[TBAA] Do not rewrite TBAA if exists, always null out !tbaa.struct

Retrieve `!tbaa` metadata via `!tbaa.struct` in `adjustForAccess`
unless it already exists, as struct-path aware `MDNodes` emitted
via `new-struct-path-tbaa` may be leveraged. As `!tbaa.struct`
carries memcpy padding semantics among struct fields and `!tbaa`
is already meant to aid to alias semantics, it should be possible
to zero out `!tbaa.struct` once the memcpy has been simplified.
`SROA/tbaa-struct.ll` test has gone out of scope, as `!tbaa` has
already replaced `!tbaa.struct` in SROA.

Fixes: llvm#95661.

Author: antoniofrighetto

llvm/include/llvm/IR/Metadata.h

@@ -846,8 +846,10 @@ struct AAMDNodes {
   AAMDNodes concat(const AAMDNodes &Other) const;
 
   /// Create a new AAMDNode for accessing \p AccessSize bytes of this AAMDNode.
-  /// If his AAMDNode has !tbaa.struct and \p AccessSize matches the size of the
-  /// field at offset 0, get the TBAA tag describing the accessed field.
+  /// If this AAMDNode has !tbaa.struct and \p AccessSize matches the size of
+  /// the field at offset 0, get the TBAA tag describing the accessed field.
+  /// If such an AAMDNode already embeds !tbaa, the existing one is retrieved.
+  /// Finally, !tbaa.struct is zeroed out.
   AAMDNodes adjustForAccess(unsigned AccessSize);
   AAMDNodes adjustForAccess(size_t Offset, Type *AccessTy,
                             const DataLayout &DL);

llvm/lib/Analysis/TypeBasedAliasAnalysis.cpp

@@ -822,16 +822,16 @@ MDNode *AAMDNodes::extendToTBAA(MDNode *MD, ssize_t Len) {
 AAMDNodes AAMDNodes::adjustForAccess(unsigned AccessSize) {
   AAMDNodes New = *this;
   MDNode *M = New.TBAAStruct;
-  if (M && M->getNumOperands() >= 3 && M->getOperand(0) &&
+  if (!New.TBAA && M && M->getNumOperands() >= 3 && M->getOperand(0) &&
       mdconst::hasa<ConstantInt>(M->getOperand(0)) &&
       mdconst::extract<ConstantInt>(M->getOperand(0))->isZero() &&
       M->getOperand(1) && mdconst::hasa<ConstantInt>(M->getOperand(1)) &&
       mdconst::extract<ConstantInt>(M->getOperand(1))->getValue() ==
           AccessSize &&
-      M->getOperand(2) && isa<MDNode>(M->getOperand(2))) {
-    New.TBAAStruct = nullptr;
+      M->getOperand(2) && isa<MDNode>(M->getOperand(2)))
     New.TBAA = cast<MDNode>(M->getOperand(2));
-  }
+
+  New.TBAAStruct = nullptr;
   return New;
 }
 

llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp

@@ -171,7 +171,7 @@ Instruction *InstCombinerImpl::SimplifyAnyMemTransfer(AnyMemTransferInst *MI) {
   IntegerType* IntType = IntegerType::get(MI->getContext(), Size<<3);
 
   // If the memcpy has metadata describing the members, see if we can get the
-  // TBAA tag describing our copy.
+  // TBAA, scope and noalias tags describing our copy.
   AAMDNodes AACopyMD = MI->getAAMetadata().adjustForAccess(Size);
 
   Value *Src = MI->getArgOperand(1);

llvm/test/Transforms/InstCombine/struct-assign-tbaa-2.ll

@@ -0,0 +1,48 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals
+; RUN: opt -passes=instcombine -S < %s | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+
+declare void @llvm.memcpy.p0.p0.i64(ptr nocapture, ptr nocapture, i64, i1) nounwind
+
+%struct.T = type { %struct.Wrapper, %struct.Wrapper }
+%struct.Wrapper = type { i16 }
+
+define void @test1(ptr %a1, ptr %a2) {
+; CHECK-LABEL: @test1(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[B:%.*]] = getelementptr inbounds i8, ptr [[A2:%.*]], i64 2
+; CHECK-NEXT:    [[TMP0:%.*]] = load i16, ptr [[A2]], align 2, !tbaa [[TBAA0:![0-9]+]]
+; CHECK-NEXT:    store i16 [[TMP0]], ptr [[A1:%.*]], align 2, !tbaa [[TBAA0]]
+; CHECK-NEXT:    [[B2:%.*]] = getelementptr inbounds i8, ptr [[A1]], i64 2
+; CHECK-NEXT:    [[TMP1:%.*]] = load i16, ptr [[B]], align 2, !tbaa [[TBAA6:![0-9]+]]
+; CHECK-NEXT:    store i16 [[TMP1]], ptr [[B2]], align 2, !tbaa [[TBAA6]]
+; CHECK-NEXT:    ret void
+;
+entry:
+  %b = getelementptr inbounds i8, ptr %a2, i64 2
+  call void @llvm.memcpy.p0.p0.i64(ptr align 2 %a1, ptr align 2 %a2, i64 2, i1 false), !tbaa !0, !tbaa.struct !6
+  %b2 = getelementptr inbounds %struct.T, ptr %a1, i32 0, i32 1
+  call void @llvm.memcpy.p0.p0.i64(ptr align 2 %b2, ptr align 2 %b, i64 2, i1 false), !tbaa !8, !tbaa.struct !6
+  ret void
+}
+
+!0 = !{!1, !4, i64 0, i64 2}
+!1 = !{!2, i64 4, !"_ZTS1T", !4, i64 0, i64 2, !4, i64 2, i64 2}
+!2 = !{!3, i64 1, !"omnipotent char"}
+!3 = !{!"Simple C++ TBAA"}
+!4 = !{!2, i64 2, !"_ZTS7Wrapper", !5, i64 0, i64 2}
+!5 = !{!2, i64 2, !"short"}
+!6 = !{i64 0, i64 2, !7}
+!7 = !{!5, !5, i64 0, i64 2}
+!8 = !{!1, !4, i64 2, i64 2}
+
+;.
+; CHECK: [[TBAA0]] = !{[[META1:![0-9]+]], [[META4:![0-9]+]], i64 0, i64 2}
+; CHECK: [[META1]] = !{[[META2:![0-9]+]], i64 4, !"_ZTS1T", [[META4]], i64 0, i64 2, [[META4]], i64 2, i64 2}
+; CHECK: [[META2]] = !{[[META3:![0-9]+]], i64 1, !"omnipotent char"}
+; CHECK: [[META3]] = !{!"Simple C++ TBAA"}
+; CHECK: [[META4]] = !{[[META2]], i64 2, !"_ZTS7Wrapper", [[META5:![0-9]+]], i64 0, i64 2}
+; CHECK: [[META5]] = !{[[META2]], i64 2, !"short"}
+; CHECK: [[TBAA6]] = !{[[META1]], [[META4]], i64 2, i64 2}
+;.

llvm/test/Transforms/InstCombine/struct-assign-tbaa.ll

@@ -38,8 +38,8 @@ define ptr @test2() {
 define void @test3_multiple_fields(ptr nocapture %a, ptr nocapture %b) {
 ; CHECK-LABEL: @test3_multiple_fields(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr [[B:%.*]], align 4, !tbaa.struct [[TBAA_STRUCT3:![0-9]+]]
-; CHECK-NEXT:    store i64 [[TMP0]], ptr [[A:%.*]], align 4, !tbaa.struct [[TBAA_STRUCT3]]
+; CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr [[B:%.*]], align 4
+; CHECK-NEXT:    store i64 [[TMP0]], ptr [[A:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -50,8 +50,8 @@ entry:
 define void @test4_multiple_copy_first_field(ptr nocapture %a, ptr nocapture %b) {
 ; CHECK-LABEL: @test4_multiple_copy_first_field(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[B:%.*]], align 4
-; CHECK-NEXT:    store i32 [[TMP0]], ptr [[A:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[B:%.*]], align 4, !tbaa [[TBAA0]]
+; CHECK-NEXT:    store i32 [[TMP0]], ptr [[A:%.*]], align 4, !tbaa [[TBAA0]]
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -86,5 +86,4 @@ entry:
 ; CHECK: [[TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0}
 ; CHECK: [[META1]] = !{!"float", [[META2:![0-9]+]]}
 ; CHECK: [[META2]] = !{!"Simple C/C++ TBAA"}
-; CHECK: [[TBAA_STRUCT3]] = !{i64 0, i64 4, [[TBAA0]], i64 4, i64 4, [[TBAA0]]}
 ;.