An Update

Busy, Busy, Busy

It’s been a busy week for me in personal stuff, so my blogging has been a bit slow. Here’s a brief summary of a few vaguely interesting things that I’ve been up to:

  • I’m now up to 37fps in the Heaven benchmark, which is an absolute unit of a number that brings me up to 68.5% of native GL
  • I discovered that people have been reporting bugs for zink-wip (great) without tagging me (not great), and I’m not getting notifications for it (also not great). gitlab is hard.
  • I added support for null UBOs in descriptor sets, which I think I was supposed to actually have done some time ago, but I hadn’t run into any tests that hit it. null checks all the way down
  • I fired up supertuxkart for one of the reports, which led me down a rabbithole of discovering that part of my compute shader implementation was broken for big shared data loads; I was, for whatever reason, attempting to do an OpAccessChain with a result type of uvec4 from a base type of array<uint>, which…it’s just not going to work ever, so that’s some strong work by past me
  • I played supertuxkart for a good 30 seconds and took a screenshot

supertuxkart.png

There’s a weird flickering bug with the UI that I get in some levels that bears some looking into, but otherwise I get a steady 60/64/64 in zink vs 74/110/110 in IRIS (no idea what the 3 numbers mean, if anyone does, feel free to drop me a line).

That’s it for today. Hopefully a longer post tomorrow but no promises.

Written on October 15, 2020