The hypothesis that JIT can be faster than C/C++ because it can adapt code during runtime is mostly nullified by PGO. Yes, you could argue that in some programs

The hypothesis that JIT can be faster than C/C++ because it can adapt code during runtime is mostly nullified by PGO. Yes, you could argue that in some programs the execution path might drastically vary and the profile obtained during compilation is not representative to the actual run. But that's probably very unlikely. And whatever gains you get from that have to offset the fact that these languages incite patterns that are generally awful for cache coherency, and generate a lot of garbage. Some C# microbenchmarks MAYBE could become faster than C++ in the future, but actual programs I think is extremely unlikely.