[Transforms] Do not drop !preserve.access.index metadata

[yonghong-song]

Summary

Currently, when a GVN or CSE optimization happens,
the llvm.preserve.access.index metadata is dropped.
This caused a problem for BPF AbstructMemberOffset phase
as it relies on the metadata (debuginfo types).

This patch added proper hooks in lib/Transforms to
preserve !preserve.access.index metadata. A test
case is added to ensure metadata is preserved under CSE.

[ast]

lgtm

llvm/trunk/lib/Transforms/Utils/Local.cpp

switch (Kind) {
         K->setMetadata(Kind,
           MDNode::getMostGenericAlignmentOrDereferenceable(JMD, KMD));
         break;
       case LLVMContext::MD_dereferenceable:
       case LLVMContext::MD_dereferenceable_or_null:
         K->setMetadata(Kind,
           MDNode::getMostGenericAlignmentOrDereferenceable(JMD, KMD));
         break;
+      case LLVMContext::MD_preserve_access_index:
+        // Preserve !preserve.access.index in K.
+        break;
     }
   }
   // Set !invariant.group from J if J has it. If both instructions have it
   // then we will just pick it from J - even when they are different.
   // Also make sure that K is load or store - f.e. combining bitcast with load
   // could produce bitcast with invariant.group metadata, which is invalid.
   // FIXME: we should try to preserve both invariant.group md if they are
   // different, but right now instruction can only have one invariant.group.
   if (auto *JMD = J->getMetadata(LLVMContext::MD_invariant_group))
     if (isa<LoadInst>(K) || isa<StoreInst>(K))
       K->setMetadata(LLVMContext::MD_invariant_group, JMD);
 }
 
 void llvm::combineMetadataForCSE(Instruction *K, const Instruction *J,
                                  bool KDominatesJ) {
   unsigned KnownIDs[] = {
       LLVMContext::MD_tbaa,            LLVMContext::MD_alias_scope,
       LLVMContext::MD_noalias,         LLVMContext::MD_range,
       LLVMContext::MD_invariant_load,  LLVMContext::MD_nonnull,
       LLVMContext::MD_invariant_group, LLVMContext::MD_align,
       LLVMContext::MD_dereferenceable,
       LLVMContext::MD_dereferenceable_or_null,
-      LLVMContext::MD_access_group};
+      LLVMContext::MD_access_group,    LLVMContext::MD_preserve_access_index};
   combineMetadata(K, J, KnownIDs, KDominatesJ);
 }
 
 void llvm::copyMetadataForLoad(LoadInst &Dest, const LoadInst &Source) {
   SmallVector<std::pair<unsigned, MDNode *>, 8> MD;
   Source.getAllMetadata(MD);
   MDBuilder MDB(Dest.getContext());
   Type *NewType = Dest.getType();
void llvm::patchReplacementInstruction(Instruction *I, Value *Repl) {
   // In general, GVN unifies expressions over different control-flow
   // regions, and so we need a conservative combination of the noalias
   // scopes.
   static const unsigned KnownIDs[] = {
       LLVMContext::MD_tbaa,            LLVMContext::MD_alias_scope,
       LLVMContext::MD_noalias,         LLVMContext::MD_range,
       LLVMContext::MD_fpmath,          LLVMContext::MD_invariant_load,
       LLVMContext::MD_invariant_group, LLVMContext::MD_nonnull,
-      LLVMContext::MD_access_group};
+      LLVMContext::MD_access_group,    LLVMContext::MD_preserve_access_index};
   combineMetadata(ReplInst, I, KnownIDs, false);
 }
 
 template <typename RootType, typename DominatesFn>
 static unsigned replaceDominatedUsesWith(Value *From, Value *To,
                                          const RootType &Root,
                                          const DominatesFn &Dominates) {
   assert(From->getType() == To->getType());

llvm/trunk/lib/Transforms/Utils/SimplifyCFG.cpp

if (isa<DbgInfoIntrinsic>(I1) || isa<DbgInfoIntrinsic>(I2)) {
                              LLVMContext::MD_fpmath,
                              LLVMContext::MD_invariant_load,
                              LLVMContext::MD_nonnull,
                              LLVMContext::MD_invariant_group,
                              LLVMContext::MD_align,
                              LLVMContext::MD_dereferenceable,
                              LLVMContext::MD_dereferenceable_or_null,
                              LLVMContext::MD_mem_parallel_loop_access,
-                             LLVMContext::MD_access_group};
+                             LLVMContext::MD_access_group,
+                             LLVMContext::MD_preserve_access_index};
       combineMetadata(I1, I2, KnownIDs, true);
 
       // I1 and I2 are being combined into a single instruction.  Its debug
       // location is the merged locations of the original instructions.
       I1->applyMergedLocation(I1->getDebugLoc(), I2->getDebugLoc());
 
       I2->eraseFromParent();
       Changed = true;

llvm/trunk/test/CodeGen/BPF/CORE/intrinsic-transforms.ll

+; RUN: opt -O2 -march=bpfeb < %s | llvm-dis | FileCheck %s
+; RUN: opt -O2 -march=bpfel < %s | llvm-dis | FileCheck %s
+;
+; Source code:
+;   #define _(x) (__builtin_preserve_access_index(x))
+;   int get_value(const int *arg);
+;   int test(int b, int *arg) {
+;     int v1 = b ? get_value(_(&arg[4])) : 0;
+;     int v2 = b ? get_value(_(&arg[4])) : 0;
+;     return v1 + v2;
+;   }
+; Compilation flag:
+;   clang -target bpf -O0 -g -S -emit-llvm -Xclang -disable-O0-optnone test.c
+
+; Function Attrs: noinline nounwind
+define dso_local i32 @test(i32 %b, i32* %arg) #0 !dbg !10 {
+entry:
+  %b.addr = alloca i32, align 4
+  %arg.addr = alloca i32*, align 8
+  %v1 = alloca i32, align 4
+  %v2 = alloca i32, align 4
+  store i32 %b, i32* %b.addr, align 4
+  call void @llvm.dbg.declare(metadata i32* %b.addr, metadata !13, metadata !DIExpression()), !dbg !14
+  store i32* %arg, i32** %arg.addr, align 8
+  call void @llvm.dbg.declare(metadata i32** %arg.addr, metadata !15, metadata !DIExpression()), !dbg !16
+  call void @llvm.dbg.declare(metadata i32* %v1, metadata !17, metadata !DIExpression()), !dbg !18
+  %0 = load i32, i32* %b.addr, align 4, !dbg !19
+  %tobool = icmp ne i32 %0, 0, !dbg !19
+  br i1 %tobool, label %cond.true, label %cond.false, !dbg !19
+
+cond.true:                                        ; preds = %entry
+  %1 = load i32*, i32** %arg.addr, align 8, !dbg !20
+  %2 = call i32* @llvm.preserve.array.access.index.p0i32.p0i32(i32* %1, i32 0, i32 4), !dbg !20, !llvm.preserve.access.index !4
+  %3 = bitcast i32* %2 to i8*, !dbg !20
+  %4 = bitcast i8* %3 to i32*, !dbg !20
+  %call = call i32 @get_value(i32* %4), !dbg !21
+  br label %cond.end, !dbg !19
+
+cond.false:                                       ; preds = %entry
+  br label %cond.end, !dbg !19
+
+cond.end:                                         ; preds = %cond.false, %cond.true
+  %cond = phi i32 [ %call, %cond.true ], [ 0, %cond.false ], !dbg !19
+  store i32 %cond, i32* %v1, align 4, !dbg !18
+  call void @llvm.dbg.declare(metadata i32* %v2, metadata !22, metadata !DIExpression()), !dbg !23
+  %5 = load i32, i32* %b.addr, align 4, !dbg !24
+  %tobool1 = icmp ne i32 %5, 0, !dbg !24
+  br i1 %tobool1, label %cond.true2, label %cond.false4, !dbg !24
+
+cond.true2:                                       ; preds = %cond.end
+  %6 = load i32*, i32** %arg.addr, align 8, !dbg !25
+  %7 = call i32* @llvm.preserve.array.access.index.p0i32.p0i32(i32* %6, i32 0, i32 4), !dbg !25, !llvm.preserve.access.index !4
+  %8 = bitcast i32* %7 to i8*, !dbg !25
+  %9 = bitcast i8* %8 to i32*, !dbg !25
+  %call3 = call i32 @get_value(i32* %9), !dbg !26
+  br label %cond.end5, !dbg !24
+
+; CHECK: tail call i32* @llvm.preserve.array.access.index.p0i32.p0i32(i32* %{{[0-9a-z]+}}, i32 0, i32 4), !dbg !{{[0-9]+}}, !llvm.preserve.access.index !{{[0-9]+}}
+; CHECK-NOT: tail call i32* @llvm.preserve.array.access.index
+
+cond.false4:                                      ; preds = %cond.end
+  br label %cond.end5, !dbg !24
+
+cond.end5:                                        ; preds = %cond.false4, %cond.true2
+  %cond6 = phi i32 [ %call3, %cond.true2 ], [ 0, %cond.false4 ], !dbg !24
+  store i32 %cond6, i32* %v2, align 4, !dbg !23
+  %10 = load i32, i32* %v1, align 4, !dbg !27
+  %11 = load i32, i32* %v2, align 4, !dbg !28
+  %add = add nsw i32 %10, %11, !dbg !29
+  ret i32 %add, !dbg !30
+}
+
+; Function Attrs: nounwind readnone speculatable willreturn
+declare void @llvm.dbg.declare(metadata, metadata, metadata) #1
+
+declare dso_local i32 @get_value(i32*) #2
+
+; Function Attrs: nounwind readnone
+declare i32* @llvm.preserve.array.access.index.p0i32.p0i32(i32*, i32 immarg, i32 immarg) #3
+
+attributes #0 = { noinline nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nounwind readnone speculatable willreturn }
+attributes #2 = { "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #3 = { nounwind readnone }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!6, !7, !8}
+!llvm.ident = !{!9}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 10.0.0 (https://github.com/llvm/llvm-project.git 4618b07fe2cede1b73512d1c260cf4981661f47f)", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !3, nameTableKind: None)
+!1 = !DIFile(filename: "test.c", directory: "/tmp/home/yhs/work/tests/llvm/cast")
+!2 = !{}
+!3 = !{!4}
+!4 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !5, size: 64)
+!5 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!6 = !{i32 2, !"Dwarf Version", i32 4}
+!7 = !{i32 2, !"Debug Info Version", i32 3}
+!8 = !{i32 1, !"wchar_size", i32 4}
+!9 = !{!"clang version 10.0.0 (https://github.com/llvm/llvm-project.git 4618b07fe2cede1b73512d1c260cf4981661f47f)"}
+!10 = distinct !DISubprogram(name: "test", scope: !1, file: !1, line: 3, type: !11, scopeLine: 3, flags: DIFlagPrototyped, isDefinition: true, unit: !0, retainedNodes: !2)
+!11 = !DISubroutineType(types: !12)
+!12 = !{!5, !5, !4}
+!13 = !DILocalVariable(name: "b", arg: 1, scope: !10, file: !1, line: 3, type: !5)
+!14 = !DILocation(line: 3, column: 14, scope: !10)
+!15 = !DILocalVariable(name: "arg", arg: 2, scope: !10, file: !1, line: 3, type: !4)
+!16 = !DILocation(line: 3, column: 22, scope: !10)
+!17 = !DILocalVariable(name: "v1", scope: !10, file: !1, line: 4, type: !5)
+!18 = !DILocation(line: 4, column: 7, scope: !10)
+!19 = !DILocation(line: 4, column: 12, scope: !10)
+!20 = !DILocation(line: 4, column: 26, scope: !10)
+!21 = !DILocation(line: 4, column: 16, scope: !10)
+!22 = !DILocalVariable(name: "v2", scope: !10, file: !1, line: 5, type: !5)
+!23 = !DILocation(line: 5, column: 7, scope: !10)
+!24 = !DILocation(line: 5, column: 12, scope: !10)
+!25 = !DILocation(line: 5, column: 26, scope: !10)
+!26 = !DILocation(line: 5, column: 16, scope: !10)
+!27 = !DILocation(line: 6, column: 10, scope: !10)
+!28 = !DILocation(line: 6, column: 15, scope: !10)
+!29 = !DILocation(line: 6, column: 13, scope: !10)
+!30 = !DILocation(line: 6, column: 3, scope: !10)