Support non-malloc functions in malloc+memset->calloc fold

Author: jamie-osec

llvm/docs/LangRef.rst

@@ -1954,6 +1954,10 @@ For example:
     The first three options are mutually exclusive, and the remaining options
     describe more details of how the function behaves. The remaining options
     are invalid for "free"-type functions.
+``"alloc-variant-zeroed"="FUNCTION"``
+    This attribute indicates that another function is equivalent to an allocator function,
+    but returns zeroed memory. The function must have "zeroed" allocation behavior,
+    the same ``alloc-family``, and take exactly the same arguments.
 ``allocsize(<EltSizeParam>[, <NumEltsParam>])``
     This attribute indicates that the annotated function will always return at
     least a given number of bytes (or null). Its arguments are zero-indexed

llvm/include/llvm/IR/Attributes.h

@@ -156,6 +156,7 @@ class Attribute {
   static Attribute getWithAllocSizeArgs(
       LLVMContext &Context, unsigned ElemSizeArg,
       const std::optional<unsigned> &NumElemsArg);
+  static Attribute getWithAllocKind(LLVMContext &Context, AllocFnKind Kind);
   static Attribute getWithVScaleRangeArgs(LLVMContext &Context,
                                           unsigned MinValue, unsigned MaxValue);
   static Attribute getWithByValType(LLVMContext &Context, Type *Ty);

llvm/lib/IR/Attributes.cpp

@@ -300,6 +300,10 @@ Attribute::getWithAllocSizeArgs(LLVMContext &Context, unsigned ElemSizeArg,
   return get(Context, AllocSize, packAllocSizeArgs(ElemSizeArg, NumElemsArg));
 }
 
+Attribute Attribute::getWithAllocKind(LLVMContext &Context, AllocFnKind Kind) {
+  return get(Context, AllocKind, static_cast<uint64_t>(Kind));
+}
+
 Attribute Attribute::getWithVScaleRangeArgs(LLVMContext &Context,
                                             unsigned MinValue,
                                             unsigned MaxValue) {

llvm/lib/IR/Verifier.cpp

@@ -2377,6 +2377,32 @@ void Verifier::verifyFunctionAttrs(FunctionType *FT, AttributeList Attrs,
       CheckFailed("'allockind()' can't be both zeroed and uninitialized");
   }
 
+  if (Attrs.hasFnAttr("alloc-variant-zeroed")) {
+    if (auto A = Attrs.getFnAttr("alloc-variant-zeroed"); A.isValid()) {
+      StringRef s = A.getValueAsString();
+      Check(!s.empty(), "\"alloc-variant-zeroed\" must not be empty");
+      if (auto *F = dyn_cast<Function>(V)) {
+        if (auto Variant = F->getParent()->getFunction(s)) {
+          auto Family = Attrs.getFnAttr("alloc-family");
+          if (Family.isValid())
+            Check(Variant->getFnAttribute("alloc-family").isValid() &&
+                      Variant->getFnAttribute("alloc-family")
+                              .getValueAsString() == Family.getValueAsString(),
+                  "\"alloc-variant-zeroed\" must name a function belonging to "
+                  "the "
+                  "same alloc-family");
+
+          Check(Variant->hasFnAttribute(Attribute::AllocKind) &&
+                    (Variant->getFnAttribute(Attribute::AllocKind)
+                         .getAllocKind() &
+                     AllocFnKind::Zeroed) != AllocFnKind::Unknown,
+                "\"alloc-variant-zeroed\" must name a function with "
+                "`allockind(\"zeroed\")`");
+        }
+      }
+    }
+  }
+
   if (Attrs.hasFnAttr(Attribute::VScaleRange)) {
     unsigned VScaleMin = Attrs.getFnAttrs().getVScaleRangeMin();
     if (VScaleMin == 0)

llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp

@@ -2028,9 +2028,18 @@ struct DSEState {
     if (!InnerCallee)
       return false;
     LibFunc Func;
+    std::optional<StringRef> ZeroedVariantName = std::nullopt;
     if (!TLI.getLibFunc(*InnerCallee, Func) || !TLI.has(Func) ||
-        Func != LibFunc_malloc)
-      return false;
+        Func != LibFunc_malloc) {
+      if (!Malloc->getFnAttr("alloc-variant-zeroed").isValid()) {
+        return false;
+      }
+      ZeroedVariantName =
+          Malloc->getFnAttr("alloc-variant-zeroed").getValueAsString();
+      if (ZeroedVariantName->empty())
+        return false;
+    }
+
     // Gracefully handle malloc with unexpected memory attributes.
     auto *MallocDef = dyn_cast_or_null<MemoryDef>(MSSA.getMemoryAccess(Malloc));
     if (!MallocDef)
@@ -2057,15 +2066,31 @@ struct DSEState {
 
     if (Malloc->getOperand(0) != MemSet->getLength())
       return false;
-    if (!shouldCreateCalloc(Malloc, MemSet) ||
-        !DT.dominates(Malloc, MemSet) ||
+    if (!shouldCreateCalloc(Malloc, MemSet) || !DT.dominates(Malloc, MemSet) ||
         !memoryIsNotModifiedBetween(Malloc, MemSet, BatchAA, DL, &DT))
       return false;
     IRBuilder<> IRB(Malloc);
-    Type *SizeTTy = Malloc->getArgOperand(0)->getType();
-    auto *Calloc =
-        emitCalloc(ConstantInt::get(SizeTTy, 1), Malloc->getArgOperand(0), IRB,
-                   TLI, Malloc->getType()->getPointerAddressSpace());
+    assert(Func == LibFunc_malloc || ZeroedVariantName.has_value());
+    Value *Calloc = nullptr;
+    if (ZeroedVariantName.has_value()) {
+      auto &Ctx = Malloc->getContext();
+      auto Attr = InnerCallee->getAttributes();
+      auto AllocKind = Attr.getFnAttr(Attribute::AllocKind).getAllocKind() |
+                       AllocFnKind::Zeroed;
+      Attr =
+          Attr.addFnAttribute(Ctx, Attribute::getWithAllocKind(Ctx, AllocKind));
+      auto ZeroedVariant = Malloc->getModule()->getOrInsertFunction(
+          *ZeroedVariantName, InnerCallee->getFunctionType(), Attr);
+      SmallVector<Value *, 3> Args;
+      for (unsigned I = 0; I < Malloc->arg_size(); I++)
+        Args.push_back(Malloc->getArgOperand(I));
+      Calloc = IRB.CreateCall(ZeroedVariant, Args, *ZeroedVariantName);
+    } else {
+      Type *SizeTTy = Malloc->getArgOperand(0)->getType();
+      Calloc =
+          emitCalloc(ConstantInt::get(SizeTTy, 1), Malloc->getArgOperand(0),
+                     IRB, TLI, Malloc->getType()->getPointerAddressSpace());
+    }
     if (!Calloc)
       return false;
 

llvm/test/Transforms/DeadStoreElimination/noop-stores.ll

@@ -374,6 +374,19 @@ define ptr @notmalloc_memset(i64 %size, ptr %notmalloc) {
   ret ptr %call1
 }
 
+; This should create a customalloc_zeroed
+define ptr @customalloc_memset(i64 %size, i64 %align) {
+; CHECK-LABEL: @customalloc_memset
+; CHECK-NEXT:  [[CALL:%.*]] = call ptr @customalloc_zeroed(i64 [[SIZE:%.*]], i64 [[ALIGN:%.*]])
+; CHECK-NEXT:  ret ptr [[CALL]]
+  %call = call ptr @customalloc(i64 %size, i64 %align)
+  call void @llvm.memset.p0.i64(ptr %call, i8 0, i64 %size, i1 false)
+  ret ptr %call
+}
+
+declare ptr @customalloc(i64, i64) allockind("alloc") "alloc-family"="customalloc" "alloc-variant-zeroed"="customalloc_zeroed"
+declare ptr @customalloc_zeroed(i64, i64) allockind("alloc,zeroed") "alloc-family"="customalloc"
+
 ; This should not create recursive call to calloc.
 define ptr @calloc(i64 %nmemb, i64 %size) inaccessiblememonly {
 ; CHECK-LABEL: @calloc(