On 10/26/20 4:55 PM, Lennart Sorensen wrote: > And it seems power8+ machines really just run ppc64el because that is > what IBM wants people to run (apparently because converting the data from > big endian to little endian before passing it to an nvidia GPU that only > does little endian made a significant performance difference and there > was no reason not to switch to little endian given almost all powerpc > designs support both ways, not to mention the amount of software that > is broken on big endian because of assumptions (often implicit) by > the developers).
I had to re-read this several times and just to be clear: You can still run big-endian code perfectly fine on POWER8 machines and newer (unless IBM removed support with POWER10). In fact, we are running the little-endian port on the POWER8 server we got from IBM to host two ppc64 big-endian KVM VMs (one buildd and one Debian porterbox). Although that causes the host kernel to crash regularly with no fix in sight yet, unfortunately [1]. Adrian > [1] https://bugzilla.kernel.org/show_bug.cgi?id=206669 -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer - glaub...@debian.org `. `' Freie Universitaet Berlin - glaub...@physik.fu-berlin.de `- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
