Hi, I need a module with physical cpus for both Linux and OS X, and I thought of adding it to cpuinfo, but I think the current API isn't going to work for anything other than linux. The reason is the CPU type has a million fields, many of them are linux-specific, and the useful functions like physicalCores require CPUs.

What I would do is change functions like physicalCores to have type IO Int. This is implemented as /usr/sbin/sysctl -n hw.physicalcpu on OS X, and by parsing /proc/cpuinfo on linux. If they all throw the same NotSupported exception, we can have a try which turns them into Maybe Int if you don't want to crash on an unsupported platform.

Personally I'd just get rid of the CPU type because I don't have a use for all those fields, but I guess it doesn't hurt to have it, as long as it's clear that it's Linux-specific. Even use CPP to make it not compile on non-linux instead of compile but crash at runtime, but I don't care much either way.

Would you be interested in a change like that? It would be basically a rewrite and incompatible API. Otherwise I can make my own package easily enough, but it's a bit of a shame to have two similar packages in hackage. You could also transfer ownership to me if you'd like.