XenevaOS – Modern Computing Reimagined

Post by rbanffy » Fri, Jul 04, 2025, 8:50 AM UTC

gsf_emergency_2 wrote: post #10

PaulHoule wrote: post #8

Earlier quoted context omitted.

Real innovation in OS is somewhere between impractical and impossible. You might think POSIX is boring and bloated but if if you really change the relationship of applications and the OS you will need all new applications. If you like microkernels, replace POSIX with Mach or maybe L3. Hypothetically you could make a POSIX implementation which beats Linux for some particular special case but Linux has so much investme…

> Real innovation in OS is somewhere between impractical and impossible. This explains Steve Jobs' saying that you have to work on hardware. The next OS will at first run on GPU/FPGA only Guess PowerPC wasn't enough of a difference (the OS innovation was thus a layer on top of BSD) & therefore neither is RISC-V, it would seem

Post by rbanffy » Fri, Jul 04, 2025, 6:27 PM UTC

gsf_emergency_2 wrote: post #12

rbanffy wrote: post #11

Earlier quoted context omitted.

> The next OS will at first run on GPU/FPGA only Modern CPUs co-evolved with the software we run on them. If we really want to rethink software, we need to rethink the CPU itself.

That's what I meant.. probably I will use your phrasing next time, it's less kooky. Thanks in advance!

Post by rbanffy » Sat, Jul 05, 2025, 12:29 PM UTC

gsf_emergency_2 wrote: post #14

rbanffy wrote: post #13

Earlier quoted context omitted.

I'd go further - we'll need to rethink the CPU, the memory (we'll need lots of memory regions with exclusive access, a way to manage scratch memory, etc) and IO (because there will be a lot going on at any given time).

(Aside from the scratch or IO maybe this isn't all that heretical :) there seems to be some nascent interest and work out there: https://en.wikipedia.org/wiki/In-memory_processing#Applicati... XenevaOS could at least have moved in that direction..