Quickstart: "Is slower" is actually faster. Why?

[tornikeo]

Can anyone explain how do these results from docs->quickstart make any sense?

How is the second cell slower? It actually shows otherwise right there. It's faster by 50ms or so. Should we update the quickstart page?

[jakevdp]

Thanks for pointing this out. I believe it has to do with the fact that in numpy x.T is a zero-cost view of the original buffer, while in JAX outside of JIT-compilation, x.T results in allocation of a new buffer to hold the transpose.

We should probably update this example to make that more clear.