[PATCH] D34311: [InstCombine] Don't replace allocas with smaller globals

[Vitaly Buka via Phabricator via llvm-commits]

vitalybuka updated this revision to Diff 103811.
vitalybuka added a comment.

Use isDereferenceableAndAlignedPointer to check global size


https://reviews.llvm.org/D34311

Files:
  lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
  test/Transforms/InstCombine/memcpy-from-global.ll


Index: test/Transforms/InstCombine/memcpy-from-global.ll
===================================================================
--- test/Transforms/InstCombine/memcpy-from-global.ll
+++ test/Transforms/InstCombine/memcpy-from-global.ll
@@ -204,3 +204,34 @@
 ; CHECK-NEXT: call void @bar(i8* bitcast (%U* getelementptr inbounds ([2 x %U], [2 x %U]* @H, i64 0, i64 1) to i8*))
   ret void
 }
+
+ at bbb = local_unnamed_addr global [1000000 x i8] zeroinitializer, align 16
+ at _ZL3KKK = internal unnamed_addr constant [3 x i8] c"\01\01\02", align 1
+
+; Should not replace alloca with global because of size mismatch.
+define void @test9_small_global() {
+; CHECK-LABEL: @test9_small_global(
+; CHECK-NOT: call void @llvm.memcpy.p0i8.p0i8.i64({{.*}}@bbb,{{.*}}@_ZL3KKK, 
+; CHECK: alloca [1000000 x i8]
+entry:
+  %cc = alloca [1000000 x i8], align 16
+  %cc.0..sroa_idx = getelementptr inbounds [1000000 x i8], [1000000 x i8]* %cc, i64 0, i64 0
+  %arraydecay = getelementptr inbounds [1000000 x i8], [1000000 x i8]* %cc, i32 0, i32 0
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %arraydecay, i8* getelementptr inbounds ([3 x i8], [3 x i8]* @_ZL3KKK, i32 0, i32 0), i64 3, i32 1, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* getelementptr inbounds ([1000000 x i8], [1000000 x i8]* @bbb, i32 0, i32 0), i8* %arraydecay, i64 1000000, i32 16, i1 false)
+  ret void
+}
+
+; Should replace alloca with global as they have exactly the same size.
+define void @test10_same_global() {
+; CHECK-LABEL: @test10_same_global(
+; CHECK-NOT: alloca
+; CHECK: call void @llvm.memcpy.p0i8.p0i8.i64({{.*}}@bbb,{{.*}}@_ZL3KKK,{{.*}}, i64 3,
+entry:
+  %cc = alloca [3 x i8], align 1
+  %cc.0..sroa_idx = getelementptr inbounds [3 x i8], [3 x i8]* %cc, i64 0, i64 0
+  %arraydecay = getelementptr inbounds [3 x i8], [3 x i8]* %cc, i32 0, i32 0
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %arraydecay, i8* getelementptr inbounds ([3 x i8], [3 x i8]* @_ZL3KKK, i32 0, i32 0), i64 3, i32 1, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* getelementptr inbounds ([1000000 x i8], [1000000 x i8]* @bbb, i32 0, i32 0), i8* %arraydecay, i64 3, i32 1, i1 false)
+  ret void
+}
Index: lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
===================================================================
--- lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
+++ lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
@@ -169,6 +169,17 @@
   return nullptr;
 }
 
+/// Returns true if V is dereferenceable for size of alloca.
+static bool isDereferenceableForAllocaSize(const Value *V, const AllocaInst *AI,
+                                           const DataLayout &DL) {
+  if (AI->isArrayAllocation())
+    return false;
+  uint64_t AllocaSize = DL.getTypeStoreSize(AI->getAllocatedType());
+  if (!AllocaSize)
+    return false;
+  return isDereferenceableAndAlignedPointer(V, 1, APInt(64, AllocaSize), DL);
+}
+
 static Instruction *simplifyAllocaArraySize(InstCombiner &IC, AllocaInst &AI) {
   // Check for array size of 1 (scalar allocation).
   if (!AI.isArrayAllocation()) {
@@ -390,7 +401,8 @@
     if (MemTransferInst *Copy = isOnlyCopiedFromConstantGlobal(&AI, ToDelete)) {
       unsigned SourceAlign = getOrEnforceKnownAlignment(
           Copy->getSource(), AI.getAlignment(), DL, &AI, &AC, &DT);
-      if (AI.getAlignment() <= SourceAlign) {
+      if (AI.getAlignment() <= SourceAlign &&
+          isDereferenceableForAllocaSize(Copy->getSource(), &AI, DL)) {
         DEBUG(dbgs() << "Found alloca equal to global: " << AI << '\n');
         DEBUG(dbgs() << "  memcpy = " << *Copy << '\n');
         for (unsigned i = 0, e = ToDelete.size(); i != e; ++i)


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D34311.103811.patch
Type: text/x-patch
Size: 3672 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170624/e7c98f1e/attachment.bin>