autovectorize spectralnorm

[llogiq]

I just added @TeXitoi's autovectorization suggestion to spectralnorm. This closes #9

Measurements are promising:

$ time bin/spectralnorm_old 5500
1.274224153

real    0m2.496s
user    0m9.159s
sys     0m0.022s

$ time bin/spectralnorm 5500
1.274224153

real    0m2.061s
user    0m7.263s
sys     0m0.027s

Since Rust uses the as-of-yet unstable simd types, there's little point in suggesting they update to this.

[TeXitoi]

A PR tourist is not relevant as it can use simd. But if you cannpropose a almost same version using simd, it'll be great.

[llogiq]

Sorry, I didn't get that. Do you mean I should add a #[cfg(simd)] feature that we can switch on and off depending on target?

[TeXitoi]

No, I badly read. I was thinking you'll doba PR to rust.

[llogiq]

Ah, I see. 😄 Ok, then.

[llogiq]

r?

[TeXitoi]

Thanks, I'll submit it tomorrow.

[llogiq]

This hasn't been submitted as of yet. Can I do something to speed up the process? This also applies to fasta-redux and the other recently updated benchmarks.

[TeXitoi]

Sorry for the delay: https://alioth.debian.org/tracker/index.php?func=detail&aid=315192&group_id=100815&atid=413122