freebsd stable - Aug 2016 - Benchmarks results for Compilers on FreeBSD 11

Il 2016-08-30 05:51 K. Macy ha scritto:
> I can't speak for the whole universe of users, but I think it's
safe
> to say that most users are not power users who individually configure
> ports tailored to their needs. I think my experiences on Ubuntu, where
> I'm definitely not a power user, are illustrative. I never compile
> *anything* that has a package in an ubuntu repo and I assume that the
> packages are configured when built to enable any performance options
> that don't potentially cause stability issues. Similarly, on FreeBSD
> most users are going to be using packages and they're going to assume
> that the packages are configured to "provide the best user
> experience". Consequently anyone using a package that could use OpenMP
> is going to legitimately just assume that "X" is slower on
FreeBSD.
> And for all intents and purposes "X" _is_ slower.
I second this 100%. 

If anyone thinks that this is not the "correct" approach then I
don't
see the point of the PKG project as a whole.

2016-08-30 9:04 GMT+02:00 Andrea Brancatelli <abrancatelli at
schema31.it>:
>
>
> Il 2016-08-30 05:51 K. Macy ha scritto:
>
> I can't speak for the whole universe of users, but I think it's
safe
> to say that most users are not power users who individually configure
> ports tailored to their needs. I think my experiences on Ubuntu, where
> I'm definitely not a power user, are illustrative. I never compile
> *anything* that has a package in an ubuntu repo and I assume that the
> packages are configured when built to enable any performance options
> that don't potentially cause stability issues. Similarly, on FreeBSD
> most users are going to be using packages and they're going to assume
> that the packages are configured to "provide the best user
> experience". Consequently anyone using a package that could use OpenMP
> is going to legitimately just assume that "X" is slower on
FreeBSD.
> And for all intents and purposes "X" _is_ slower.
>
>
> I second this 100%.
>
> If anyone thinks that this is not the "correct" approach then I
don't see
> the point of the PKG project as a whole.
>
I would also vote for "best performance per default". On a second
thought,
this would actually mean "average performance per default", because we
should be conservative as to what optimizations are enabled that still work
on older CPUs. I would say enabling all those compiler optimizations would
be a safe bet (simply going from -O to -O2).

As for pkg, if it can provide a sufficiently rich set of package options,
then I'm all in. The main reason I still compile ports is the hope of
gaining a bit of performance. Secondarily, compiling away features I don't
need. Admittedly, this is a bit of being a control freak, but I can see
servers were security is a concern who would want the bare minimum, and
desktops that would want all the bells and whistles. I think that is pretty
hard to achieve with a binary distribution, so I'll stick to building my
own ports with poudriere and then using pkg just out of convenience.