CERN's migration path from CentOS Linux to Debian
22 points by zg
22 points by zg
This is happening while Debian, always happy to kowtow to their corporate leaders, is discussing raising the baseline for amd64 / x86-64 from v1 to v3.
The discussions have been rather handwavy about how all the "benefits" of additional performance and power savings will somehow be real, with no actual data. The corporate minded people also talk of the "cost for retaining support". Get that - they're marketing so hard to non-technical people that they actually can't help but repeat their marketing points to technical people, as if programmers will need to sit around and spend programming hours figuring out how to support v1 instead of v3.
It makes me sad that they're not even pretending to care about facts and actual, real world data.
I could see wanting v2 at least to get support for POPCNT.
Is the motivation for going from v1 straight to v3 to target AVX support?
From what I found, and here just for reference...
v3 - since 2013
This is very misleading. My GPD MicroPC, released in late 2019 and purchased in late 2023 has Intel Celeron N4120 (also released late 2019), that does not support v3.
It is powerful enough to be an excellent personal computer, even for modern internet. I run full blown NixOS on it, and even occasionally compile Emacs, works great.
NixOS community had a short discussion to go v3 last year but was effectively shot down. Ubuntu has not been an option for me for a decade. Debian has always been my 'safe' fallback, being as conservative as they are (a plus point). If Debian and NixOS both switch over, its a step change for people like me. There are A LOT of people on hardware perfectly powerful enough for their needs. Especially with Intel's stagnation decade. Things like performance improvements should not be thrown around without considering drawbacks and solid numbers backing both sides.
This is happening while Debian, always happy to kowtow to their corporate leaders
And who are those?