Another example of this is found in comparing generic MCMC methods to HMC. The former work for discrete or continuous parameters while the latter requires a gra

Another example of this is found in comparing generic MCMC methods to HMC. The former work for discrete or continuous parameters while the latter requires a gradient. Regardless of how fast a generic MCMC implementation is, the HMC method is asymptotically faster, so for some problem sizes the language or low level optimizations are irrelevant.