Ah right; is that still bound by FM-index throughput then? Just walking through exact matches of a single read should be 10-100Mbp/s on a core.

Whoa, 96MB for 8 cores (12MB/core) is crazy indeed! That does kinda blast away my meagre 12MB total (2MB/core), and actually gives a much smoother transition to RAM.

I wonder what's the difference between this 64MB V-cache and 32MB on-die cache.

Yeah all constrained to total memory of course.
Kinda crazy though, the gap between 12MB L3 and 64GB RAM on my machine. Plateau of 5000x where once you're well above L3, it doesn't really matter how much space you actually use.

(Why get 8GB cloud instances though... Then laptops will do.)

The list goes on and on; my god.

> local youth shoots at immigrants with some airgun.
-> to make up for it, they end up playing a soccer tournament together

In first week of it opening: immigrant is exploring the area on a bike; the most Dutch thing one can do.
-> online narrative: immigrant stole a bike, which is just false 😠

> locals: install cameras, scared of bike theft.
-> reality: bike theft numbers are down after the opening.

If you're benchmarking in a loop, what you're really measuring is: how many iterations of this function fit in the execution window?

Oh didn't know of this! In principle sounds good but it seems more like it wants to manage the notifications in github itself (which I never use) rather than the mail notifications?