Hacker News Reader

Hacker News

Comment by unixhero | Hacker News Reader

Parent Full thread

unixhero

Those architectures help OpenBSD to smoke out bugs.

Obscure architectures is important also to keep the world online.

Also obscure architectures are damn fun.

View on HN

I bet the disappearance of big endian systems means a notable amount of bugs sit out there waiting.

I would imagine the new PowerPC64 port is running big endian.

Cool! Do you know which machines that port would boot?

http://www.openbsd.org/powerpc64.html

Current status:

Currently OpenBSD/powerpc64 only supports POWER9 processors and runs stably on PowerNV machines based on the Raptor Computing Systems Talos II and Blackbird boards. SMP is not yet supported. Support for POWER8 processors is planned but won't happen unless someone provides us access to suitable hardware.

OpenBSD/powerpc64 does not run under a hypervisor such as PowerVM or PowerKVM.

Are there advantages to big endian systems?

Network protocols usually encode things as big endian since that was the dominant order when the internet was being developed. It's a minor point now but everything has to be swapped around on little endian systems.

But is there some kind of computational advantage down to the processor architecture level? Can certain instructions or set of instructions be executed faster on big endian systems?

No. There may have been micro optimizations on specific chips, but there's nothing intrinsically faster about one encoding or the other.

Sorta like English reads right to left, but left to right or top to bottom would work just as well. left to right was what won.

Not sure about today's CPUs, but back in the days of the 68000 I'm pretty certain there was a small advantage in both designs but for different reasons. I wouldn't normally write such a vapid post without supporting details (that I can't recall anyway), but I wanted to pipe up to encourage you to do more research if you're curious.

No, no advantage. Just variety. You used to occasionally see bugs where things would work fine on little and not big, or vice-versa. Usually due to assumptions about data.

Reply on news.ycombinator.com