[llvm-branch-commits] [llvm] [libsycl] Add liboffload kernel creation (PR #188794)

[Yury Plyakhin via llvm-branch-commits]

================
@@ -119,7 +141,6 @@ static bool isImageTargetCompatible(const DeviceImageWrapper &Image,
 DeviceImageWrapper *ProgramManager::getDeviceImage(std::string_view KernelName,
                                                    const kernel_id &KernelID,
                                                    DeviceImpl &Device) {
-  std::lock_guard<std::mutex> Guard(MImageCollectionMutex);
   auto [Begin, End] = MKernelIDToDevImageJIT.equal_range(KernelID);
----------------
YuriPlyakhin wrote:

`MKernelIDToDevImageJIT` is `unordered_map`, so `equal_range` can return only 0 or 1 element. Based on code below it seems more than 1 kernel might be expected. should `MKernelIDToDevImageJIT` be `unordered_multimap`?

https://github.com/llvm/llvm-project/pull/188794