Job van der Zwan

More testing also revealed that loading from cache caused slowdown, so that was reduced from three cache loads (`pad`, early return and `ch`) back to one. ```js // Reduced usage...

I wanted to verify the behaviour across a more elaborate range than the old benchmarks: in theory this version should have a more gradual performance drop, rather than steep a...

A quick check to see what happens if the cache is bumped to length 64, and it's clear that we've hit the point of diminishing returns: ``` -> pad 600...

What we were doing: - check value of `ch`, default to `' '` - check if `ch` is `' '`, if so replace it with cache. We can smush that...

Eh... I don't understand why test 41 is failing? It works fine off-line.

For the record, I'm still working on this but it has been on the back-burner. Also, when I try to benchmark it I notice that *the order in which I...

I am, and I had found that changing the order changes results

Look, you should not trust benchmarks that use scores in this way anyway, they're the equivalent to dynamite plots, and the problems with dynamite plots are... well, a picture says...

> First, the Node.js benchmarks let you plot the results - fon scatter plots - see scatter.R and for the box plot you show above see compare.R - all these...

And *of course* this is not linked anywhere from the main Node.js website... Thanks Benjamin, you're a lifesaver