16:47scientes: my computer crashes randomly with 4.14+, I went back to 4.9 and it is stable again
16:47scientes: the problem is that I don't get anything in dmesg. Sysrq works, but nothing else
16:48scientes: I just bought a new video card so maybe I will be able to get some more info
16:51mooch: nyef, no 386 emulator currently runs most debuggers correctly, as they use 386 debug registers, which none of them have implemented correctly
16:51mooch: the ONLY one that has, is my pull request to MAME
16:52mooch: ...which isn't merged yet
16:52scientes: qemu choose not to implement ptrace() because it is a can of worms
16:52scientes: qemu-user-static that is
17:04karolherbst: scientes: are you able to ssh into the machine when it happens?
17:10nyef: mooch: Thus, YOU have an emulator that can run such debuggers.
17:11mooch: eh, true
17:11mooch: but even then, my debug register support doesn't FULLY work :/
17:12mooch: at least, not when it comes to loadalls
17:12mooch: also, in mame, you can't break into soft-ice from the dos prompt
17:12mooch: seriously, i've tried
17:16nyef: Have you tried figuring out why you can't?
17:17scientes: karolherbst, i forgot i could try that
17:20mooch: also, nyef, mame in its current state is REALLY not that good for pc emulation in general, so :/
17:20mooch: also, i can't even write asm code
17:20mooch: i've tried, and my asm code is absolute horseshit
17:22nyef: ... So, mixed with celulose, it's a promoter of plant growth? The cure for that is practice.
17:22nyef: It's still code. If you can code in C, it's not a big stretch to be able to code in ASM.
17:23scientes: karolherbst, yes i can
17:23scientes: just happened
17:23scientes: and firefox is still chewing up my processor even though graphics has died
17:24scientes: and the screen is frozen
17:28nyef: mooch: It seems clear that you don't WANT to, for whatever reason. Either OWN that reason, or acknowledge that it's not an important reason and start to pick up the skills that you claim not to have (and there is plenty of help available there for the asking), but please stop making excuses.
17:28nyef: (And, yes, the "making excuses" thing is somthing that I struggle with as well.)
17:29scientes: karolherbst, is there something I can do, cause I want to reboot my system and get back to work
17:29mooch: bruv, i once made a demo called copper2. the code was so fucking AWFUL that it spawned a meme on pouet
17:29mooch: and yes, it was in asm
17:30mooch: then again, that was in 2012 or so :/
17:30nyef: I once tried to make an SNES emulator. Around '96 or so, maybe? It was so bad that I swore off writing emulators forever.
17:31nyef: "Clearly, this is not a thing that I can do."
17:31JayFoxRox: nyef: work on Xbox emulation :P we barely do any traditional emulation tbh. it's mostly just doing research and writing tools, because we know so little about hardware (+ PC emulation is more or less a solved problem, so we don't have to deal with it)
17:31nyef: By '98 I had managed to trick myself into writing an NES emulator that worked well enough to publish.
17:32mooch: JayFoxRox, i REALLY wouldn't call it a "solved" problem
17:32mooch: there's a game from like, 1994 or so called power dolls that was on pc hardware. it doesn't work in even the MOST accurate pc emulators
17:33mooch: not even in 86box
17:33mooch: but it DOES work on real hardware
17:33mooch: in fact, not even the installer works in emulators
17:33JayFoxRox: mooch: that's why I said "more or less"
17:33mooch: ah lol
17:33JayFoxRox: mooch: huh? any idea what's wrong with "power dolls"? that sounds interesting - I didn't know that one yet
17:33karolherbst: scientes: I assume nothing shows up inside dmesg still?
17:34karolherbst: what about your sys logger?
17:34nyef: JayFoxRox: As much as I'd like to have an Xbox emulator, I don't find PC-type hardware all that interesting to emulate for some reason.
17:34mooch: JayFoxRox, it's some sort of dumb mmu bullshit
17:34scientes: damn i just rebooted
17:34mooch: i have no idea other than that
17:34karolherbst: scientes: no worries, just check whenever it happens the next time or something
17:34mooch: also, in mame, quitting the emulator once you've started the installer almost invariably crashes the emulator
17:35mooch: also, it seems to only work in dos/v
17:35mooch: so i hope you know japanese :p
17:35JayFoxRox: nyef: again: we don't work on PC style emulation there. We do work on DSP, audio playback, GPU emulation etc. I've worked on Citra (3DS emulator) for a while, and it was basically the same kind of work - but Xbox is trickier in many ways.
17:36mooch: also, no pc emulator is even functionally perfect much less has good timings
17:36mooch: most of them literally just do 1 instruction = 1 cycle
17:36mooch: even bochs and qemu
17:37nyef: The simpler the CPU, the easier it is to get accurate timing.
17:37JayFoxRox: that's also why I don't call Xbox a PC anymore either. you *can* use it like a PC (which Linux does). but then you don't get to use the audio hardware or GPU, which games use to great extend. Games don't even use legacy features like AC97 - they directly use another audio chip. As xbox runs entirely in kernel mode and games are statically linked, they also access the hardware directly and are basically bare metal
17:37nyef: As soon as you involve a cache, the timing complexity goes way up.
17:38nyef: Overall, I would expect that a 286 emulator might have accurate timing, but a 386 emulator almost-certainly won't.
17:38mooch: nyef, nah, even 286 emulation doesn't have accurate timings
17:39mooch: hell, we STILL don't know everything about the 8086
17:39mooch: oh sorry
17:39mooch: jeez, i must be a bit loopy today, i'm missing even obvious nuance :p
17:39nyef: The 386 is about when mainboards started getting L2 cache.
17:39nyef: (Or is it L1 cache? I forget.)
17:40nyef: There's also TLB modeling to consider at that point.
17:41nyef: ... Seems there are some nine games in the "power dolls" series?
17:43mooch: i'm talking about the original dos/v release here
17:43mooch: even the original game was on many platforms
17:48nyef: Hrm... I wonder if I could take a Riva 128, and re-use the chips to build a NuBus video card?
17:51mooch: uh, those chips have a LOT of pins on them, so i kinda doubt that you'd be able to solder them very well
17:53nyef: Yeah, looks like it's a PGA or BGA?
17:54nyef: Plus, speaks PCI/AGP like a native, so bridging to NuBus would be even more of a trick than the bridge for connecting NuBus cards to PCI machines.
18:07nyef: Now I'm wondering if I already have anything in the NV03-NV05 range.
18:07nyef: Hrm. Not likely to have an NV0a, am I?
18:15nyef: ... Two MGA cards ("Millenium"?), two ATI 3D Rage cards, one Trident TVGA card...
18:16nyef: Also an ne2k-pci.
18:27nyef: ... An ATA controller, two SCSI controllers, a cheapo sound card, another ATI 3D Rage IIC, and more ne2k-pci cards than I care to count. Hrm.
18:40nyef: ... GeForce 2 MX?
18:41nyef: NV11, maybe?
18:41nyef: (The system wouldn't boot for whatever reason, but the card clearly works enough for the BIOS, and I can read the chip labels.)
19:01nyef: NV11. Kernel 3.2.0, some version of ubuntu. Hrm.
19:07nyef: Anything else I have is likely to be considerably newer... and probably for PPC macs.
19:11nyef: Okay, modulo the one laptop with the dead battery, the broken backlight, and the NV18(?).
19:30JayFoxRox: mwk: do you know the the width and representation of register combiner registers on NV2A? are they just 8-bit two's complement fixed point [what range?] ? NV_texture_shader said 9-bit signed fixed point [for reg comb], but we are not sure if this also applies to NV2A
19:31mwk: JayFoxRox: that's an interesting question
19:31mwk: I do not know
19:32mwk: but I would say at least 9-bit signed
19:32mwk: fixed point
19:32mwk: mayyyybe 10-bit
19:32JayFoxRox: I just realized the reg combine spec also says: "The fact that NV10 does the computations as 9-bit signed fixed point is not mentioned in the spec."
19:33mwk: yeah, well
19:33JayFoxRox: so for NV10 we do at least have a strong hint that this holds true. I intend to do some experiments with it soon [probably just putting a bitmask in the constants and shifting them left / right]
19:33mwk: you're welcome to experiment
19:33mwk: I'm betting on 9-bit fixed point signed
19:34JayFoxRox: yeah, me too :P and for texture shaders 8 and 16 bit [although that's going to be a pita to debug]
19:34JayFoxRox: I'm looking at shrek right now - which does deferred rendering on NV2A. I did trace the framebuffers yesterday and it's really amazing imo, such madlads
19:36JayFoxRox: if you care, there's plenty of cool stuff here. including code they used: https://sites.google.com/site/richgel99/index#TOC-Shrek-Xbox-and-Deferred-Shading
19:38mooch4: holy fuck that was made by dice
19:42nyef: mooch: So, if you're not working on 86box, what ARE you working on these days?
19:47mooch4: not much tbh, i'm still trying to figure out how i'm going to make an accurate 386 pc emulator
19:47mooch4: i wanted to use mame's i386 core as a base, but that's proving too hard
19:47mooch4: so i'm trying to make my own 386 core based on the official manuals from intel
19:48mooch4: though tbh, i'm thinking about trying to find a version of linux that has python 2.6 AT LEAST, and runs on a 386 so i could, say, use sandsifter on mame's 386 emulation :p
19:49nyef: Hm. MAME's emulation not worth trying to fix?
19:53mooch4: well, more like it's such complex code that i can't do fix mame's pci shit myself, and nobody else wants to right now either
19:53mooch4: besides, if you did all the mmu checks you're supposed to in mame, it would likely cripple performance
19:53mooch4: in my own emulator? not so much
19:53nyef: Fair enough.
19:53mooch4: mame's 386 emulation is slow enough already
19:54mooch4: mostly due to mame's memory system
19:54nyef: qemu also a non-starter here?
19:55mooch4: yeah, qemu is also a non-starter
19:55mooch4: and bochs? euugh
19:55mooch4: that code is a MESS
19:55mooch4: also, i want to use c++ here
19:55mooch4: not c
19:55nyef: I thought that that particular epithet ("MESS") would've applied more to MAME than bochs?
19:56mooch4: nah, bochs is way messier than mame, code-wise
19:56nyef: I can believe it.
20:17JayFoxRox: stupid question: how do I find out when an opengl extension was first published / registered? the revision history only contains when the first update was done
21:28nyef: Damnit. Turns out I have a "Diamond Viper V550", but it's an AGP card that's not compatible with any of the AGP slots that I have.
21:45nyef: Hrm. Voltage, huh?
22:21nyef: ... And just turned up what looks to be a GeForce2 GTS (NV15?). This is getting ridiculous.
22:31rhyskidd: skeggsb: do you have a sense of timing on the GV100 fuc in linux-firmware.git, now that 4.18 is released and looks for those files?
22:32rhyskidd: (appreciate that it's ultimately up to NVIDIA)