The obfuscation in question was just name mangling, so it was probably ~zero-cost. Possibly slightly negative cost, depending on the relative verbosity of method names.