faster-mult-arity-variadic.md

Currently:

    cljs.core.complement = (function cljs$core$complement(f) {
        return (function() {
            var G__10928 = null;
            var G__10928__0 = //impl
            var G__10928__1 = //impl
            var G__10928__2 = //impl
            var G__10928__3 = (function() {
                var G__10929__delegate = function(x, y, zs) {
                    // variadic implementation
                };
                var G__10929 = function(x, y, var_args) {
                    // dispatcher calling above delegate fn
                };
                G__10929.cljs$lang$maxFixedArity = 2;
                G__10929.cljs$lang$applyTo = // snip
                G__10929.cljs$core$IFn$_invoke$arity$variadic = G__10929__delegate;
                return G__10929;
            })();
            G__10928 = function(x, y, var_args) {
                // Dispatcher fn
                // it'll eventually call:
                return G__10928__3.cljs$core$IFn$_invoke$arity$variadic(x, y, G__10932);
            };
            G__10928.cljs$lang$maxFixedArity = 2;
            /// !!! Copies applyto
            G__10928.cljs$lang$applyTo = G__10928__3.cljs$lang$applyTo;
            G__10928.cljs$core$IFn$_invoke$arity$0 = G__10928__0;
            G__10928.cljs$core$IFn$_invoke$arity$1 = G__10928__1;
            G__10928.cljs$core$IFn$_invoke$arity$2 = G__10928__2;
            // !!! Only copies variadic impl
            G__10928.cljs$core$IFn$_invoke$arity$variadic = G__10928__3.cljs$core$IFn$_invoke$arity$variadic;
            return G__10928;
        })()
    }); 

New:

    cljs.core.complement = (function cljs$core$complement(f) {
        return (function() {
            var G__9681 = null;
            var G__9681__0 = // impl
            var G__9681__1 = // impl
            var G__9681__2 = // impl
            var G__9681__3__impl = // variadic impl
            var G__9681__3$cljs$lang$applyTo = // snip
            var G__9681__3$cljs$core$IFn$_invoke$arity$variadic = G__9681__3__impl;
            G__9681 = function(x, y, var_args) {
              // The fn dispatcher, now using:
              return G__9681__3$cljs$core$IFn$_invoke$arity$variadic(x, y, G__9684);
            };
            G__9681.cljs$lang$maxFixedArity = 2;
            // Copy applyTo:
            G__9681.cljs$lang$applyTo = G__9681__3$cljs$lang$applyTo;
            G__9681.cljs$core$IFn$_invoke$arity$0 = G__9681__0;
            G__9681.cljs$core$IFn$_invoke$arity$1 = G__9681__1;
            G__9681.cljs$core$IFn$_invoke$arity$2 = G__9681__2;
            // Reusing above variadic
            G__9681.cljs$core$IFn$_invoke$arity$variadic = G__9681__3$cljs$core$IFn$_invoke$arity$variadic;
            return G__9681;
        })()
    }); 

Benefits:

• Faster compile times?
• Smaller code
• Faster code by avoiding IIFE
• Better optimization by GCC

Benchmarks

    (enable-console-print!)
    (dotimes [_ 2]
      (simple-benchmark []
        (set! js/xxxx (comp inc dec +))
        1000000))

;; NEW: FF: 360ms   Chrome: 1300
;; OLD: FF: 470     Chrome: 1900

New benefit, probably better DCE by Closure Compiler.