There is architecture specific code in CodeGen_ARM.cpp that can be accomplished using LLVM's vector predication intrinsics. However doing so requires investigating the final code that is generated and making sure there are no failures inside LLVM. (At the time of filing the issue, this is almost certainly not going to work, hence the way the code was written.)
There is architecture specific code in CodeGen_ARM.cpp that can be accomplished using LLVM's vector predication intrinsics. However doing so requires investigating the final code that is generated and making sure there are no failures inside LLVM. (At the time of filing the issue, this is almost certainly not going to work, hence the way the code was written.)