Handle scalable store size in MemCpyOptimizer

The compiler crashes with an ICE when it tries to create a `memset` with
scalable size.

Author: momchil-velikov

llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp

@@ -800,8 +800,9 @@ bool MemCpyOptPass::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
     // in subsequent passes.
     auto *T = V->getType();
     if (T->isAggregateType()) {
-      uint64_t Size = DL.getTypeStoreSize(T);
       IRBuilder<> Builder(SI);
+      Value *Size =
+          Builder.CreateTypeSize(Builder.getInt64Ty(), DL.getTypeStoreSize(T));
       auto *M = Builder.CreateMemSet(SI->getPointerOperand(), ByteVal, Size,
                                      SI->getAlign());
       M->copyMetadata(*SI, LLVMContext::MD_DIAssignID);

llvm/test/CodeGen/AArch64/memset-scalable-size.ll

@@ -0,0 +1,56 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S --passes=memcpyopt < %s | FileCheck %s
+target triple = "aarch64-unknown-linux"
+
+define void @f0() {
+; CHECK-LABEL: define void @f0() {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[P:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 2
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 2 [[P]], i8 0, i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    call void @g(ptr [[P]])
+; CHECK-NEXT:    ret void
+;
+entry:
+  %p = alloca { <vscale x 16 x i1>, <vscale x 16 x i1>}, align 2
+  store { <vscale x 16 x i1>, <vscale x 16 x i1> } zeroinitializer, ptr %p, align 2
+  call void @g(ptr %p)
+  ret void
+}
+
+define void @f1() {
+; CHECK-LABEL: define void @f1() {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[P:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 48
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[P]], i8 0, i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    call void @g(ptr [[P]])
+; CHECK-NEXT:    ret void
+;
+entry:
+  %p = alloca {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
+  store {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } zeroinitializer, ptr %p, align 16
+  call void @g(ptr %p)
+  ret void
+}
+
+define void @f2() {
+; CHECK-LABEL: define void @f2() {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[P:%.*]] = alloca { <vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x double> }, align 16
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 192
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[P]], i8 0, i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    call void @g(ptr [[P]])
+; CHECK-NEXT:    ret void
+;
+entry:
+  %p = alloca {<vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x double> }, align 16
+  store {<vscale x 8 x double>, <vscale x 8 x double>, <vscale x 8 x double> } zeroinitializer, ptr %p, align 16
+  call void @g(ptr %p)
+  ret void
+}
+
+declare void @g(ptr)