transfer /opt/local to another machine
jonesc at hep.phy.cam.ac.uk
Mon Apr 13 03:30:49 PDT 2015
> And it's more than likely that it will happen. It would surprise me somewhat if Atlas doesn't check for `-march=native` support, and that would give you AVX (and later) support if your i7 supports that.
> You ought to ask the question the other way round: how much of a (real-world) performance hit would you see when running Atlas built on a C2Duo on an i7.
> BTW: I do feel that "base" could add -march=native in user installations (with an optional override and not on the build bots) since AFAIU locally built packages are not intended to be distributed. The only times I've seen this flag cause issues is when running in a VM hosted on an i7, where the VM doesn't support the AVX instruction set and the compiler presumes it to be available.
For me that would just cause more problems that it would solve. Ports
would then be built differently based on where they are being built,
which is not a nice situation for debugging etc. I am also willing to
bet the number of ports that would significantly benefit from compiling
with higher instruction set support, AVX etc., is a very very select
few. For most there simply will not be much in the way of an appreciable
gain to the end user. Better to just for those few (like Atlas) handle
it specifically in the port file, as is done).
More information about the macports-users