> any optimizations that can be preformed by the cumbersome JVM can be preformed in LLVM
That cumbersome JVM can optimise based on runtime knowledge of the characteristics of the running code. In many cases, that kind of optimisation is more effective than the static compile-time sort.
I am curious as to what exactly can be optimized by the JVM runtime vs. LLVM. Do you mean targeting specific machine architectures? Because LLVM can generate images for different processors (e.g., different version of Intel architectures).
That cumbersome JVM can optimise based on runtime knowledge of the characteristics of the running code. In many cases, that kind of optimisation is more effective than the static compile-time sort.