Long Time No See

Yes, yes, it’s been a while, some number of weeks or mesa-years since I last blogged. Lots of important things have happened in that time. I’ve generated enough blog content for an entire month of posts, in fact. Maybe I’ll manage to maintain enough motivation to write about them.

Let’s kick off the return by looking at some progress updates.

  • maintenance5 finally released (but nobody can implement it yet)
  • host_image_copy finally released (lavapipe world-first open source best driver #1)
  • alyssa delete code big
  • some other things happened?

It’s all very exciting, and there’s definitely gonna be lots of posts once I remember what happened when I was melting into a puddle during the heatwave.

In the meanwhile, I want to talk about something else. Something a lot of people ask me about.

I want to talk about working for Valve.


I work in Open Source, so obviously I can only comment on that, and I only work on certain projects, so obviously I can only comment on my experience working on them, and I’m only one non-hivemind mortal being, so obviously I can only comment on what I’ve personally experienced, but I’m nearly through my third year here and it feels like a good time for a post of this sort. You know, because three.

So what’s it really like here?

In a word, working here is great.

Imagine you’ve got three or twenty projects you enjoy working on. Now imagine your task is to make them better. Better how, exactly? However you want. Which one do you work on? Whichever one you want. How many hours per week do you work? However many you want. Who checks your progress? You do. How do biannual performance evaluations happen? They don’t. And on top of all that you get paid.

It sounds too good to be true, doesn’t it? Surely working here can’t be such a complete state of anarchy, and indeed it isn’t. In my experience, the Open Source team here is like a big jigsaw puzzle: there’s a ton of different pieces, each of them with its place, each of them making the others better.

Let me explain.

The Team

There’s a lot of people working here, all of them smarter than me (but none of them blogging more than me). Most of them have been here longer than me too. Every one of them has fallen into their niche, the place they like to tinker where they can excel. Here’s a few just out of the old-timers:

  • Hans-Kristian works on VKD3D-proton, Vulkan spec, self-help guides, and is generally the graphics bible
  • Joshie has a frog in most of the Linux gaming ecosystem components in addition of being a top frog in DX layering and frogging occasional patches into Mesa
  • Timothy Arceri can make Mesa’s GLSL compiler roll over, sit, shake, and compile shaders in finite time
  • Daniel “fyi I’ll be on vacation until Sunday” Schürmann is our ACO architect and resident workaholic
  • Bas isn’t technically on the team but also he is
  • Samuel has written eight new RADV extension implementations in the time it took me to write this post
  • And I know I’ve just left out a ton of people because this is like 5% of the team here—really just the ones I could meme about with my last functioning brain cell

Everyone has distinct roles that they play on the team. Project areas they specialize in, as per my “anything goes” claim above. Some people work on lots of things, some don’t, but the niches are filled. Everyone’s got their “spot”.

Put another way, everyone on the team is a piece of the puzzle, the puzzle being “make Linux gaming great”. Everyone fits into their spot, and by doing so things get better.


There’s another way of looking at things though. While everyone here can be a puzzle piece, everyone has their own puzzles too. I work on zink, but that doesn’t mean “Mike is the one working on zink”. What it really means is that I’m able to work on zink because the puzzle pieces have been assembled such that I’m able to work on zink. It’s like how you wouldn’t try benching four plates at the gym without having a spot (I would, but I’m huge).

Sometimes getting a spot is a production. You know, the kind of thing that makes headlines where Joshie throws me the rock, but it’s too hot so I fire off a pass to Georg, and he slows down the tempo while we wait for hardware vendors to understand how their thinking sand interacts with complex ancient graphics specificiations, but then we get in the zone, and Georg throws me an alleyoop, and then Joshie takes it back to fix the original problem and now games work a little better.

Sometimes it’s all rockstars all day long.

But it’s the other times that make working here really great. The times when you’re struggling to grind out that last rep because you told your buddy you were definitely gonna hit 10x315 on front squat this week and you don’t wanna go back and admit you had too much preworkout.

I’m talking about times like when Timur picked up my massive CPU optimization series and wrangled it into a bunch of MRs because I was foolishly stretching myself too thin across too many projects.

I’m talking about the unsung heroes who make working here truly great.

Case Study: Rhys (not that one, the other one) Perry

Everyone knows Rhys. Everyone here, anyway. Outside the team it might be a different story; he has no blog, and searching for his many and varied accomplishments in the depths of the internet yields only one article written before I was born.

IYKYK, as they say. Just in the past week he’s quietly fixed Rage 2 and WWZ. A glance through his extensive patch history is a litany of complex optimizations and tweaks which aren’t flashy enough to be newsworthy on their own but add up to significant gains through consistent improvements.

But it’s still not any of these that (to me, at least) make Rhys one of the unsung heroes of the team. The glue that holds parts of it together.

All my very high IQ readers know what it’s like to get stuck on something. That feeling when you come across a problem, and you know it’s a problem, and you can handwave some half-functional solution that lets you limp across the finish line to collapse in a broken, battered heap with 0 regressions as the output of your refactoring branch’s latest CTS run, but you can’t quite figure out the “right” way to fix the problem. The way that won’t get your patches NAKed harder than proposing the addition of registers to NIR right now.

At times like these, who’s there to help you out? Who is it that gives the bar that tiny, it-was-all-you-bro-I-didn’t-even-touch-it nudge to help you finish that last rep?

It’s Rhys Perry. It’s always the Rhyses, the unsung heroes. The ones who answer complex questions on IRC at 2am because they’re watching an historic cricket match and happened to glance over and see you flailing away at your keyboard. The ones who step in and say “sure, I’ll review this absolute disaster you fingerpainted into gitlab using the webui without regard to formatting, or coding conventions, or even the right programming language, and we’ll get through this together with a fix that’ll make everyone happy” when you’re staring down Yog-Sothoth in the abyss of the compiler stack at the end of a week and have exactly one functioning brain cell remaining that tells you only SHOWER. NOW. IT’S BEEN DAYS.

And it’s the mixing of all these people, rockstars and not, unsung heroes and not, working on so many projects, enabling each other and making us all better at what we do that makes working at Valve great.

To me, at least.

Tune in next time when I’ll be MS Painting some XFB memes and raging about Big Triangle since that’s apparently my niche.

Written on August 3, 2023