Question

Based on language shootouts in a past few years, Clojure and other Lisps consistently perform better than most other dynamic languages. Why is that so?
Is it because of its homoiconicity?

Edit:

I did not know that Clojure is compiled into bytecode just like Java and Scala.
This stackoverflow thread threw light on why Clojure reaps the advantage of being both compiled and interpreted.

Answer 1

It is almost impossible to answer this question – it depends very much on how well the benchmark code is written, what exactly is being tested, whether you are allowed to use libraries that exploit native code, whether you are writing “idiomatic” code or optimising for performance etc.

So as always, you should treat all microbenchmarks with caution.

Having said that, the following reasons might give Clojure an advantage however in certain circumstances:

• Clojure is always compiled – first down to bytecode, then down to native code by the JVM JIT compiler. This can give it a speed advantage in many cases, particularly over langauges that rely on some form of interpretation. In theory at least, you should be able to match pure Java speed in any circumstance where Clojure enables you to produce the same bytecode (which is reasonably often, though not always….)
• Clojure can exploit JVM primitives and static typing – despite being a dynamic language, Clojure will compile statically types or primitive code if you give it enough hints. This can easily give a 10x boost in performance, though at the expense of making your code a bit longer/uglier…..
• Clojure has heavily optimised certain data structures and operations – in particular the immutable persistent data structures and certain functional programming constructs like “reduce”.
• Macros enable powerful compile-time optimisations – if you use macros cleverly, you can do some quite sophisticated optimisations at compile time, effectively using code-generation to produce the code that will be most efficient at runtime. This is an advantage many Lisps share (especially Common Lisp, which was one of the big inspirations for Clojure). As nulvinge points out, homoiconicity isn’t strictly necessary to achieve this (C++ also has macros!), but being a homoiconic language makes macros a lot easier.
• The modern JVMs are brilliantly engineered – Clojure takes advantage of the thousands of man-years of engineering that have gone into the JVM, the Java runtime libraries, the garbage collection algorithms etc. Non-JVM languages don’t get these benefits.