[llvm] [Analysis] Add Scalable field in MemoryLocation.h (PR #69716)

[Nikita Popov via llvm-commits]

================
@@ -101,22 +101,23 @@ bool BasicAAResult::invalidate(Function &Fn, const PreservedAnalyses &PA,
 //===----------------------------------------------------------------------===//
 
 /// Returns the size of the object specified by V or UnknownSize if unknown.
-static uint64_t getObjectSize(const Value *V, const DataLayout &DL,
-                              const TargetLibraryInfo &TLI,
-                              bool NullIsValidLoc,
-                              bool RoundToAlign = false) {
+/// getObjectSize does not support scalable Value
+static LocationSize getObjectSize(const Value *V, const DataLayout &DL,
+                                  const TargetLibraryInfo &TLI,
+                                  bool NullIsValidLoc,
+                                  bool RoundToAlign = false) {
   uint64_t Size;
   ObjectSizeOpts Opts;
   Opts.RoundToAlign = RoundToAlign;
   Opts.NullIsUnknownSize = NullIsValidLoc;
   if (getObjectSize(V, Size, DL, &TLI, Opts))
-    return Size;
-  return MemoryLocation::UnknownSize;
+    return LocationSize(Size);
+  return LocationSize(MemoryLocation::UnknownSize);
----------------
nikic wrote:

What do we gain by returning LocationSize from this function? I think this code mostly only works by accident (MemoryLocation::UnknownSize gets converted to LocationSize::afterPointer() by being out of range, and then the comparison below also does this conversion via the non-explicit ctor).

If this should be something other than uint64_t, then it should probably be TypeSize, not LocationSize.

https://github.com/llvm/llvm-project/pull/69716