AArch64 Options in clang

raphael · December 29, 2022, 6:42pm

I stumbled upon some options for the arm64/aarch64 architecture, such as aarch64-use-tbi and so on. I noticed that they seem to be available for the Swift compiler but not for clang.

I looked into the libLLVM.dylib on my system (I am aware that Swift links the LLVM mostly statically into the compiler) and the shared library seems to atleast contain the symbols. In an additional step, I build clang myself, and wrapped the initialisation of the variable UseAddressTopByteIgnored that is repsonsible for adding the option aarch64-use-tbi. I am a bit dumb-founded as to why the variable has been initialised and the flag has been “added” but I cannot find the option with clang.

tschuett · December 29, 2022, 6:49pm

There is an LLVM option and not a clang option:

github.com

llvm/llvm-project/blob/5eec8dfc2b663bed7a90199f00277afb208bbd32/llvm/lib/Target/AArch64/AArch64Subtarget.cpp#L44


      
          #define GET_SUBTARGETINFO_CTOR
          #define GET_SUBTARGETINFO_TARGET_DESC
          #include "AArch64GenSubtargetInfo.inc"
          
          
static cl::opt<bool>
          EnableEarlyIfConvert("aarch64-early-ifcvt", cl::desc("Enable the early if "
                               "converter pass"), cl::init(true), cl::Hidden);
          
          
// If OS supports TBI, use this flag to enable it.
          static cl::opt<bool>
          UseAddressTopByteIgnored("aarch64-use-tbi", cl::desc("Assume that top byte of "
                                   "an address is ignored"), cl::init(false), cl::Hidden);
          
          
static cl::opt<bool>
              UseNonLazyBind("aarch64-enable-nonlazybind",
                             cl::desc("Call nonlazybind functions via direct GOT load"),
                             cl::init(false), cl::Hidden);
          
          
static cl::opt<bool> UseAA("aarch64-use-aa", cl::init(true),
                                     cl::desc("Enable the use of AA during codegen."));

The Swift way is swiftc -Xllvm aarch64-use-tbi. So it passes the option to LLVM. It is available to Clang. You have to use the same trick to pass the option to LLVM.

raphael · December 29, 2022, 7:01pm

Thank you for the fast answer. I kept using -Xllvm instead of clang’s -mllvm.

I stumbled on another problem in this context as well. With my self-built version I get following error

clang (LLVM option parsing): Unknown command line argument '--aarch64-use-tbi'.  Try: 'clang (LLVM option parsing) --help'
clang (LLVM option parsing): Did you mean '--aarch64-use-aa'?

Would you happen to have a pointer for this problem as well?

tschuett · December 29, 2022, 7:12pm

Unfortunately no. It tried clang-16 -mllvm aarch64-use-tbi ..., but I got similar problems.