[AMDGPUAttributor][FIX] No endless recursion for recursive initializers

Fixes: #63956
llvm · Jul 19, 2023 · d015018 · d015018
1 parent c3f3068
commit d015018
Show file tree

Hide file tree

Showing 2 changed files with 29 additions and 4 deletions.
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
@@ -208,7 +208,8 @@ class AMDGPUInformationCache : public InformationCache {
   }
 
   /// Get the constant access bitmap for \p C.
-  uint8_t getConstantAccess(const Constant *C) {
+  uint8_t getConstantAccess(const Constant *C,
+                            SmallPtrSetImpl<const Constant *> &Visited) {
     auto It = ConstantStatus.find(C);
     if (It != ConstantStatus.end())
       return It->second;
@@ -223,10 +224,10 @@ class AMDGPUInformationCache : public InformationCache {
 
     for (const Use &U : C->operands()) {
       const auto *OpC = dyn_cast<Constant>(U);
-      if (!OpC)
+      if (!OpC || !Visited.insert(OpC).second)
         continue;
 
-      Result |= getConstantAccess(OpC);
+      Result |= getConstantAccess(OpC, Visited);
     }
     return Result;
   }
@@ -241,7 +242,8 @@ class AMDGPUInformationCache : public InformationCache {
     if (!IsNonEntryFunc && HasAperture)
       return false;
 
-    uint8_t Access = getConstantAccess(C);
+    SmallPtrSet<const Constant *, 8> Visited;
+    uint8_t Access = getConstantAccess(C, Visited);
 
     // We need to trap on DS globals in non-entry functions.
     if (IsNonEntryFunc && (Access & DS_GLOBAL))

diff --git a/llvm/test/CodeGen/AMDGPU/recursive_global_initializer.ll b/llvm/test/CodeGen/AMDGPU/recursive_global_initializer.ll
@@ -0,0 +1,23 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals --version 2
+; RUN: opt -mtriple=amdgcn-amd-amdhsa -amdgpu-attributor -S %s | FileCheck %s
+
+%struct.foo = type { %struct.pluto, ptr, i64 }
+%struct.pluto = type { [512 x i8], ptr }
+
+@global.2 = internal addrspace(1) global %struct.foo { %struct.pluto zeroinitializer, ptr addrspacecast (ptr addrspace(1) @global.2 to ptr), i64 0 }
+
+;.
+; CHECK: @[[GLOBAL_2:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(1) global [[STRUCT_FOO:%.*]] { [[STRUCT_PLUTO:%.*]] zeroinitializer, ptr addrspacecast (ptr addrspace(1) @global.2 to ptr), i64 0 }
+;.
+define void @hoge()  {
+; CHECK-LABEL: define void @hoge
+; CHECK-SAME: () #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:    [[LOAD:%.*]] = load i64, ptr addrspace(1) getelementptr inbounds ([[STRUCT_FOO:%.*]], ptr addrspace(1) @global.2, i64 0, i32 2), align 8
+; CHECK-NEXT:    ret void
+;
+  %load = load i64, ptr addrspace(1) getelementptr inbounds (%struct.foo, ptr addrspace(1) @global.2, i64 0, i32 2), align 8
+  ret void
+}
+;.
+; CHECK: attributes #[[ATTR0]] = { "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
+;.