Hello, Simon Josefsson, le lun. 07 sept. 2026 10:46:33 +0200, a ecrit: > Federico Vaga <[email protected]> writes: > > > As of today, we should be covered by Trixie (LTS + ELTS). However, we > > would like to have the possibility of upgrading to Forky. > > > > Having Forky support x86_64-v1 would give us a more modern system and > > a larger margin to both replace our x86_64-v1 hardware and cover the > > accelerator RUN (when beams will actually run). > > I think that is an argument for adding a 'amd64old' (or x86_64?) port > with a lower CPU baseline, possibly even a release arch. > > I would disagree it is a valid argument against increasing the baseline > for the normal amd64 port. > > If nobody manufacture and sell hardware for some profile any more (*), > the overall cost for retaining support for that hardware
What will be that cost? Do we e.g. have bugs that happen only on x86_64-v1 and not x86_64-v2? > will likely be overwelmingly offseted by the gains of increased > performance (and thus reduced energy consumption) for the remaining > fleet. How much gain is expected? > The cost of maintaining the 'amd64old' should also be a factor. Adding an arch means recompiling everything and one more arch to track. > I'm not convinced the cost of introducing the 'amd64old' port would > offset the cost to discard and replace any x86_64-v1 machines I advise you to watch the debconf talk about it https://debconf25.debconf.org/talks/129-approaching-the-speed-of-light-with-debian-debians-role-in-the-worlds-largest-particle-accelerator/ Replacing the x86_64-v1 machines *is* extremely costly in that environment. That's public money, I believe we should care. Samuel
