[llvm] 57ad2e1 - [OpenMP] Prevent OpenMPOpt from internalizing uncalled functions

[via llvm-commits]

Author: Joseph Huber
Date: 2021-06-28T16:47:53-04:00
New Revision: 57ad2e106767f298a0248a5894fcb3581f533f93

URL: https://github.com/llvm/llvm-project/commit/57ad2e106767f298a0248a5894fcb3581f533f93
DIFF: https://github.com/llvm/llvm-project/commit/57ad2e106767f298a0248a5894fcb3581f533f93.diff

LOG: [OpenMP] Prevent OpenMPOpt from internalizing uncalled functions

Currently OpenMPOpt will only check if a function is a kernel before deciding not to internalize it. Any uncalled function that gets internalized will be trivially dead in the module so this is unnnecessary.

Depends on D102423

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D104890

Added: 
    

Modified: 
    llvm/lib/Transforms/IPO/OpenMPOpt.cpp
    llvm/test/Transforms/OpenMP/remove_globalization.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
index 3765378634611..ef86fe04708e3 100644
--- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -2636,15 +2636,25 @@ PreservedAnalyses OpenMPOptPass::run(Module &M, ModuleAnalysisManager &AM) {
 
   KernelSet Kernels = getDeviceKernels(M);
 
-  // Create internal copies of each function if this is a kernel Module.
+  auto IsCalled = [&](Function &F) {
+    if (Kernels.contains(&F))
+      return true;
+    for (const User *U : F.users())
+      if (!isa<BlockAddress>(U))
+        return true;
+    return false;
+  };
+
+  // Create internal copies of each function if this is a kernel Module. This
+  // allows iterprocedural passes to see every call edge.
   DenseSet<const Function *> InternalizedFuncs;
   if (isOpenMPDevice(M))
     for (Function &F : M)
-      if (!F.isDeclaration() && !Kernels.contains(&F))
+      if (!F.isDeclaration() && !Kernels.contains(&F) && IsCalled(F))
         if (Attributor::internalizeFunction(F, /* Force */ true))
           InternalizedFuncs.insert(&F);
 
-  // Look at every function definition in the Module that wasn't internalized.
+  // Look at every function in the Module unless it was internalized.
   SmallVector<Function *, 16> SCC;
   for (Function &F : M)
     if (!F.isDeclaration() && !InternalizedFuncs.contains(&F))

diff  --git a/llvm/test/Transforms/OpenMP/remove_globalization.ll b/llvm/test/Transforms/OpenMP/remove_globalization.ll
index 2edf5d3c219d2..62dc54f52b5fe 100644
--- a/llvm/test/Transforms/OpenMP/remove_globalization.ll
+++ b/llvm/test/Transforms/OpenMP/remove_globalization.ll
@@ -6,6 +6,7 @@ target triple = "nvptx64"
 
 ; CHECK-REMARKS: remark: remove_globalization.c:4:2: Could not move globalized variable to the stack. Variable is potentially captured.
 ; CHECK-REMARKS: remark: remove_globalization.c:2:2: Moving globalized variable to the stack.
+; CHECK-REMARKS: remark: remove_globalization.c:6:2: Moving globalized variable to the stack.
 
 @S = external local_unnamed_addr global i8*
 
@@ -30,7 +31,7 @@ define internal void @foo() {
 ; CHECK-NEXT:    ret void
 ;
 entry:
-  %0 = call i8* @__kmpc_alloc_shared(i64 4), !dbg !11
+  %0 = call i8* @__kmpc_alloc_shared(i64 4), !dbg !12
   call void @use(i8* %0)
   call void @__kmpc_free_shared(i8* %0)
   ret void
@@ -46,7 +47,7 @@ define internal void @bar() {
 ; CHECK-NEXT:    ret void
 ;
 entry:
-  %0 = call i8* @__kmpc_alloc_shared(i64 4), !dbg !12
+  %0 = call i8* @__kmpc_alloc_shared(i64 4), !dbg !13
   call void @share(i8* %0)
   call void @__kmpc_free_shared(i8* %0)
   ret void
@@ -69,6 +70,14 @@ entry:
   ret void
 }
 
+define void @unused() {
+entry:
+  %0 = call i8* @__kmpc_alloc_shared(i64 4), !dbg !14
+  call void @use(i8* %0)
+  call void @__kmpc_free_shared(i8* %0)
+  ret void
+}
+
 ; CHECK: declare i8* @__kmpc_alloc_shared(i64)
 declare i8* @__kmpc_alloc_shared(i64)
 
@@ -87,8 +96,10 @@ declare void @__kmpc_free_shared(i8*)
 !5 = !{void ()* @kernel, !"kernel", i32 1}
 !6 = !{i32 7, !"openmp", i32 50}
 !7 = !{i32 7, !"openmp-device", i32 50}
-!8 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !10, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
-!9 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 1, type: !10, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
-!10 = !DISubroutineType(types: !2)
-!11 = !DILocation(line: 2, column: 2, scope: !8)
-!12 = !DILocation(line: 4, column: 2, scope: !9)
+!8 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !11, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!9 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 1, type: !11, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!10 = distinct !DISubprogram(name: "unused", scope: !1, file: !1, line: 1, type: !11, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!11 = !DISubroutineType(types: !2)
+!12 = !DILocation(line: 2, column: 2, scope: !8)
+!13 = !DILocation(line: 4, column: 2, scope: !9)
+!14 = !DILocation(line: 6, column: 2, scope: !9)