15:02 <alyssa> Ella[m]: I have not observed restriction, MAX_PPP_UPDATES is just to make sure we have enough space allocated in the worst case with minimum wasteage otherwise

15:02 <Ella[m]> ah oky makes sense :3 thank you!

15:02 <alyssa> :)

15:03 <alyssa> Ella[m]: Also, the good folk at XDC convinced me that I should write a Vulkan driver actually

15:03 <Ella[m]> Oo

15:03 <alyssa> assuming you've gotten enough done that helping you is less work than starting from scratch myself

15:03 <alyssa> let me know how I can help

15:03 <alyssa> or would it be more helpful for me to stay on compiler + r/e duty and continue with the GL driver as reference code for you?

15:04 <alyssa> (There is a LOT of work to do for both compiler and r/e stuff and it's all needed regardless of API so I guess there's that..)

15:05 <alyssa> (and honestly I much prefer those two topics to either GL or VK driver dev so..)

15:06 <Ella[m]> uhhh not sure, currently I'm trying to figure out how best to do descriptor sets and I haven't done much more than hook up the compiler and do state emission. I haven't implemented vk_sync on top of macos stuff either yet.

15:09 <Ella[m]> I prefer working on VK driver dev and doing a little r/e but I'm not very experienced on the compiler side. Starting VK from scratch might be an option if you want idk?

15:10 <alyssa> Mmh, ok

15:10 <alyssa> We should get on a call with jekstrand about descriptor sets

15:11 <alyssa> I know the hw and jekstrand knows the Vulkan so together we should figure out the right way to do it

15:11 <alyssa> but neither is clear from the GL driver

15:11 <alyssa> [If I were writing the VK driver myself, I'd figure this out over a 1-on-1 call with jekstrand :-p]

15:14 <bluetail> I really like your blog alyssa. What have you found out later, that proven really valuable, that should I know in life?

15:18 <Ella[m]> alyssa: Yeah we should. I'm not really sure at all how to map it to the hw. I have some ideas but they don't seem pretty and probably aren't the correct way of going about these things.

15:36 <alyssa> nod

15:36 <alyssa> I think I know but jekstrand is the expert here

16:44 <phire> BTW Ella, I was reading that commit which tells me not to read it. Sorry ;-)

16:45 <Ella[m]> ah that's very messy code :3

16:46 <phire> I'm tempted to also offer to help work on the vulkan driver.... Something in the water at xdc

16:59 <bluetail> I'mma gonna stop being a fan bird. Doesn't seem to any helpful. Effectively I don't have the time and health to contribute in a meaningful way. That's why I tried to cheer / motivate. Perhaps I should not be so selective. Everybody who contributes is indeed, great. Sigh.

17:12 <Ella[m]> phire: I've pushed my cleanups from today if you want to have a look at something a little more clean. If you are going to read it, feel free to tell me when I've done something completely wrong.

17:24 <phire> You are the expert compared to me. The most gpu driver I've ever actually done was a hello triangle on videocore 4

17:25 <phire> Literally just poking registers from userspace

17:26 <phire> Wrote shaders in raw machine code

17:26 <jn> phire: the shaders run on the QPUs, right?

17:27 <phire> Yeah

17:27 <jn> three branch delay slots…

17:27 <phire> Yep. And pretty brain dead threading