Originally posted by starshipeleven
View Post
Reverse engineering efforts done outside Intel get into the tree as soon as they're usable and become the default configuration over "official" binaries (see Ivybridge).
Originally posted by starshipeleven
View Post
That makes it still rather x86-focused but it's more diverse than it used to be.
I really like to see all of that stuff improve: more reverse engineering, more non-x86 hardware, more open source contribution by the silicon vendors themselves. There are things I can't do (reverse engineering while having access to relevant information under NDA has rather bad optics), there are other things I do that aren't very visible (but that I'm quite certain have some positive impact).
Raptor considers x86 a lost cause and so they moved to Power9. More power to them (pun intended) but due to costs it remains a 1%er solution (while Raptor managed to bring it more into mainstream from the 0.1%er solution it was before, so yay). I'm not so sure that things will remain that open there once it becomes mainstream hardware. As soon as a Netflix client becomes a "hard" requirement for the platform's user base, Power will undergo lockdown attempts as well (and an open ISA won't help: the ISA was never the problem, fab access is). We need systematic solutions to such requests, not just ISA owners that are desperate to increase their marginal slice of market share but will throw you under the bus as soon as things take off.
I think x86 can be salvaged (not this chip generation, but the general direction) and through x86 and ARM (and their sway on the market, and the vendors, and everything) it might be possible to set healthier trends and standards for the entire industry.
Comment