: [[META1:![0-9]+]] = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: [[META2:![0-9]+]], producer: "{{.*}}clang version {{.*}}", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: [[META3:![0-9]+]], splitDebugInlining: false, nameTableKind: None)
-; ZERO-BASED-SHADOW: [[META2]] = !DIFile(filename: "alloca.cpp", directory: {{.*}})
+; ZERO-BASED-SHADOW: [[META2]] = !DIFile(filename: "{{.*}}alloca.cpp", directory: {{.*}})
 ; ZERO-BASED-SHADOW: [[META3]] = !{}
 ; ZERO-BASED-SHADOW: [[META4:![0-9]+]] = !{i32 7, !"Dwarf Version", i32 4}
 ; ZERO-BASED-SHADOW: [[META5:![0-9]+]] = !{i32 2, !"Debug Info Version", i32 3}