But it's not really compatible with the GCC IFUNC scheme ... PTL entries will be permanently remapped to the most appropriate code on the CPU where the function is first called, and never thereafter remapped. So you end up with a coin toss whether you get the optimized function or not.
Personally I don't find the e-cores on my alder lake CPU to be of any value. They're more of a hazard than a benefit.
