Posts and replies

@AsahiLinux As someone considering buying an m2 Mac studio mainly for compiling aarch64 RPMS, is native much faster than via UTM? (e.g. x86_64 Linux running x86_64 Linux under qemu isn't particularly noticeable).

Or can you dual boot asahi, seems a shame to pay inflated apple prices and not have macos to play with too.

Do you have any compiling a kernel RPM timings? Colleagues say about 70mins on M1 max under UTM, about the same as Xeon+qemu

@sej7278 @AsahiLinux the fedora asahi remix kernel rpm (6.5.6 plus asahi patches) takes on a m1 ultra around 75 minutes. that are 4 sequential kernel builds, 4k + 16k variants with and without debug. rpmbuild itself is unfortunately mostly just using a single core and the post build processing takes a considerable amount of time. average CPU use is 1200% while the kernel build itself uses all 20 cores.
The M2 Ultra should be a little faster.
Dual booting is no problem and for now the expected setup.

@janne @AsahiLinux thanks. Seems RPM is the problem, as pretty sure building a kernel shouldn't take that long and dual xeons or absurd amounts of ram don't help, I've seen nothing build any kernel RPM in under 73mins (baremetal, mock, VM....) and M1 max/ultra sound the same

@sej7278 @AsahiLinux the m1/m2 ultra should be able to build the plain fedora aarch64 kernel rpm under 40 minutes. the asahi kernel rpm build takes twice as much since it builds the kernel once with 4k page size like the fedora build and once with 16k page size (optimal/required for apple silicon devices).

@janne @AsahiLinux oh right so the default 4kb in most distro's (or additionally 64kb) page sizes just don't work on apple silicon like they do on e.g. qemu/raspberrypi/aws kit?

Sorry for all the questions, i feel like i should be reading this on a wiki somewhere....