[llvm] [Offload] Full AMD support for olMemFill (PR #154958)

[Matt Arsenault via llvm-commits]

================
@@ -1216,7 +1222,12 @@ struct AMDGPUStreamTy {
     assert(Args->Dst && "Invalid destination buffer");
     assert(Args->Src && "Invalid source buffer");
 
-    std::memcpy(Args->Dst, Args->Src, Args->Size);
+    auto BasePtr = Args->Dst;
+    for (size_t I = 0; I < Args->NumTimes; I++) {
+      std::memcpy(BasePtr, Args->Src, Args->Size);
+      BasePtr = reinterpret_cast<void *>(reinterpret_cast<uintptr_t>(BasePtr) +
----------------
arsenm wrote:

Why does this need to cast to uintptr_t? 

https://github.com/llvm/llvm-project/pull/154958