galtsev at kicp.uchicago.edu
Fri May 1 13:06:56 UTC 2020
On 5/1/20 5:51 AM, Polytropon wrote:
> On Fri, 1 May 2020 11:21:56 +0200, Matthias Gamsjager wrote:
>> We have binary package so you don't have to compile your self.
>> Of course it is a choice to compile everything but why would you want to do
>> that on a small machine?
> Why? Especially because! :-)
> In ye olden times, you often used source-based installation
> methods to tweak the amount of what gets installed (memory
> footprint), and you dealt with cimpile-time options to get
> faster software - faster than what the default configuration
> allowed. For example, system tools could be omitted, or the
> kernel could be configured in a way to only contain the
> stuff needed for a particular system. It was also useful
> for ports where you needed to deviate from the default
> options, or where you were forced (!) to use source-based
> installs due to licensing restrictions.
> For those who wish to track -STABE or -HEAD, source-based
> installations are mandatory. Maybe someone wants to check
> if a specific patch works as intended - the whole system
> or just one of its components can be built and installed.
> This currently is impossible with binary packages.
> While I personally enjoy using binary packages, they are
> not an answer to every scenario, because there simply is
> no "one size fits all egg-laying wool-milk-sow".
> Machines equipped with slower disks and less memory will
> of course need more time to build something. This is why
> several users keep their machine running at night where
> it can compile happily. On a 150 MHz Pentium with 64 MB
> RAM, building a kernel required a few hours, and the whole
> system needed 24 hours to build. With today's hardware,
> compile times are faster. And especially for building ports,
> some people use their own build servers (real or VM) for
> this task.
>> If you really want to see how fast it could go. Spin up a machine on AWS
>> with the memory and CPUs you would compare it to.
> Comparing bare metal to virtual metal is like cheating
> in statistics - choose your test subjects in a specific
> way to get any result you want. :-)
Agree with the sentimet. That aside, Unixbench, maybe?
Sr System Administrator
Department of Astronomy and Astrophysics
Kavli Institute for Cosmological Physics
University of Chicago
More information about the freebsd-questions