Added benchmark info

README.md

@@ -8,10 +8,10 @@
 
 __[Live demo](http://jonschlinkert.github.io/remarkable/demo/)__
 
-- Supports the [CommonMark](http://commonmark.org/) spec + extentions
-  (URL autolinking, typographer).
+- Supports the [CommonMark](http://commonmark.org/) spec +
+  [syntax extensions](#syntax-extensions) + sugar (URL autolinking, typographer).
 - Configurable syntax! You can add new rules and even replace existing ones.
-- High speed! See the [benchmarks](./benchmark).
+- [High speed](#benchmark)!
 
 
 ## Install
@@ -152,7 +152,7 @@ var md = new Remarkable({
 ```
 
 
-### Syntax extentions
+### Syntax extensions
 
 Enabled by default:
 
@@ -162,12 +162,12 @@ Enabled by default:
 
 Disabled by default:
 
-- [\<sup](http://johnmacfarlane.net/pandoc/README.html#superscripts-and-subscripts) - `19^th^`
+- [\<sup>](http://johnmacfarlane.net/pandoc/README.html#superscripts-and-subscripts) - `19^th^`
 - [\<sub>](http://johnmacfarlane.net/pandoc/README.html#superscripts-and-subscripts) - `H~2~0`
 - __\<ins>__ - `++inserted text++` (experimental)
 - __\<mark>__ - `==marked text==` (experimental)
 
-__*__ Experimental extentions can be changed later for something like
+__*__ Experimental extensions can be changed later for something like
 [Critic Markup](http://criticmarkup.com/), but you will still be able to use
 old-style rules via external plugins if you prefer.
 
@@ -262,6 +262,25 @@ Remarkable.renderer
 Remarkable.renderer.rules
 ```
 
+## Benchmark
+
+Here is result of CommonMark spec parse at Core i5 2.4 GHz (i5-4258U):
+
+```bash
+$ benchmark/benchmark.js spec
+Selected samples: (1 of 27)
+ > spec
+
+Sample: spec.txt (110610 bytes)
+ > commonmark-reference x 40.42 ops/sec ±4.07% (51 runs sampled)
+ > current x 74.99 ops/sec ±4.69% (67 runs sampled)
+ > current-commonmark x 93.76 ops/sec ±1.23% (79 runs sampled)
+ > marked-0.3.2 x 22.92 ops/sec ±0.79% (41 runs sampled)
+```
+
+As you can see, `remarkabe` doesn't pay with speed for it's flexibility. Because
+it's written in monomorphyc style and use JIT inline caches effectively.
+
 
 ## Authors