[GlobalMerge] Exit early if only one global is to be merged

[haicheng]

Summary

I think this can save some compilation time and prevent some unnecessary changes.

Two test cases are changed. In AArch64/global-merge-3.ll, y is not merged with any other globals, but is changed to @_MergedGlobals.1 (or __MergedGlobals_y in the case of ios). In ARM/global-merge.ll, g3 is not merged with other globals, but is changed to _MergedGlobals.1.

[dmgreen]

lib/CodeGen/GlobalMerge.cpp

473 ↗

(On Diff #145916)

Can you add a comment for this.

[haicheng]

Add the requested comment. Thank you for reviewing this.

[qcolombet]

Looks sensible to me.

test/CodeGen/ARM/global-merge.ll

63 ↗

(On Diff #146186)

Could you keep the original layout of the test (i.e., g1, g2, ..., in this order)

[haicheng]

Fix the test case. Thank you, Quentin.

[qcolombet]

test/CodeGen/ARM/global-merge.ll

63 ↗

(On Diff #146186)

What I meant here is don’t touch the input test, just modify the check lines.

[haicheng]

Change the test case. Thank you.

llvm/trunk/lib/CodeGen/GlobalMerge.cpp

bool GlobalMerge::doMerge(const SmallVectorImpl<GlobalVariable *> &Globals,
   assert(Globals.size() > 1);
 
   Type *Int32Ty = Type::getInt32Ty(M.getContext());
   auto &DL = M.getDataLayout();
 
   LLVM_DEBUG(dbgs() << " Trying to merge set, starts with #"
                     << GlobalSet.find_first() << "\n");
 
+  bool Changed = false;
   ssize_t i = GlobalSet.find_first();
   while (i != -1) {
     ssize_t j = 0;
     uint64_t MergedSize = 0;
     std::vector<Type*> Tys;
     std::vector<Constant*> Inits;
 
     bool HasExternal = false;
     StringRef FirstExternalName;
     for (j = i; j != -1; j = GlobalSet.find_next(j)) {
       Type *Ty = Globals[j]->getValueType();
       MergedSize += DL.getTypeAllocSize(Ty);
       if (MergedSize > MaxOffset) {
         break;
       }
       Tys.push_back(Ty);
       Inits.push_back(Globals[j]->getInitializer());
 
       if (Globals[j]->hasExternalLinkage() && !HasExternal) {
         HasExternal = true;
         FirstExternalName = Globals[j]->getName();
       }
     }
 
+    // Exit early if there is only one global to merge.
+    if (Tys.size() < 2) {
+      i = j;
+      continue;
+    }
+
     // If merged variables doesn't have external linkage, we needn't to expose
     // the symbol after merging.
     GlobalValue::LinkageTypes Linkage = HasExternal
                                             ? GlobalValue::ExternalLinkage
                                             : GlobalValue::InternalLinkage;
     StructType *MergedTy = StructType::get(M.getContext(), Tys);
     Constant *MergedInit = ConstantStruct::get(MergedTy, Inits);
 
for (ssize_t k = i, idx = 0; k != j; k = GlobalSet.find_next(k), ++idx) {
       if (Linkage != GlobalValue::InternalLinkage || !IsMachO) {
         GlobalAlias *GA =
             GlobalAlias::create(Tys[idx], AddrSpace, Linkage, Name, GEP, &M);
         GA->setDLLStorageClass(DLLStorage);
       }
 
       NumMerged++;
     }
+    Changed = true;
     i = j;
   }
 
-  return true;
+  return Changed;
 }
 
 void GlobalMerge::collectUsedGlobalVariables(Module &M) {
   // Extract global variables from llvm.used array
   const GlobalVariable *GV = M.getGlobalVariable("llvm.used");
   if (!GV || !GV->hasInitializer()) return;
 
   // Should be an array of 'i8*'.

llvm/trunk/test/CodeGen/AArch64/global-merge-3.ll

 ; RUN: llc %s -mtriple=aarch64-none-linux-gnu -aarch64-enable-global-merge -global-merge-on-external -disable-post-ra -o - | FileCheck %s
 ; RUN: llc %s -mtriple=aarch64-linux-gnuabi -aarch64-enable-global-merge -global-merge-on-external -disable-post-ra -o - | FileCheck %s
 ; RUN: llc %s -mtriple=aarch64-apple-ios -aarch64-enable-global-merge -global-merge-on-external -disable-post-ra -o - | FileCheck %s --check-prefix=CHECK-APPLE-IOS
 
 @x = global [1000 x i32] zeroinitializer, align 1
 @y = global [1000 x i32] zeroinitializer, align 1
 @z = internal global i32 1, align 4
 
 define void @f1(i32 %a1, i32 %a2, i32 %a3) {
 ;CHECK-APPLE-IOS: adrp  x8, __MergedGlobals_x@PAGE
 ;CHECK-APPLE-IOS-NOT: adrp
 ;CHECK-APPLE-IOS: add x8, x8, __MergedGlobals_x@PAGEOFF
-;CHECK-APPLE-IOS: adrp  x9, __MergedGlobals_y@PAGE+12
-;CHECK-APPLE-IOS: str w1, [x9, __MergedGlobals_y@PAGEOFF+12]
+;CHECK-APPLE-IOS: adrp  x9, _y@PAGE+12
+;CHECK-APPLE-IOS: str w1, [x9]
   %x3 = getelementptr inbounds [1000 x i32], [1000 x i32]* @x, i32 0, i64 3
   %y3 = getelementptr inbounds [1000 x i32], [1000 x i32]* @y, i32 0, i64 3
   store i32 %a1, i32* %x3, align 4
   store i32 %a2, i32* %y3, align 4
   store i32 %a3, i32* @z, align 4
   ret void
 }
 
 ;CHECK: .type .L_MergedGlobals,@object // @_MergedGlobals
 ;CHECK: .p2align  4
 ;CHECK: .L_MergedGlobals:
 ;CHECK: .size .L_MergedGlobals, 4004
 
-;CHECK: .type .L_MergedGlobals.1,@object // @_MergedGlobals.1
-;CHECK: .local  .L_MergedGlobals.1
-;CHECK: .comm .L_MergedGlobals.1,4000,16
+;CHECK-APPLE-IOS: .zerofill __DATA,__common,_y,4000,2
 
 ;CHECK-APPLE-IOS: .p2align  4
 ;CHECK-APPLE-IOS:  __MergedGlobals_x:
 ;CHECK-APPLE-IOS: .long 1
 ;CHECK-APPLE-IOS: .space  4000
 
-;CHECK-APPLE-IOS: .zerofill __DATA,__common,__MergedGlobals_y,4000,4
 
 ;CHECK: .set z, .L_MergedGlobals
 ;CHECK: .globl  x
 ;CHECK: .set x, .L_MergedGlobals+4
 ;CHECK: .size x, 4000
-;CHECK: .globl  y
-;CHECK: .set y, .L_MergedGlobals.1
-;CHECK: .size y, 4000
 
 ;CHECK-APPLE-IOS-NOT: .set _z, __MergedGlobals_x
 ;CHECK-APPLE-IOS:.globl _x
 ;CHECK-APPLE-IOS:.set _x, __MergedGlobals_x+4
-;CHECK-APPLE-IOS:.globl _y
-;CHECK-APPLE-IOS:.set _y, __MergedGlobals_y

llvm/trunk/test/CodeGen/ARM/global-merge.ll

 declare i32 @__gxx_personality_sj0(...)
 
 declare i32 @llvm.eh.typeid.for(i8*)
 
 declare i8* @__cxa_begin_catch(i8*)
 
 declare void @__cxa_end_catch()
 
+; CHECK: g3:
 ; CHECK: _MergedGlobals:
 @g1 = internal global i32 1
 @g2 = internal global i32 2
 
 ; Make sure that the complete variable fits within the range of the maximum
 ; offset.  Having the starting offset in range is not sufficient.
-; When this works properly, @g3 is placed in a separate chunk of merged globals.
-; CHECK: _MergedGlobals.1:
+; When this works properly, @g3 is not merged.
 @g3 = internal global [30 x i32] [ i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10 ], align 4
 
 ; Global variables that can be placed in BSS should be kept together in a
 ; separate pool of merged globals.
-; CHECK: _MergedGlobals.2
+; CHECK: _MergedGlobals.1
 @g4 = internal global i32 0
 @g5 = internal global i32 0
 
 ; Global variables that are constant can be merged together
-; CHECK: _MergedGlobals.3
+; CHECK: _MergedGlobals.2
 @g6 = internal constant [12 x i32] zeroinitializer, align 4
 @g7 = internal constant [12 x i32] zeroinitializer, align 4