if you want to see an interesting example of the kind of machine we had in mind for NIX: https://www.esperanto.ai/technology/
4 superscalar risc-v cores, and about 1000 basic scalar cores. Pretty interesting. On Fri, Feb 7, 2025 at 9:21 AM ron minnich <rminn...@gmail.com> wrote: > I checked in execac system call support this morning. not tested. > > I am bringing this back to life because I hate to see it lost. 9front is a > good vehicle and, once it is in, keeping it alive as a fork or branch is > doable. > > There are several ideas in here that may be worth keeping. > The idea of "core roles" seems useful, beyond NIX itself. Especially since > for 10 years now we've had differentiated cores on our CPUs. That won't go > away. > 2M pages is an idea whose time may have come > the idea of scheduling procs non-premptively on cores, with no kernel or > interrupts there, is interesting. And, in an age of hundreds of cores on a > cpu, might be worth it. > > my osx command line as of just now > qemu-system-x86_64 -kernel ~/Downloads/9pc64 -cdrom > ~/Downloads/9front-10811.amd64.iso -smp 4 -nographic -serial mon:stdio > -cpu qemu64,+monitor -append console=0 -m 8192 > > webbox is nice. > > I'll test the execac command next week unless someone else does it first. > > > > > On Fri, Feb 7, 2025 at 12:25 AM Eli Cohen <echol...@gmail.com> wrote: > >> I never really understood what this was supposed to be. seems like an >> interesting idea. very asymmetric multi-processor stuff. >> >> what do you all think we should do? >> >> On Thu, Feb 6, 2025, 10:34 PM Cody Holliday <c...@codysse.us> wrote: >> >>> Hey Ron, >>> >>> Looks like you didn't check in the implementation of execac. >>> >>> Happy February :) >>> Cody >>> >>> >>> On 07/02/2025 01:39, ron minnich wrote: >>> >>> and for the last thing for the day: >>> 1. mwait usage is back in >>> 2. original nix bug from 2011 fixed (see commits) >>> 3. execac command from nix is added back >>> >>> ok, somebody, give it a whirl. >>> >>> >>> On Thu, Feb 6, 2025 at 3:59 PM ron minnich <rminn...@gmail.com> wrote: >>> >>>> ok we have nix >>>> >>>> >>>> core 0:Calling testicc ... >>>> core 0:calling core 1... Set testicc to ffffffff80110d87 >>>> core 1:acsched: cpu1: fn 0xffffffff80110d87 >>>> >>>> core 1:called: <1> >>>> >>>> core 1:acsched: cpu1: idle >>>> core 1:acstackok is ok >>>> core 1:m is ffffffff8002b000 >>>> core 1:m->icc is ffffffff80042680 >>>> core 1:m->icc->fn is 0 >>>> core 1:&m->icc->fn ffffffff80042680 >>>> core 1:wait ... >>>> core 0:done >>>> core 0:Done >>>> >>>> I need to get mwait usage back in, I see what was wrong with my qemu >>>> command now. >>>> >>>> I'm pushing that now. >>>> >>>> >>>> >>>> On Thu, Feb 6, 2025 at 3:51 PM ron minnich <rminn...@gmail.com> wrote: >>>> >>>>> Calling testicc ... >>>>> Done >>>>> >>>>> that's the first attempt to call an icc in quite a while. >>>>> >>>>> one weird thing: either qemu does not implement mwait (unlikely) or >>>>> we're not enabling it (likely) >>>>> >>>>> I'll see. >>>>> >>>>> >>>>> On Thu, Feb 6, 2025 at 3:06 PM ron minnich <rminn...@gmail.com> wrote: >>>>> >>>>>> I'm doing a tiny bit at a time. >>>>>> qemu-system-x86_64 -kernel ~/Downloads/9pc64-3 -cdrom >>>>>> ~/Downloads/9front-10811.amd64.iso -smp 4 >>>>>> starts NIX, and I have set up core 1 as an AC. >>>>>> >>>>>> It panics in acmmuswitch >>>>>> >>>>>> So there's something for someone to chew on. >>>>>> >>>>>> I've been able to ditch a bit of NIX infra that is done better in >>>>>> 9front. >>>>>> >>>>>> >>>>>> On Wed, Feb 5, 2025 at 2:08 PM ron minnich <rminn...@gmail.com> >>>>>> wrote: >>>>>> >>>>>>> FWIW, that boots in qemu. But I've not tested turning on an AC yet. >>>>>>> >>>>>>> >>>>>>> On Wed, Feb 5, 2025 at 11:02 AM ron minnich <rminn...@gmail.com> >>>>>>> wrote: >>>>>>> >>>>>>>> all my work is in https://github.com/rminnich/9front/tree/ron_nix >>>>>>>> >>>>>>>> On Wed, Feb 5, 2025 at 9:05 AM <tlaro...@kergis.com> wrote: >>>>>>>> >>>>>>>>> On Mon, Feb 03, 2025 at 01:40:29PM -0800, ron minnich wrote: >>>>>>>>> > apropos this, I just now pushed a fixup for acmmuswitch that's >>>>>>>>> better. >>>>>>>>> > >>>>>>>>> > we're getting pretty close to 'turn on an AC and watch it burn' >>>>>>>>> point. >>>>>>>>> >>>>>>>>> What is now the master sources? >>>>>>>>> >>>>>>>>> https://github.com/rminnich/nix-os >>>>>>>>> >>>>>>>>> or >>>>>>>>> >>>>>>>>> https://github.com/rminnich/9front/tree/ron_nix >>>>>>>>> >>>>>>>>> ? >>>>>>>>> >>>>>>>>> > >>>>>>>>> > >>>>>>>>> > On Sun, Feb 2, 2025 at 8:55?PM ron minnich <rminn...@gmail.com> >>>>>>>>> wrote: >>>>>>>>> > > from my point of view, what is lacking in NIX is that it's not >>>>>>>>> integrated >>>>>>>>> > > into a living code base, rather it is its own thing. That >>>>>>>>> means it can not >>>>>>>>> > > take advantage of the last 14 years of plan 9 improvements and >>>>>>>>> bug fixes. >>>>>>>>> > > >>>>>>>>> > > Paul has done excellent work pulling it into 9front and I've >>>>>>>>> tried to >>>>>>>>> > > build on it, see >>>>>>>>> https://github.com/rminnich/9front/tree/ron_nix >>>>>>>>> > > >>>>>>>>> > > I am not sure we need a list, maybe we do, but NIX is very, >>>>>>>>> very simple, >>>>>>>>> > > by design, and I think what's more important is to dig into >>>>>>>>> the code and >>>>>>>>> > > get it going. >>>>>>>>> > > >>>>>>>>> > > I think we're within a month or maybe sooner of having the NIX >>>>>>>>> based on >>>>>>>>> > > 9front working as NIX used to work. I'm following the rule >>>>>>>>> that, for every >>>>>>>>> > > change, the kernel must build and boot. I am not at the point >>>>>>>>> where we can >>>>>>>>> > > enable an AC, but we're getting there. >>>>>>>>> > > >>>>>>>>> > > It would be very useful if you can take a look at the code and >>>>>>>>> understand, >>>>>>>>> > > for example, what tcore.c and acore.c are doing, and where >>>>>>>>> I've messed up >>>>>>>>> > > pulling that code over. >>>>>>>>> > > >>>>>>>>> > > Also I'm still not completely understanding how 9 front does >>>>>>>>> startup IPI >>>>>>>>> > > -- we added SIPI in nix, vector 62, but I don't see that in >>>>>>>>> 9front. I added >>>>>>>>> > > VectorSIPI to pc/io.h, but maybe that's a botch. >>>>>>>>> > > >>>>>>>>> > > This means staring at code, and it's a real grind, but there's >>>>>>>>> just about >>>>>>>>> > > no other way to do it. >>>>>>>>> > > >>>>>>>>> > > In particular, the path in acmmuswitch needs more eyes. >>>>>>>>> > > >>>>>>>>> > > On Sun, Feb 2, 2025 at 4:16?PM Cody Holliday < >>>>>>>>> > > cody.parker.holli...@gmail.com> wrote: >>>>>>>>> > > >>>>>>>>> > >> Hello everyone, >>>>>>>>> > >> >>>>>>>>> > >> I created an issue on the GitHub but I figure using the >>>>>>>>> mailing list was >>>>>>>>> > >> more the projects speed. >>>>>>>>> > >> >>>>>>>>> > >> Is there already a list of things to do? Shall we maintain a >>>>>>>>> file to >>>>>>>>> > >> track tasks needing to be completed? >>>>>>>>> > >> >>>>>>>>> > >> I would like to at least be able get a handle on what is >>>>>>>>> still lacking in >>>>>>>>> > >> NIX. >>>>>>>>> > >> >>>>>>>>> > >> Hope you all had a good Groundhog Day! >>>>>>>>> > >> Cody >>>>>>>>> >>>>>>>>> -- >>>>>>>>> Thierry Laronde <tlaronde +AT+ kergis +dot+ com> >>>>>>>>> http://www.kergis.com/ >>>>>>>>> http://kertex.kergis.com/ >>>>>>>>> Key fingerprint = 0FF7 E906 FBAF FE95 FD89 250D 52B1 AE95 6006 >>>>>>>>> F40C >>>>>>>>> >>>>>>>>> ------------------------------------------ >>>>>>>>> 9fans: 9fans >>>>>>>>> Permalink: >>>>>>>>> https://9fans.topicbox.com/groups/9fans/T7e645b96f1f70c95-Mf1b65f60232dce4dcaeba7e9 >>>>>>>>> Delivery options: >>>>>>>>> https://9fans.topicbox.com/groups/9fans/subscription >>>>>>>>> >>> *9fans <https://9fans.topicbox.com/latest>* / 9fans / see discussions >> <https://9fans.topicbox.com/groups/9fans> + participants >> <https://9fans.topicbox.com/groups/9fans/members> + delivery options >> <https://9fans.topicbox.com/groups/9fans/subscription> Permalink >> <https://9fans.topicbox.com/groups/9fans/T7e645b96f1f70c95-M102028ad03288edbadefaed1> >> ------------------------------------------ 9fans: 9fans Permalink: https://9fans.topicbox.com/groups/9fans/T7e645b96f1f70c95-M3bdb5d590d36d70db23503b8 Delivery options: https://9fans.topicbox.com/groups/9fans/subscription
