Code generated by managed language runtimes tend to have checks that are required for safety but never fail in practice. In such cases, it is profitable to make the non-failing case cheaper even if it makes the failing case significantly more expensive. This asymmetry can be exploited by folding such safety checks into operations that can be made to fault reliably if the check would have failed, and recovering from such a fault by using a signal handler.
For example, Java requires null checks on objects before they are read from or written to. If the object is null then a NullPointerException has to be thrown, interrupting normal execution. In practice, however, dereferencing a null pointer is extremely rare in well-behaved Java programs, and typically the null check can be folded into a nearby memory operation that operates on the same memory location.
Information about implicit checks generated by LLVM are put in a special "fault map" section. On Darwin this section is named __llvm_faultmaps.
The format of this section is
Header { uint8 : Fault Map Version (current version is 1) uint8 : Reserved (expected to be 0) uint16 : Reserved (expected to be 0) } uint32 : NumFunctions FunctionInfo[NumFunctions] { uint64 : FunctionAddress uint32 : NumFaultingPCs uint32 : Reserved (expected to be 0) FunctionFaultInfo[NumFaultingPCs] { uint32 : FaultKind uint32 : FaultingPCOffset uint32 : HandlerPCOffset } } FailtKind describes the reason of expected fault. Currently three kind of faults are supported:
FaultMaps::FaultingLoad- fault due to load from memory.FaultMaps::FaultingLoadStore- fault due to instruction load and store.FaultMaps::FaultingStore- fault due to store to memory.
The ImplicitNullChecks pass transforms explicit control flow for checking if a pointer is null, like:
%ptr = calli32*@get_ptr() %ptr_is_null = icmpi32*%ptr, nullbri1%ptr_is_null, label%is_null, label%not_null, !make.implicit!0 not_null: %t = loadi32, i32*%ptrbrlabel%do_something_with_t is_null: callvoid@HFC() unreachable!0 = !{}to control flow implicit in the instruction loading or storing through the pointer being null checked:
%ptr = calli32*@get_ptr() %t = loadi32, i32*%ptr;; handler-pc = label %is_nullbrlabel%do_something_with_t is_null: callvoid@HFC() unreachableThis transform happens at the MachineInstr level, not the LLVM IR level (so the above example is only representative, not literal). The ImplicitNullChecks pass runs during codegen, if -enable-implicit-null-checks is passed to llc.
The ImplicitNullChecks pass adds entries to the __llvm_faultmaps section described above as needed.
Making null checks implicit is an aggressive optimization, and it can be a net performance pessimization if too many memory operations end up faulting because of it. A language runtime typically needs to ensure that only a negligible number of implicit null checks actually fault once the application has reached a steady state. A standard way of doing this is by healing failed implicit null checks into explicit null checks via code patching or recompilation. It follows that there are two requirements an explicit null check needs to satisfy for it to be profitable to convert it to an implicit null check:
- The case where the pointer is actually null (i.e. the "failing" case) is extremely rare.
- The failing path heals the implicit null check into an explicit null check so that the application does not repeatedly page fault.
The frontend is expected to mark branches that satisfy (1) and (2) using a !make.implicit metadata node (the actual content of the metadata node is ignored). Only branches that are marked with !make.implicit metadata are considered as candidates for conversion into implicit null checks.
(Note that while we could deal with (1) using profiling data, dealing with (2) requires some information not present in branch profiles.)
