MINTIA (not vibecoded)

[twilit smelt]

plus you can do like mfcr zero, scratch0 to reset the nmi masking period

or whatever

[raven drift]

rv now has a well defined NMI model

NMIs are masked in NMI handlers

there is a distinct instruction to return from NMI handlers (mnret)

FRED will fix NMIs on x86 by doing pretty much the same (= distinguish between return from NMI and return from interrupt)

[twilit smelt]

@icy bridge basically just don't worry about nmis lol

they aren't used by xrstation for anything functional

[icy bridge]

Yeah but they still don't exist in S-mode right

Like the whole nmi extension only affects M mode

[raven drift]

well, NMIs are only taken into M mode but there is a mechanism in SBI to forward them to S mode (not forward the literal NMI but forward an event that corresponds to an NMI)

it's called SSE

Supervisor Software Events

[icy bridge]

ah

[dense vigil]

@icy bridge btw u probably want CONFIG_SLOB instead of slub

Iirc it's designed for very low memory amounts

[icy bridge]

i'm not doing any config changes until it works with this current one but noted

[dense vigil]

Slub might oom quickly

But not sure

[icy bridge]

eh i can increase mem up to 256M in the emulator

[dense vigil]

Wait really

[icy bridge]

i highly doubt it'll need that much

yeah

[dense vigil]

Why is 256 the max

[icy bridge]

on the default 4M ooms in the bootloader lmao doesn't even get into the kernel

[dense vigil]

Lol

[icy bridge]

limits of the fictional design

each ram bank is placed at a 32M boundary regardless of the capacity of the previous one (even if the previous one has 0 capacity) and there are 8 ram banks

[dense vigil]

Interesting

[icy bridge]

neither of those numbers can be increased since it'd break the firmware's abi:

• the boot info contains a static array whose size is the number of ram banks
• the boot info only contains a total amount of memory and the size of each ram bank, so everything that interfaces with the firmware has the 32M interval thing baked in

[twilit smelt]

yeah and this is all fairly typical for like 1989 workstation

including the weirdly specific firmware interface

they'd just write new firmware

lol

[icy bridge]

luckily neither of those things will be compiled into linux since the bootloader translates the firmware-provided info (including memory map) into a dtb

[dense vigil]

That's good

[icy bridge]

one weird thing that isn't because of an architectural limitation though: the emulator seemingly tries to spread out the given amount of memory evenly over the 8 ram banks

so if you specify like 64M you get 8 banks with 8M each

which might cause problems in the future for linux since even if the total amount of memory is sufficient to load the kernel there might not be enough contiguous memory

[twilit smelt]

Yeah that's for testing purposes to make sure the system software deals with non contiguous physical memory right

[icy bridge]

for now it's fine since the kernel image is ~3.5MiB including bss which is less than the minimum amount of space per memory bank

but if that grows too much it could become an issue very easily

especially considering this is a minimal config of the kernel

i might have to switch out the boot protocol to a paged one in the future for this reason

where the kernel can be noncontiguous physically

[dense vigil]

Which drivers did you enable?

In the config

[icy bridge]

but idk how much linux code (if any) assumes physically contiguity

[icy bridge]

none of them

i started with an empty config and added stuff until it compiled

arch Kconfig: ```

SPDX-License-Identifier: GPL-2.0-only

config XR17032
def_bool y
select GENERIC_ATOMIC64
select THREAD_INFO_IN_TASK
select HAVE_PAGE_SIZE_4KB
select ARCH_SPARSEMEM_ENABLE
select GENERIC_IOREMAP
select FUNCTION_ALIGNMENT_4B
select GENERIC_IRQ_SHOW

config MMU
def_bool y

config GENERIC_CALIBRATE_DELAY
def_bool y

config GENERIC_HWEIGHT
def_bool y

menu "Kernel features"

source "kernel/Kconfig.hz"

endmenu

and defconfig is literally just an empty file

[dense vigil]

Probably doesnt need sparsemem

[icy bridge]

which results in this .config

[icy bridge]

oh yeah that one is just because i wanted to be able to deal with it, i don't think it's enabled by default

[dense vigil]

Since its only 256 mib and the ranges are next to each other the vmemmap can be contiguous one blob ig

[icy bridge]

like architecturally speaking the cpu itself can handle a 4G physical address space

it's just the overall system design of xrstation that limits it to 256M

I know that xrstation is most likely the only configuration of xr17032 that will ever exist but I wanted to be able to deal with theoretical other configs as well

[dense vigil]

A qemu backend could not have this limitation

Theoretically

[icy bridge]

Yeah exactly

Which is actually something I'm interested in exploring

Maybe I'll even port OVMF to xr17032 (probably not)

[dense vigil]

Yeah that would be cool af since it comes with tons of devices

And a jit

[icy bridge]

And gdbstub support which would be very useful considering xremu's builtin debugging is very limited

[night needle]

how does the dtb for this look like

[icy bridge]

it's generated at runtime give me like a minute to dump it

@night needle https://hst.sh/dequruxezu.properties

[night needle]

oh nice

[icy bridge]

this is in the 32M ram 1 cpu configuration

[night needle]

very nice

[mortal thunder]

did you already write all the pte manip code yet

if not, then how come you dont have missing symbols for those

[icy bridge]

it's just a header full of static inline functions

but yes though i am not certain they are fully correct

[dense vigil]

did u use the builtin header which flattens all levels but 2?

other than that it should be just a few inlines yeah

and a vm prot to pte prot or whatever

[icy bridge]

if you mean pgtable-nopmd.h yes

[dense vigil]

probably that

[icy bridge]

annoyingly even with that header you have to define some pmd stuff

[dense vigil]

also whats kinnowfb?

[icy bridge]

it's xrcomputer's video card

it's just an 8-bit static palette framebuffer

[dense vigil]

ah

[dense vigil]

that should be fun to expose as a drm fb

[icy bridge]

love how the register interface for kinnowfb supports 4095x4095 resolution setups

4k on my late 80s workstation

[dense vigil]

1 fps/hour

[pastel flume]

4k ahh 4k

[icy bridge]

although that register is unfortunately readonly

still

[twilit smelt]

concept: guy who says "hi" whenever he opens a discord channel and "bye" before he clicks to another one

@icy bridge still playing silksong

[shadow ridge]

port hollow knight to mintia

[twilit smelt]

hi

bye

[shadow ridge]

hi

[shadow ridge]

ok if i did this, it would spam every single chat

because i just check unread messages in rapid succession before deciding which ones to engage in

[dense vigil]

Did you change your name

[icy bridge]

yes

[ocean valley]

I didn't recognize him until you said that

[shadow ridge]

i also saw the name liam in my DMs and i was like wait who is this

though i didnt bother to check until i messaged in here

and saw it was will

[queen torrent]

@twilit smelt mr hyena man did you spin in your old pushlock implementation before sleeping and if so what was the optimal number of spins

[twilit smelt]

i did

[twilit smelt]

did i spin? hope not or id be dizy

but like no i dont think i did, i do spin in my turnstiles impl

theres a law whose name i forget where if youre spinwaiting for a while to avoid a costly transition (such as blocking and selecting a new thread and a context switch)

its optimal (for some statistical reason) to spinwait for about half the amount of time the transition would take

so figure out how long your full circuit scheduling path tends to take and cut that in half and try to spin that long

[queen torrent]

I see, thanks for your wisdom hyena man

[dense vigil]

This is a Liam

[twilit smelt]

I switched to the second half of william

Wanted to shake things up

[dense vigil]

Ohh

[shadow ridge]

but will has 2 ls

so you have to be iam

not liam

(if you're dropping the will)

[queen torrent]

Gotta have half of the l, 'iam

[mortal thunder]

@twilit smeltspotted in the wild: Windows NT Modified Page Writer car

[twilit smelt]

wel there's a blacked out letter

That could say anything

It could say NT SUCKS IM A HUGE LINUX FAN MPW

Had a dream about working on mintia and woke up feeling very bad I've barely done any work this year

[mortal thunder]

number

[shadow ridge]

sometimes.. i dream about mint

[icy bridge]

@twilit smelt are non-wired tb entries architecturally guaranteed to persist as long as they aren't overwritten by a tb handler or should i treat them as ephemeral at all times unless backed by the page tables

for context i am doing the unpaged-to-paged transition and i was wondering if this was possible by manually adding a non-wired itb entry that identity maps the page containing the jump to the proper virtual addresses

if non-wired tb entries should be treated as emphemeral at all times i would need to back this itb entry in the page tables which i would like to avoid

or just use a wired tb entry ig but that feels icky

[twilit smelt]

if you can guarantee no other tb miss is taken

then theres no reason it would disappear

[icy bridge]

yeah i can guarantee that

thanks

[twilit smelt]

but theres no guarantee about the replacement policy for example

[icy bridge]

i was just wondering if things like speculative execution might mess it up

[twilit smelt]

perhaps

but not in 1989

[icy bridge]

fair enough lol

[icy bridge]

this is what the code in question looks like, and considering interrupts are disabled at the time i'm pretty sure a lack of tb misses is guaranteed

[icy bridge]

speaking of which, i'm curious, on modern software-tb arches how do tb misses interact with speculative execution?

i can't really think of a reliable way to do speculative tb fills with software tb

do they just abort speculation if a speculative tb lookup misses?

[twilit smelt]

yes

[icy bridge]

fair enough

[twilit smelt]

thats one of the reasons software tlb refill disappeared i think

[icy bridge]

yeah thatd make sense

[icy bridge]

btw how's the implementation of adr going? for now i'm perfectly fine without it (the kernel itself doesn't need pic) but once i start doing userspace stuff it'll very much be a necessity

[twilit smelt]

ohhh it's harder than I thought. Beeen wrestling with it for weeks

Jk

I could do it in like 5 min at any time I've just been super busy and when im not busy im tired

[icy bridge]

thats fair yeah

i was in a situation like that for months with my kernel

[dense vigil]

imagine upstreaming the xr port

that would be funny

[icy bridge]

it would

[dense vigil]

like pretend to be some crazy rich croporation with a new arch

[icy bridge]

anyway, update: the kernel successfully executes C code! it now gets to the __builtin_trap() in my setup_arch stub

not much to see for now but hopefully that will change tomorrow because i am going to try and figure out earlyprintk

if everything goes well then i should get the first real linux output because this executes before setup_arch

[azure tinsel]

is there a c compiler for xr?

[icy bridge]

yes, i ported gcc to it

for now it's contained within https://github.com/monkuous/xrlinux

note that while the compiler does work and seems relatively reliable i haven't done any super robust testing (beyond "make stuff with it until a bug is found") and libgcc does not compile yet

[azure tinsel]

cool, might check it out when i'm done hurting clang

[icy bridge]

but on reliability, fwiw it can compile a somewhat simple bootloader i made (which includes memory management, device tree generation, page table creation, and an ext2 driver) without any miscompilations and linux without compile-time crashes (though i haven't verified a lack of miscompilations)

[azure tinsel]

cool

[queen torrent]

hell yeah good work monkuous

[dense vigil]

This probably executes before early param parsing which is the thing that enables earlycon no?

But great work

Or well nothing prevents you from registering a console as the first line I guess

[icy bridge]

Yeah that was my plan

[heady bobcat]

you can do even earlier param parsing before you call start_kernel

[icy bridge]

that was much easier than i thought lol

it's very temporary code and you can tell (it just uses a wired dtb entry with a constant vaddr to get access to the registers so that it doesn't have to rely on early_ioremap)

but it works!

[dense vigil]

lol nice

[twilit smelt]

That's wild

Leenoox

[icy bridge]

does xrcomputer have any timing facilities with sub-millisecond precision? the only time stuff i can find is the rtc which only deals with milliseconds

i don't need interrupts that often ofc but being able to stall for less than a millisecond would be nice

[twilit smelt]

What's the stalling used for

[icy bridge]

just general delay stuff

linux wants udelay

rn i'm just doing a loop-based thing (where the speed of each loop is calibrated during boot) but it'd be nice to base it on an actual timer instead of execution speed

[twilit smelt]

Sadly you'll have to just do that

[icy bridge]

fair

[dense vigil]

i mean old x86 hardware also just used the in instruction for delays or whatever

which guaranteed at least 1 us or something?

[icy bridge]

also means CLOCK_MONOTONIC will just be jiffies so userspace can say goodbye to precise timing ig

[dense vigil]

probably not true for current hw

yeah vdso will be useless as well

[twilit smelt]

ur porting it to something meant to be from a time before Linux existed at all

[icy bridge]

i mean yeah but the ibm pc from 1981 had a microsecond-precise timer

it's not like that level of precision was unheard of in the late 80s

[twilit smelt]

I might have added something when/if mintia needed it

[icy bridge]

some questions about i/o caching and ordering:

• do dcachectrl writes just invalidate modified cache lines or do they also write them back into memory?
  • if they do, how do you ensure that after a noncoherent dma read (device -> memory) the newly read data isn't overwritten by the cache flush?
  • if they don't, how do you ensure that before a noncoherent dma write (memory -> device) the cache lines are written back to main memory?
• can accesses to uncached memory be reordered wrt one another? (aka do you need memory barriers between device register accesses)
• can accesses to uncached memory be reordered wrt normal accesses? (aka do you need memory barriers for coherent dma)

oh and for noncoherent dma, does the dcachectrl write function as a memory barrier or do you need both

[twilit smelt]

The cache is write through with a write buffer

Not writeback

For questions 2 and 3 the answer is no and yes

[icy bridge]

alr ty

and I suppose the memory barriers drain the write buffer

[raven drift]

the first bullet (how do i ensure that newly read data isn't overwritten by the cache flush) is usually not an issue on mainstream archs

[twilit smelt]

yes

[raven drift]

You'd expect a writeback to also clean the cache (i.e., move the cache lines into a state != Modified)

I don't know of any arch that has a writeback-but-keep-dirty instruction

[twilit smelt]

this is because this is how the cache worked on comparable mips workstations in 1989, writeback was more fancy and didnt become ubiquitous until a couple years later

[icy bridge]

how does that prevent overwrite of dma'd data?

[raven drift]

you just writeback+clean before and after the transfer

For example, that's how you need to do it with non coherent memory on x86

[icy bridge]

what if userspace writes to the cache line during the operation (this can happen with e.g. O_DIRECT i believe)

[raven drift]

that's simply not supported

[icy bridge]

fair enough

[twilit smelt]

whats the status of things right now, where are you at @icy bridge

[icy bridge]

haven't done much since getting earlyprintk output, just looking over the various stuff I've put in the headers to see if it's actually consistent with architectural behavior

once that's done I'll start actually filling out the init sequence

[twilit smelt]

should be interesting writing a disk driver and stuff

and interrupt controller support and so on

are you planning on actually supporting SMP

[icy bridge]

probably yeah

[queen torrent]

When xrstation linux works send a message here from it 🔥

[twilit smelt]

xrstation has no networking so that would be tough

i was gonna design a fictional NIC when i got there in mintia

i guess he could do some strange kind of ethernet over serial or something

[icy bridge]

Maybe yeah

also, maybe i'm doing something wrong, but -serialrx and -serialtx seem to be no-ops for some reason?

oh wait it's for serial port b not a

nevermind

[twilit smelt]

they go over the second serial port

yeah

[icy bridge]

that's better

that'll help with ethernet over serial if i get there

9600 bps ethernet sounds like a pain though

[twilit smelt]

i mean its in principle infinitely fast as long as you dont supply -asyncserial (which should probably be called -seriallatency instead or something)

[icy bridge]

ah ok

or I could just add an xr17032 target to qemu (which I am interested in doing anyway) and use some PCI ethernet card

or virtio

not as fun though

[safe hollow]

SLIP is really period accurate too

was already widespread in 1988 according to Wikipedia

[twilit smelt]

are you gonna give it a nice tux bitmap for the boot picker

[sterile frost]

holy shit thats so sexy

[queen torrent]

Fr

[icy bridge]

i should

[sterile frost]

btw have you considered making a separate repo for linux instead of it being a single patch?

[icy bridge]

considered yes but having everything incl. jinx recipes in a single repo makes it easier for now

mostly because I'm effectively porting 3 things at once: binutils, gcc, and linux

[icy bridge]

just to confirm, xr17032 has the same "dependent loads can be reordered" thing as alpha, right?

[twilit smelt]

Unlikely because xr17032 probably wouldn't have the weird dual dcache thing

[icy bridge]

yeah but like architecturally

[twilit smelt]

ill say no

[icy bridge]

alr ty

[night needle]

car

[shadow ridge]

thats a weird lookin dog

if i had the time i would write an SDL backend for mintia

would be interesting to see how it would turn out i think

@twilit smelt xrstation emulator doesnt have an audio device currently right

i think it would be really cool to flex standard frequency audio playback

[twilit smelt]

another thing i wasnt gonna do until i got there in mintia

[shadow ridge]

it could be a good stress test

[twilit smelt]

i was gonna do it double buffered with buffer refill interrupts

2 sample bufr

I WAS GONNA DO IT DOUBLE BUFFERED WITH BUFFER REFILL INTERRUPTS.

2 SAMPLE BUFR.

GOD!!!

[shadow ridge]

slay

by 2 sample you dont mean 2 audio samples right

do you mean 2 channels

i think mintia should support 7.1 surround sound setups

only supporting stereo is some weak unix clone shit

you dont want xr station to be one of those "enables a20 line, loads gdt and idt + page table support!!" projects do you

anyways if it cant mix multiple 44100 hz 24 bit audio files real time, it'll be a very uncool project that is very not cool at all

[twilit smelt]

2 sample buffers

2 (sample buffers)

there are sample buffers

two of them

its not 2 channle

thats per channel

two buffesr per bchahenl

the buffer swould contian liek idk like 0.5 sec ond s of audio each and they woudl then be like yeah

[shadow ridge]

aah

[devout geyser]

what happened to this message 😭

[shadow ridge]

it's still readable lol

somehow

[shadow ridge]

can xr station handle mmio writes that are like ~50 kb per interrupt

[twilit smelt]

it could do like 40 megabytes per second iirc at 20mhz

you can map the sample buffers writecombine like a framebuffer

[shadow ridge]

note that with sound you're likely not playing a single stream

so you need to mix a shit ton of buffers at interrupt time

[twilit smelt]

not at interrupt time

[shadow ridge]

well you could do it async

[twilit smelt]

you can do it ahead of time

[shadow ridge]

and copy from a buffer at interrupt time

[twilit smelt]

yes that

[shadow ridge]

how do you prevent buffer underruns with that method btw

[icy bridge]

wait how do you do wc on xr17032

[twilit smelt]

any cached mapping is sort of like write combine cuz its writethrough

[icy bridge]

oh fair

[mortal thunder]

barely

it ran at 1,193,182 hz

[dense vigil]

@icy bridge btw any progress on the port?

[icy bridge]

not since friday, haven't been working on it since then

[dense vigil]

ah

[icy bridge]

actually I don't think I sent an update then? it currently gets to successfully parsing the dtb

[dense vigil]

damn

where does it die atm

[icy bridge]

at panic("TODO: setup_arch")

[dense vigil]

ah

[icy bridge]

also I'm pretty sure this is the only linux port where the kernel image is not physically contiguous which is interesting

it used to be but I changed the boot protocol due to the xremu ram spreading behavior

[shadow ridge]

ram spreading??

[icy bridge]

it spreads the given ram size over all banks (with at least 4M per bank)

so if you specify 32M ram you get 8 banks of 4M

[shadow ridge]

ah

[icy bridge]

which is problematic because each bank's starting offset is determined only by its index

so you get a bunch of small regions with lots of space between them instead of one big contiguous region

[shadow ridge]

what is the coefficient

[icy bridge]

32M

[shadow ridge]

so you get 8 4M banks with 28M padding?

wtf

[icy bridge]

yep

and considering an extremely barebones nonfunctional version of linux already has a ~3.6M kernel image I figured making the kernel image physically noncontiguous was important

right now the boot protocol involves the bootloader creating page tables and passing them to the kernel without enabling mmu itself

so the kernel is free to set up the tlb however it wants before enabling paging

but this does mean that the pre-paging part of the entrypoint isn't allowed to cross a page boundary (right now I just use the first 256 bytes of the exception block for this)

[shadow ridge]

cant believe we're getting linux on limnstation before mintia 2

[icy bridge]

https://github.com/monkuous/xrlinux/blob/main/bootloader%2FPROTOCOL.md

[twilit smelt]

I'm slow.

[queen torrent]

hi slow

[twilit smelt]

asshole.

[queen torrent]

where I dont see any

[shadow ridge]

liamstation

[twilit smelt]

kdebug> threads
            name  thid  team cpu state     sem
   idle thread 1     1     2   0 rdy
   psycho_killer     2     2   0 rcv
   kernel_daemon     3     2   0 slp
    1 idle angel     5     2   0 wai debug idle thread 1(29)
         dprintf     6     2   0 slp
           dgets     7     2   0 wai request for input(48)
              sh     9     3   0 RUN

report of the threads command on BeOS's built in kernel debugger

i like the psycho killer thread and the kernel demon and idle angel

[mortal thunder]

They were having fun

[devout geyser]

lol

[twilit smelt]

@icy bridge this might be a pain but do you think you can write out a list of all the stuff you pointed out to me

For example the mistakes in the book, the changes in jalr signedness, and ADR

I think one reason I haven't gotten around to any of this is laziness about getting all these things together

[icy bridge]

• docs+emu: ADR instruction (ADR RA, IMM16, Reg[RA] = PC + (IMM16 << 16), immediate operate format, opcode = 0x30, RB field is ignored or MBZ)
• docs+emu: JALR instruction should zero-extend instead of sign-extend
• docs: instruction summary for LSH, RSH, ASH, ROR list registers in incorrect order (RA, RB, RC instead of RA, RC, RB - detailed description lists them correctly)
• docs: LSH/RSH/ASH/ROR should mention that operand is masked with 31
• docs: MTCR pseudocode names incorrect source register (should be RB instead of RA)
• docs: MB and WMB descriptions should probably be rewritten to match #1322494605819908156 message
• docs: PAUSE function code is listed incorrectly in hexadecimal (but binary is correct)
  iirc this is everything but i might be forgetting some stuff

[dense vigil]

Btw any new progress?

[icy bridge]

been too busy to work on it

[twilit smelt]

Thank you very much

Embarrassingly huge number of docs errata

[twilit smelt]

all done finally

it was indeed helpful to put it all in one spot, thank you

[dense vigil]

Did you add the instruction monkuous was asking for?

[twilit smelt]

yes

static void XrExecuteAdr(XrProcessor *proc, XrIblock *block, XrCachedInst *inst) {
    proc->Reg[inst->Imm8_1] = inst->Imm32_1;
    XR_NEXT();
}

static XrCachedInst *XrDecodeAdr(XrProcessor* proc, XrCachedInst *inst, uint32_t ir, uint32_t pc) {
    inst->Func = &XrExecuteAdr;
    inst->Imm8_1 = (ir >> 6) & 31;
    inst->Imm32_1 = pc + (ir & 0xFFFF0000);
    return inst + 1;
}

very shrimple

[dense vigil]

nice

[twilit smelt]

Not
Interesting or
Cool
End your project now

thats pretty mean

[dense vigil]

ill say that when xr gets acpi support

[twilit smelt]

it wont but mintia2 for amd64 will use uacpi

[dense vigil]

interesting and cool then

[twilit smelt]

in like 2027 probably

[dense vigil]

but a madt extension for xr would be funny too

[twilit smelt]

wait 2027 is actually pretty close now

thats been my go-to "impossibly far off" year for like 10 years

[dense vigil]

yeah

lmao

[twilit smelt]

im 23 and i already feel like my youth is far behind me and my life is coming to a close

my mental age is like 60

[mortal thunder]

my god it's less than 2 yrs away

[twilit smelt]

someone once called me "the oldest young person [they] ever met"

and that was like 4 years ago when my mental age was only 40

[dense vigil]

thats pretty offensive

[twilit smelt]

i mean

it was true

they were saying that primarily cuz i was a very long time member of a community they were relatively newer in and i had stories from like 7 years earlier that id tell like a village elder with the kids gathered around the campfire despite being 19 and most of the newer people being much older than me lol

[dense vigil]

lol

23 on this server is very old

but im even older

[twilit smelt]

cant believe i started mintia when i was 18

[dense vigil]

most people are like 15

here

[twilit smelt]

when i was 18 i was suffering about how long mintia kernel was taking before anything interesting would happen

im now 23

sufferinga bout how long mintia2 kernel is taking before anything interesting will happen

[dense vigil]

wait u started mintia2 at 18?

[mortal thunder]

mintia 1

[dense vigil]

that was old mintia right

[twilit smelt]

yeah

old mintia was 2020-2023

almost the entirety of 2023 and first part of 2024 was spent on the new toolchain

mintia2 started in summer 2024

[dense vigil]

2020 feels like it was a year ago lmao

[twilit smelt]

the new toolchain did NOT need to take that long

i got seriously demotivated

i still am

i have yet to recover the motivation levels i had during old mintia

[dense vigil]

yeah i can see that

[twilit smelt]

at the final old mintia pace i think mintia2 kernel would be literally done by now

since i already know how to do everything this time

it would have gone much faster

[dense vigil]

for old mintia u would have 12 hour sessions where u would implement 999 things

[twilit smelt]

i had like an all consuming hyperfixation with it which is no longer present

[dense vigil]

thats less healthy i guess

[twilit smelt]

no i liked it better when i had the hyperfixation

[mortal thunder]

i kinda get this hyperfixation everytime i start a new project

[twilit smelt]

i liked when every few weeks id run stress tests and spend like 10 hours straight fixing like 40+ bugs

[mortal thunder]

which means i can get a new project "released" in about 2 months

[twilit smelt]

old mintia still has no known bugs

[mortal thunder]

thats very cool

[twilit smelt]

"known bug" was not in my vocabulary

if there was the slightest whiff of a bug i would drop everything and fix it

mintia2 has been the same so far to the extent possible with a like 20% done kernel

i think thats the only way to do it if youre a one-man team working on something of that scale

if you allow a "bug backlog" to build up itll collapse under its own weight pretty fast

[twilit smelt]

i got back to this a little bit inbetween summer and fall semester when i spent like 12 hours a day for 2 weeks straight speeding up the emulator

so ive confirmed my brain is still capable of entering that mode lol

[raven drift]

You pretty much have to accept the existence of bugs when you run on a diverse set of real hardware though

[mortal thunder]

its capable but the thing you're working on has to be "new", at least in my case

or "important enough"

[twilit smelt]

only with code that meshes with the hardware

i disagree with generic code

like scheduling, filesystems, etc

most vmm logic

[raven drift]

well yeah

[mortal thunder]

i think those are also fixable

[twilit smelt]

i think korona is saying theres always gonna be some weird hardware that your driver doesnt quite work well with even though you followed the spec perfectly

[dense vigil]

only if you can set up ci for every existing piece of hardware

only then u can know that it works everywhere

[mortal thunder]

no?

[dense vigil]

yeah thats how hardware is

[mortal thunder]

if you encounter a bug in real hardware that is replicable you can fix it

if it's not replicable then it may stick around for a while, sure

but still fixable

[dense vigil]

regressions are a thing

[mortal thunder]

that's just another bug

or a bug resurfacing

[dense vigil]

hence the need for CI

[mortal thunder]

no

[dense vigil]

otherwise u have no way to prove that a given piece of code works everywhere

[mortal thunder]

sure

but those bugs arent "known bugs" 😛

if you dont know of a bug it's not a "known bug"

[dense vigil]

works on my machine

[mortal thunder]

exactly

[mortal thunder]

also it could work on your CI machine but not someone else's

[dense vigil]

that would mean the hardware is broken

either mine or theirs

[twilit smelt]

wonder how many man hours are in my project

i think it could be 10k+

[dense vigil]

what if you count the time u spent thinking about it

[twilit smelt]

then it becomes the majority of my waking moments for the last 7 years

[twilit smelt]

iykyk

[mortal thunder]

an incomprehensible amount of obsidian

how was it mined? it takes like 7 seconds to break one piece of obsidian with a normal diamond pickaxe and probably still like 3 with an enchanted netherite pickaxe and haste

[acoustic sparrow]

nothing is truly incomprehensible when using //cyl /j

[twilit smelt]

not what they did

that was survival

lol

[acoustic sparrow]

yeah obviously lol

[mortal thunder]

so my guess was right

[acoustic sparrow]

but yk

[mortal thunder]

oh even with haste 2 and an efficiency 5 netherite pickaxe it still takes 1.6s to mine

must have a lot of patience

[acoustic sparrow]

you dont have to mine the obsidian though?

[mortal thunder]

to place it?

[acoustic sparrow]

yeah

you can create it in place

[twilit smelt]

yeah you do, it was botted

but placed by hand

[acoustic sparrow]

ah okay

[mortal thunder]

well you could use lava buckets instead but buckets are not stackable so you need to pick up some lava, then swap to an empty bucket, pick up some more lava, etc

[acoustic sparrow]

ah yeah

right lol

im stupid

[mortal thunder]

and you can only have a couple thousand lava buckets at most

thats a damn big ocean

[twilit smelt]

been doing lots of high res renders lately

theyre fun

this one is normous

in particular

open in browser or download it

[raven drift]

At some point the "you" in this sentence becomes a problem. It's too much work for a single person

Sure, it's theoretically possible but not in practice

[mortal thunder]

sure

[raven drift]

We were talking about keeping the system bug free in practice though, not only in theory

[icy bridge]

nice, thank you!

[queen torrent]

Wait wtf how old are you

[dense vigil]

26

[queen torrent]

Damn

[dense vigil]

Yes

[teal trench]

old man

[mortal thunder]

https://bsky.app/profile/invoxiplaygames.uk/post/3lza6iqn5d22w

Emma (IPG) (@invoxiplaygames.uk)

perfect really is the enemy of the good. procrastinating too long getting certain stuff done on projects cause i spent too long obsessing over making them "perfect". realised they don't have to be perfect, sat down and wrote some code, and now they exist when they previously did not

Bluesky · Sep 20, 2025, 1:00 AM

[tiny swallow]

is there any sign of dwl's whereabouts

[dense vigil]

Turkey

[tiny swallow]

[shadow ridge]

they're probably riding around on their camel and trading pottery or some shit

[pastel flume]

I feel like this may be slightly racist

[dense vigil]

Very

[pastel flume]

Yeah ….

Is that allowed here or

[shadow ridge]

it's ok it's only 14% racist

the rules say you cant go above 15%

[dense vigil]

This person is Turkish

[shadow ridge]

ewww whaaaat

[pastel flume]

Ahhh

In that case I hate Bosnians!

Balija fucks

Cevapi munching

[shadow ridge]

imagine living in the balkans and caring about racism

this is what american influence does to people

[dense vigil]

He lives in America iirc lol

[shadow ridge]

my condolences

[pastel flume]

I don’t care about racism

I’m Balkan

I just know that shit don’t always fly around here

[shadow ridge]

dont worry most westoids only care about shit like racism/bigotry if it's against the minorities they approve of

[pastel flume]

To je smiješno

[pastel flume]

What are these minorities

Westoid is funny

[shadow ridge]

idk these days antisemitism seems to fly more cuz of the palestine stuff

[shadow ridge]

wellll maybe not cuz people have complained about random balkan drama in the past

[pastel flume]

That’s why I avoid Balkan farts in this server

Balkan culture is insane

[earnest zenith]

fr

this is literally the Nazi's wet dream

[golden meteor]

antisemitism has been alive and well for a while, well before the most recent wave of palestine problems

[earnest zenith]

where people can say "Israelis are occupying our government" without fear

[golden meteor]

but most pro-palestine people are anti-israel/israeli

at the more common extreme at least

[earnest zenith]

like

uh

Candance Owens

Anna Kasparian

Nick Fuentes

[golden meteor]

idk probably, she says a bunch of dumb shit so I tend not to listen to her

[earnest zenith]

and she is (nominally) left btw

and is parroting ZOG and "Israel did 9/11"

[earnest zenith]

ah yes, and Pax Tube

another left-wing antisemite is GDF

[twilit smelt]

i think theyre referring to the insane level of unconditional deference shown to israel no matter what they do

[golden meteor]

sometimes, but it's also common that they literally think there is a shadow government of israelis controlling every world government

[twilit smelt]

youre crossing the line from criticizing antisemitism into claiming a foreign state cant acquire any influence over another's government which is false

like aipac exists this stuff isnt even secret or a conspiracy theory

its done in the open and called good

[earnest zenith]

we're not talking about any influence, we're talking about literal occupation of the government

[twilit smelt]

also i understand there are antisemitic connotations to ZOG bc the term was in use by the alt right for a long time

but it is denotationally true that america's government seats are occupied by zionists and thats probably why leftists started saying it

the only mainstream political opinions in the usa that you can possibly be elected under contain zionism

this is just true

[earnest zenith]

still wouldn't explain the 9/11 shit

https://x.com/AnaKasparian/status/1958648473533939769

Ana Kasparian (@AnaKasparian)

It is antisemitic to acknowledge FBI records. So please, in the same of protecting Israel, DO NOT look for or read what our own FBI reported.

X · Aug 21, 2025, 9:52 PM

She also said: "Our country is occupied and controlled by the Israelis."

note not

zionist politicians

but THE Israelis

[twilit smelt]

its like inherently radical and islamic terrorist marxist vibes to do anything other than total deference to israel for some reason no matter which side of the aisle youre on

which is weird and bad

theres been some total israeli PR victory among the older generations

[golden meteor]

what makes that message antisemitic?

[twilit smelt]

younger generations on both sides are increasingly questioning the state of israel though so this might change in the coming decades

[earnest zenith]

see the whole thread

she's basically claiming that Israeli nationals somehow knew about 9/11 ahead of time

and were celebrating

[twilit smelt]

think they screwed themselves over by acting like whiny little professional victim bitches without realizing we only like that when we're doing it about ourselves and when someone else does it its just repulsive

lol

[earnest zenith]

see?

[twilit smelt]

its just braindead to imagine that "israeli nationals" as a whole would have special foreknowledge lol

secret israeli info distribution network where their millions of citizens receive special knowledge about mossad ops and they all stay quiet about it

somewhat insane anybody could believe that

it doesnt stand up to 2 seconds of thinking

[earnest zenith]

yup

And why would she think that if not antisemitism or trying to signal to far-right people?

[twilit smelt]

if you want to point out individual israelis who were celebrating 9/11 after the fact and use that to say "maybe they arent such good friends of ours after all"

thats different

[twilit smelt]

idc about that woman

she may be antisemitic

[glass sparrow]

kasparian is like a grifter

not much interesting about her

[earnest zenith]

TYT is so weird

they produced Cenk, Ana, Hasan, and Dave Rubin

[glass sparrow]

TYT didnt produce as much as the money that people paid to their pundents

[twilit smelt]

@glass sparrow look at this transition i got. i screern recorded it because i thought the songs went well together

thoughts?

[glass sparrow]

banger

[twilit smelt]

@earnest zenith at least youre not a "antizionism is inherently equivalent to antisemitism" person

[earnest zenith]

Yeah

[glass sparrow]

no she is not

she grifted away

[glass sparrow]

this depends a lot on what you define as anti-israel, if I had a wayback machine and infinite power id probably not made israel

not because I hate jews or anything, I think jewish people are really cool

it just seems like a bad colonial project that we cannot pull out of

[twilit smelt]

there was no real reason for it to exist to begin with

and one of the most dangerous mistakes humanity has ever made especially was allowing that weird evil apartheid state to get nukes :((((

if ever interventionism was justified it would have been to blow up the reactor at dimona like jfk wanted

sadly many average americans are in a religious hysteria about israel and think its actually a command from god that we unconditionally support them

[glass sparrow]

I mean my personal take is there should be a kilometer wide line of blue helmets between every IDF soldier and every palestinian like we went into bosnia for way less, its pretty interventionalist

and idk we can see what to do after that

not that theres gonna be much palestinian to put behind the km wide line at this rate if we take much more

[twilit smelt]

ideally IDF would be disbanded and its commanders put on trial for genocide and hanged (including netanyahu and ben gvir and others)

government of israel dissolved and put under temporary UN control

[twilit smelt]

also the ones posting gleefully about stealing wedding dresses from bombed out houses and doing mock weddings and putting that on tiktok because they think theyre safe from ever being prosecuted should probably be tried too

stuff like that

[glass sparrow]

idk I kinda feel thats second degree priorities, but I second the sentiment

first degree priorities being you know stopping it

[twilit smelt]

if ireland could borrow the entire american navy for 10 days this is probably what theyd do

what i described

lol

[glass sparrow]

this is what we need to facilitate

just do 1800s level corsair politics and have 3 aircraft carriers lower the US flag for no flag

[earnest zenith]

Yeah, definitely
But we now live in a reality where Israel is probably here to stay for a long time

[maiden lance]

okay but how is AIPAC not being registered as a foreign lobby and actively funding big money to US politicians not "running the country"

like explain how the US, EU and most EU members can look at the genocide happening before our eyes and not actively oppose Israel if they are not dependent on them be it politically or financially or whatever

also on the question of Israeli nationals vs zionist politicians the zionist machine has done good and hard work to convince its people that all Palestinians from young to old and all neighboring Muslim citizens want to kill them

this is all to establish support for an apartheid regime and to one day realize the dream of great israel a la nazi germany

[earnest zenith]

they haven't actively opposed any genocide in a while

simple: it's not in their national/political interest

[maiden lance]

Rwandan Genocide?

Darfur conflict?

wrong reply

[earnest zenith]

What did they actually do there actively?

[maiden lance]

send troops

intelligence operations

for the "good" side

not sure if they've done sanctions but that's often the first thing thing that happenes

[earnest zenith]

Are you referring to UNAMIR?

[maiden lance]

yeah

and french troops

well the french troops helped the genocide but

it was active involvement

[earnest zenith]

UNAMIR was deployed before the genocide and didn't do shit

And as you said, the French did the opposite of stopping it

[maiden lance]

it is a shame that the UN has a big history of failing to prevent genocides despite active deployment

in the case of Israel UN members are just letting them do their colonial dirty work passively supporting with intelligence and arms

[maiden lance]

I don't get this religious hysteria when Trump but more importantly Peter Thiel, and co. are eerily accurate representations of the antichrist

and as if Jesus would have ever stood for any zionist policy

[earnest zenith]

Afaik it's about where the money comes from, which supposedly is mostly wealthy private US citizens

[maiden lance]

anyway these last 10+ messages will get me interrogated if I ever visit a US airport

[earnest zenith]

And anyway, that's incredibly far from running the country

There are tons of lobbies

The gulf states could probably massively outspend the pro-Israel lobby on that issue if they wanted to

[maiden lance]

fuck all lobbies

I don't believe Israel runs the US either

just that Trump is the least independent president in a long time

[tiny swallow]

aint no way dwl's whereabouts triggered this convo

[mortal thunder]

Trump is the least president in a long time

Like he does absolutely nothing about foreign threats

And basically other people do the work for him like P. Bondi, S. Miller, K. Patel etc

[mortal thunder]

there is no religious argument, the bible states that the jews were left without a country because of what that guy who was the brother of the father of muslims did

i forgot his name

isaac i think

nvm ignore this

[maiden lance]

lol ain't Patel honeytrapped tho?

[warm mural]

If Jesus went to the US he would get deported

[twilit smelt]

ok I think that's enough of that. @icy bridge please linuxpost more.

[twilit smelt]

does anyone even remember what i was supposed to work on next

i still dont have time to but it would be nice to be reminded.

[shadow ridge]

docs+emu: ADR instruction (ADR RA, IMM16, Reg[RA] = PC + (IMM16 << 16), immediate operate format, opcode = 0x30, RB field is ignored or MBZ)
docs+emu: JALR instruction should zero-extend instead of sign-extend
docs: instruction summary for LSH, RSH, ASH, ROR list registers in incorrect order (RA, RB, RC instead of RA, RC, RB - detailed description lists them correctly)
docs: LSH/RSH/ASH/ROR should mention that operand is masked with 31
docs: MTCR pseudocode names incorrect source register (should be RB instead of RA)
docs: MB and WMB descriptions should probably be rewritten to match ⁠progress-report⁠MINTIA
docs: PAUSE function code is listed incorrectly in hexadecimal (but binary is correct)
iirc this is everything but i might be forgetting some stuff

[twilit smelt]

finished all that

[shadow ridge]

wtf

dont have the time my ass

[twilit smelt]

all of that took me like 1 hour

total

lmfao

if you actually read it youll see its very shrimple tasks

correcting a few typos and adding a new instruction to the emulator and toolchain

[twilit smelt]

ohy eah

i am supposed to

have the bootloader load all of the driver config files

so they can be used to initialize the IO catalog in the kernel with a set of all known drivers present at boot time

which can then be used to dynamically load the needed ones

im supposed to be doing something else but i guess ill just get that over with rn

oh god thats not a good sign

im stumbling over jackal syntax a bit

I MADE THE LANGUAGE!

[shadow ridge]

did you make syntax / semantics docs

i have a relatively incomplete one for my language but if i need syntax help i can look at it

[twilit smelt]

no bc thatd take effort and people who already know C seem good at figuring it out by just reading whats already there

@warm root is rewriting fox32os's file browser in jackal rn

[shadow ridge]

if this were the 80s, how good would this project be overall compared to competition

(in its finished state)

[twilit smelt]

unrealistically good

if you mean mintia2

[shadow ridge]

yea

[golden meteor]

you should port it to alpha....

[twilit smelt]

dont wanna

alpha didnt have 8 or 16 bit loads and stores

[shadow ridge]

add virtualization instructions and run mintia within mintia

[twilit smelt]

so compilers would synthesize 8 and 16 bit stores from 32 bit read-modify-writes

which made them silently no longer atomic

requiring extra locks in places no other architecture needed them

[shadow ridge]

wait arent you doing most atomic ops on 32 bits anyways

[twilit smelt]

no i mean like

just storing a byte

is now racey

just by compiling a program for alpha

racey with all other bytes in the same aligned 32 bits

like it doesnt even need to be something youd usually consider explicitly "atomic"

you might have a packed set of booleans which are each 1 byte and are modified under different locks

thats now broken

[twilit smelt]

i think fastfat still has one of these places lol

the sorta open source windows fat driver from microsoft

yep

https://github.com/microsoft/Windows-driver-samples/blob/main/filesys/fastfat/allocsup.c#L4050-L4060

that lock is taken only on alpha because that 16 bit write is racey only on alpha

thats so scary they kept it around decades after they dropped alpha support just in case it ever rises from the dead

            //
            //  We need extra synchronization here for broken architectures
            //  like the ALPHA that don't support atomic 16 bit writes.
            //

lmao i hadnt noticed that

they call it broken

[shadow ridge]

wait until sigma rolls along and has a native word size of 37

[twilit smelt]

alpha actually added 8 and 16 bit memory accesses later but like

it was an extension that was only implemented in the very last chips

which there are almost none of and which arent emulated even in commercial emulators ppl use to run old versions of VMS and stuff iirc

interestingly we have an early build of Windows XP for DEC Alpha

and it uses the new 8 and 16 bit access instructions

[shadow ridge]

didnt u get that one running or soemthin

or was that another project

i remember you getting some early windows build running on an emulator

[twilit smelt]

so they actually added support for them to their toolchain during the interim period between the last public release for Alpha and internally dropping support for it

meaning they were serious about Alpha development

they didnt drop Alpha because they wanted to either. Compaq bought it and unilaterally ended its development

it was still seen as a viable workstation and server architecture til they did that

and MS was fully committed to supporting it for at least a few more years

[twilit smelt]

that was an extremely early NT build from april 1991 that ran on the MIPS R3000

still the earliest build we've yet seen

its so early it has a "balance set" like VMS in the kernel still

by the next one we have (september 1991) thats gone

[shadow ridge]

maybe microsoft has some tapes lying around somewhere that contain an earlier version

[twilit smelt]

the released versions of NT for MIPS ran on the 64-bit MIPS architecture (R4000 and up) and basically pretended the upper 32 bits didnt exist

so this fully 32 bit version was unique

NT ran on R3000 internally but that port was dropped before release

[twilit smelt]

they do yes

someone probably still has an i860 build of the kernel on a floppy disk in their attic somewhere

i have extensively emailed around and stuff trying to track these things down

i860 NT is my lost backyardigans pilot

[shadow ridge]

wtf is a backyardigans

[golden meteor]

a kids show from a while back

like teletubbies

well i guess its not that old

[twilit smelt]

if it ever drops NT/860 on the internet or i otherwise find someone willing to share a copy i hope it doesnt happen in the middle of a semester because im going to forget school exists for a full 2 weeks while i spend 12 hours a day writing a Dazzle emulator to get it to boot

thats what happened with the R3000 NT build and that wasnt even half as exciting

i remember after a full week of working on it i was suddenly like "wait... im in college...... uh oh"

[golden meteor]

where can I find information about it? docs, blog posts, books, etc

[twilit smelt]

about what

[golden meteor]

i860 nt

[twilit smelt]

i mean its just NT but really early and for the intel i860

kind of explains itself

[golden meteor]

yeah but I mean, do we have any info on nt from that time period?

[twilit smelt]

some

mostly from my labors

[golden meteor]

i assume its earlier than the r3k port right?

[twilit smelt]

yes

[shadow ridge]

if you miss that many classes wont you get booted off

[twilit smelt]

no

i havent gone to a real analysis lecture since the second one which was like 4 weeks ago now

the midterm is on friday

im still in it

unfortunately

[shadow ridge]

hahahahaha

go to the complex analysis lecture instead

[twilit smelt]

dont think thatll help

[shadow ridge]

why not

[twilit smelt]

the rough timeline as ive gathered it is:

summer 1988:

• the PRISM architecture is cancelled at DEC
• Dave Cutler leaves DEC and takes most of the PRISM and Mica OS team with him to try to create a new workstation startup
• intel is about to come out with the RISC i860 and wants to make it the PC platform of the future, to replace elderly CISC x86. they call up Microsoft to try to arrange that a portable OS be prepared on their end that will be able to run on the new i860 PC platform
• a project at microsoft called Psycho begins, headed up by Nathan Myhrvold, intending to research the feasibility of using Mach as a new kernel for a portable version of OS/2. the "Portable Systems Group" (PSG) is founded

fall 1988:

• Myhrvold hears that Cutler is on the market and convinces Bill Gates to call him up and try to get him to join the PSG (Cutler was famous as a hyper competent OS guy due to his work on VMS). Gates does and Cutler isn't initially interested in PC software.
• Ballmer takes Cutler to breakfast at Denny's and manages to change his mind.
• Cutler's team settles in at the PSG at Microsoft with Cutler taking the role of director
• design begins for the NT kernel, taking heavily after a flattened version of Mica
• an i860 toolchain distributed by intel (pictured below) is acquired, containing a simulator, runtime library, compiler (written by Metaware), and other tools

march 1989:

• NT is being written for a simulator using a convoluted build process ive described elsewhere
• first code written for NT, in the Ke subcomponent (exception handling & the scheduler). Ke is almost completely finished by the end of march

mid 1989:

• first Dazzle and Frazzle (i860 server and PC reference platform, respectively) boards are manufactured by Olivetti and development switches to these

late 1989:

• problems with the i860 are apparent by now
• port to a MIPS R3000-based version of the Dazzle board known as "Jazz" begins

early 1990:

• port to i386 begins
• i860 port is dropped

mid 1990:

• NT is ported to new R4000-based Jazz

late 1990:

• direction is officially switched from "NT OS/2" to "Windows NT"
• port to R3000-based DECstation 5000 begins, worked on by Alexandre Bronstein

late 1991-early 1992?:

• R3000 port dropped along with DECstation 5000 port

early 1992:

• DEC Alpha port begins

alexandre bronstein demos the DECstation 5000 port to Bill Gates and Dave Cutler:

i860 Dazzle development board:

4 tapes of NT's design being presented by Cutler and others to DEC engineers who are about to work on NT's codebase (including Alexandre Bronstein):
https://www.youtube.com/watch?v=UVKKhF9pGpI
https://www.youtube.com/watch?v=mtF-X1YnIDg
https://www.youtube.com/watch?v=19I2ffadc3Y
https://www.youtube.com/watch?v=L5ODGQdZZhA

tape of NT's design being presented by Alexandre Bronstein at DEC (includes a demo of the never-released DECstation 5000 port):
https://www.youtube.com/watch?v=bMJIhN4xCro

yes im autistic.

@golden meteor this is the broad strokes of what ive learned about early NT over the last few years

[golden meteor]

awesome, I'll be reviewing all of that Shortly™

and anything else you feel like sending

ideally related to early NT and the platforms it ran on, but if you just feel like talking about newer NT thats fine too

[twilit smelt]

i got an autistic fixation on NT because its early history is underappreciated and shockingly fascinating

i think what first caught my eye about it was like

hang on this was a brand new OS being written for a brand new platform using an incredibly shitty buggy simulator and toolchain. thats kinda like what im doing.

nobody ever talked about NT so i assumed it was boring and stupid and lame and then one day i got curious and googled it and i was like WOW! THIS ISNT BORING, STUPID, OR LAME!

[twilit smelt]

also the like

5000 coincidences had to align at the same moment in late 1988

aspect of it

it is one of the unlikeliest things to exist

feels a little homeric

or preordained...

[cunning mortar]

https://computerparkitecture.substack.com/p/the-long-mode-chronicles

idk maybe you find this interesting

[twilit smelt]

ai generated cover image is sus

makes me suspect the entire thing is gonna be ai

[cunning mortar]

techie people generate images with AI

it just is what it is at this point

[twilit smelt]

this is the paper that blog post cites

since its now behind some weird intranet login page

THEY CITED WIKIPEDIA!!!!

[cunning mortar]

death

[shadow ridge]

kill.

[twilit smelt]

i dont think the "dont cite wikipedia" thing was common at the time this was written (2006)

i think people citing wikipedia was what led to "dont cite wikipedia" rather than that always having been the attitude

so thats probably why they boldly cited wikipedia

it wasnt seen as a problem

yet

[twilit smelt]

anyone know how people tend to implement speedy userspace condvars

given that userspace mutexes are usually managed entirely in userspace when theres no contention

using atomics

and youll want to pass this to your condvar for release that is atomic with waiting

[raven drift]

When wikipedia was new, its content was seen as unreliable

[acoustic sparrow]

futex has FUTEX_REQUEUE, which may help?

another pretty reasonable solution is to implement your primitives on top of a "add an unpark token to thread X, if they don't have on already" / "consume my unpark token, possibly blocking to do so" pair of primitives

[raven drift]

Aka a binary semaphore

[twilit smelt]

boom https://github.com/xrarch/mintia2/commit/0d215bbd80e5388ceacacc99aecde9ec726afe98

bootloader loads all driver config files in preparation for initializing the IO catalog later in the kernel

thats the next thing ill have to work on

omg mintia2 io system real???

[night needle]

fake

[twilit smelt]

2026 will be the year of the mintia2 desktop

if im not mounting a filesystem by january 2026 i will drop out of uni and become homeless

[twilit smelt]

sometimes feels incredibly bleak that i started old mintia when i was 18 and im now 23 and i still cant mount a filesystem. i mean i could like 4 years ago. but now i cant again

this shit is taking toooooooo looooooong

wasted almost the entirety of 2025

in 9 months in 2025 there have been 193 commits

in 6 months in 2024 there were 347 commits

in 2023 there were 625 commits on old mintia

i considered that a very inactive year too

because i had stopped working on it in favor of the jackal toolchain

thats gnarly

ive lost the juice

[shadow ridge]

progress is NOT going according to the june plan

[night needle]

if you graduate and go into an IT job you will also be homeless

[night needle]

btw in this image i can see XR/computer XR/station and XR/system

what's the difference

[tame phoenix]

XR/server/320 when?

😄

[twilit smelt]

XR/computer is the broader underlying platform architecture

XR/station is a particular model implementing it

There is no "XR/system" that's XrSystem which is the mintia2 ISD (integral system driver) for XR/computer, basically the module that contains all of the drivers for devices that are invariant (ie the chipset)

There will also be FoxSystem

And I suppose PcSystem

later

[twilit smelt]

Also probably the thing most reminiscent of NT about my project is the obscene miscalculations of how long everything will take

I have a message in my discord from like April 2023 where I estimate the new toolchain will be done and mintia rewritten by the end of the year

toolchain alone wasn't even done by the end of the year

Finished that in February of the next year or so

After taking my sweet ass time on it (that was another period of very low motivation)

Thing is in hindsight this was impossible but I literally could have had the toolchain and mintia2 DONE by the beginning of THIS year if I'd kept a pace consistent with work on old mintia the entire time

Could have finished the toolchain in probably 4-6 months and mintia2 in another year and a half

Easy

Annoying to look back on

[shadow ridge]

i dont really regret not doing shit when i had negative motivation levels

because you literally cannot do something unless you have motivation

[golden meteor]

holding you to this

[twilit smelt]

Also I predicted it could kill the project

I knew there were gonna be huge motivation issues

Here's another prediction where there's parity with old mintia by January of this year

This was doable if I worked on it manically like I thought I would

[warm pine]

do you think yourself more productive in jakla at least?

vs dragonfruit

[craggy spire]

Can you though just in theory be more productive with dragonfruit

[twilit smelt]

No

I got Stockholm syndrome toward dragonfruit

I miss it

Jackal is too sterile

the zaniness is gone

[blissful smelt]

jackal 2.0 now messier than ever with direct df backwards compatibility

[queen torrent]

create a variant of C ez

[twilit smelt]

Jackal basically already is this

[warm pine]

it looks zany enough to me

NOTHING and such

the macros

[blissful smelt]

even better do att asm

[twilit smelt]

It fits my mental picture of like the temples of mica too closely.

[warm pine]

i had a close look at pillar and it's quite distinctive

a bit loud frankly

CONVERT_ARITHMETIC to cast between integers for instance

[twilit smelt]

yeah

I mean I have a CONTAINEROF keyword

my casts are CAST expr TO type

[warm pine]

this is so rare it's fair enough

[twilit smelt]

Also im probably being silly being so upset that my third OS in a custom language wasn't done by the time I was 23

MOST people have only done one, maybe two.

by that age

[blissful smelt]

one upping mintia2 seems like quite the endeavor

[blissful smelt]

mintia3: "It's basically NT 5.2 in a different programming language"

[twilit smelt]

mintia2 is the third one

[blissful smelt]

will u plan to make a fourth one

[twilit smelt]

It will never need to be rewritten because it is now in a state where there is nothing stopping it from evolving eternally or at least as far as any of the modern kernels have gotten

It's in an acceptable language and will be ported to real hardware

[blissful smelt]

final form

[queen torrent]

fourth one will be aisix2

[blissful smelt]

aidsix

[queen torrent]

after like 10 years you should add a 64 bit xrstation

to simulate hardware getting better over time

[blissful smelt]

and add custom cores for dedicated AI compute and add Mintia Recall as a necessary feature

[heady bobcat]

I realised yesterday that there's "NT" in "MINTIA".

[twilit smelt]

arabic hyenasky be like: mintiyya

thank god im not hyenasky if i were itd be weird i just made a self-referential joke as if i were another person

[twilit smelt]

god you can just see the unspoken "... unlike fucking MIPS." here

so much hatred.

its funny that all the other riscs made sure to put their "zeroth-level" exception vectors in untranslated memory especially the TLB refill traps

and in xr17032 i just make them virtual as well. including, yes, the TLB refill handler

how? thats a secret. patent pending. this WILL be the greatest architecture of 1990

im not letting sun get their grubby little fingers on my trickz

[mortal thunder]

another thing we have in common

I have a lot of stuff I need to implement, but just not the motivation to do it

[shadow ridge]

if you dont have motivation it's because it's not fun anymore

go do a few side projects

and come back when you're feeling it

[mortal thunder]

Okay, maybe I misspoke. I do have some motivation, but not enough to have 12 hour coding sessions like I did at the start of every large scale project of mine

[shadow ridge]

12 hour coding session is insane

[mortal thunder]

That's what happens when I start a project

Emphasis on start

[shadow ridge]

still insane

[twilit smelt]

when i got that R3000 NT build booting i was pulling 20+ hour sittings

[mortal thunder]

because I want something to work as soon as possible

[shadow ridge]

if you dont code 40 hours a day are you even a programmer

[twilit smelt]

i had never known how badly that could actually wreck your body until i did that a few days in a row

parts of me were sore that had never been sore before

[mortal thunder]

my monkey brain likes to get validation and I kinda got used to it with my relatively recent projects

[twilit smelt]

i listened to this album on loop the entire time and now i get like a ptsd response when i hear a song from it on shuffle https://open.spotify.com/album/1BVQ0iqg1mFg3Q48OLlwOT?si=blkeHwSATGqTCqvFQ8rvrg

[shadow ridge]

can you say the artist name and album name cuz my discord is broken with spotify links

[twilit smelt]

death's dynamic shroud - ill try living like this

[shadow ridge]

thx

[twilit smelt]

actually a very good album if i hadnt ruined it by listening to it probably 50+ times

[shadow ridge]

wtf is this

[mortal thunder]

i have a few songs that are so good i didn't get tired of listening to them for like 2 years

which is crazy

[shadow ridge]

the first track sounds really bad with the weird ass oh yeaaah voice

bad as in not that it's a bad song it just sounds dirty af

[twilit smelt]

thats my least fav song on the album

[shadow ridge]

i can tell lol

[twilit smelt]

also its all kind of an acquired taste

[shadow ridge]

the scat is insane

love it

[twilit smelt]

https://open.spotify.com/track/7KTbdZRAZ3q3wYGz5mD4tR?si=2790912946fd4797

this one was probably my favorite

after the intro

https://open.spotify.com/track/3rYcZ3jBj86xfF0l1tXzxE?si=58e5a0dd56ef445c

also this one has a beat switch at about 2:30 thats extremely pretty

experimental music is worth listening to the only issue is sometimes the experiment fails.

but sometimes it works

which is why its worth listening to

[shadow ridge]

you should make AI songs about mintia and xr station

[twilit smelt]

never say those words in that order again

[shadow ridge]

you should make AI songs about xr station and mintia

[twilit smelt]

theres another beat switch at around 4:55 thats even prettier

[mortal thunder]

no

[twilit smelt]

also im like 90% sure it actually samples the Windows NT 4.0 shutdown sound chopped and spliced

so it seemed appropriate

er startup sound

https://www.youtube.com/watch?v=6suWwnjGKgQ

tell me if u can hear it

[shadow ridge]

i dont hear it

[twilit smelt]

not there

[shadow ridge]

still sounds 4:4

[twilit smelt]

at 4:55

or a bit after that

[shadow ridge]

still in 4/4

[twilit smelt]

wdym 4/4

thats not what i meant by beat switch

i meant like

the beat

changes

not

the rhythmic meter changes

or whatever

[shadow ridge]

oh like

the uh

drum track or whatever u call it?

[twilit smelt]

listen to the NT 4.0 startup sound and then go to like 5:00 in that song and tell me if u can hear it

[shadow ridge]

i dont hear it

[twilit smelt]

my cousin is like a music production guy and he said something like "yes thats either a heavily edited sample of that sound or it sounds remarkably like it in a 1 in a million coincidence"

[shadow ridge]

it does have the same vibe

do you mean exactly 5:00

or before

[twilit smelt]

not exactly

a little after maybe

more like 5:05

[shadow ridge]

if thats what they're sampling, it's extreeemely edited

[twilit smelt]

well yeah it would be

https://open.spotify.com/track/5Oh0xRjwd2rteQU68hAbha?si=2a0abebdfb7142b8

this song is cool from the beat switchup around 1:10 until a while later

think that part sounds cool

this is an artist where i mostly only like "parts of" their songs lmao

their songs are full of switchups

then theres 18 carat affair which is just weird but strangely attractive https://open.spotify.com/track/59jVonr9Ko6gszCYD2DeAw?si=8492157cc7b4461e

i was a like top 0.01% listener for them in 2022

https://open.spotify.com/track/336QHH275JidjharPfi5PU?si=5a36dec151324e79