00:58 <gog> https://i.imgur.com/kzHfRqC.png

01:15 <jimbzy> gog, I just thought about myself wearing that and vomited a little bit.

01:23 <gog> lol

01:26 <klange> Not into chokers.

01:32 <gog> what about the skirt and the gloves?

01:45 <klange> gog: more defined pleating, please; and those are arguably just arm covers and not gloves, gloves need at least a bit more intentional hand coverage

01:46 <klange> oh no my opinions on skirts knocked out gog

01:50 <vdamewood> Yay, gog's back.

01:51 <gog> yes

01:52 <gog> me marries the fishy

07:42 <arahael> "gloves"?

07:44 <klange> They're definitely arm wraps.

07:44 <klange> Not to be confused with ARM wraps.

08:23 <wleslie> I think I may be beginning to understand PCI. I've been comparing open source drivers and the osdev wiki and I think I've got something. I would love to hear if I'm missing something. Here goes:

08:25 <wleslie> so if you're implementing PCI, there's not much that is genuinely needed; but seems to be a whole stack of optional, useful interfaces it provides. you can get by simply enumerating the devices on each bus, querying their vendor and product or class and subclass, starting the relevant driver and mapping physical addresses and io ports described by

08:25 <wleslie> any BAR into the driver's address space and calling it done.

08:27 <wleslie> now it's a good idea to go through and compare the resources claimed by each BAR to ensure that there are no overlaps, especially if you have hotplug devices, but BIOS systems should already be set up without any resource conflicts

08:28 <Mutabah> Yep, that's pretty much it

08:28 <Mutabah> On PCs (which have a pretty advanced pre-boot firmware), all devices will have had IO and memory spaces allocated, so all you need to do is enumerate it and use the pre-existing regions

08:31 <wleslie> superb. so I can go back and add support for power management or MSI when I need them, and get on with porting drivers for now.

08:32 <klange> I still need to set up mapping for devices the BIOS didn't set up for me - my ExpressCard XHCI card is unconfigured.

08:33 <wleslie> thanks, that's good to know. so if I have a driver for some hardware, it is worth asking the driver which BARs should be configured?

08:35 <Mutabah> Iirc I've done two approaches

08:35 <Mutabah> 1. The PCI bus handler detects the BAR types (by checking the set bits), and determines the size (by doing a read/write/restore)

08:36 <Mutabah> 2. The device driver queries the bus handle for the BARs it actually needs

08:37 <Mutabah> (Example driver: https://github.com/thepowersgang/rust_os/blob/master/Kernel/Modules/nic_rtl8139/lib.rs#L448-L474)

08:41 <wleslie> right so this driver is requesting that 2 BARs be set up?

08:41 <Mutabah> Just one, and and an IRQ

08:41 <Mutabah> `bind_io` is the function that decodes a BAR

08:42 <wleslie> ok, cool

13:52 <vai> hi

15:04 <Oli> Hello, vai!

18:11 <c2a1> What is a typical use case for SIMD architectures

18:11 <c2a1> And also could they be used in something like a scheduler?

18:12 <zid> for doing simd on

18:12 <GeDaMo> Arithmetic

18:26 <superleaf1995> hmmm

18:35 <superleaf1995> xtensa causes itnernal ldscript error

19:03 <geist> never looked into xtense

19:03 <geist> xtensa. i assume it's just another embedded like risc machine

19:04 <geist> ah, founded by one of the founders of MIPS, so probably pretty mipsy in design

19:09 <Bitweasil> geist, your team hiring remote?

19:10 <geist> Yeah I think so. Also remote is kinda a floating concept nowadays

19:19 <kazinsal> openbsd 7.0 just released, adds rv64 support and partial apple silicon support

20:33 <geist> noice

20:34 <j`ey> Linux 5.16 should have a bunch more support too.. still no nvme though

20:35 <geist> oh for the sifive unmatched? may be mostly to deal with the PCIe bridge

20:38 <j`ey> https://www.openbsd.org/riscv64.html yup, unmatched

21:41 <kazinsal> updated my testing procedure, instead of needing to manually power my test VM off, scp the new bootable ISO to the ESXi box, and power it back on, I just copy the new boot images to the TFTP root and punch "reload" into the serial console