How do you deliver a feature on the biggest stage in the world?
summary
Disguise uses Single Large Canvas to calculate exactly which input pixels each of 23 disguise servers needs to fill the Sphere’s 16k interior so a single 16k source becomes many machine-sized files. The team ships a slicer (vue.js prototype with C++ back end) that emits a JSON spec, runs UK/US overlapping shifts, misses an internal Sep 1 target but lands a fix on Sep 24, the show goes live, and later a texture allocation removal yields a measurable performance win.
AI-written summary · about
chapters (8)
- 0:00Sphere project introduction
- 2:04Content delivery challenge
- 3:51Single Large Canvas
- 5:40Slicer tool development
- 7:24Team leadership challenges
- 8:49Loss of oversight
- 9:50Final sprint and delivery
- 10:33Post-show improvements
related talks
32mMaking of this Stage by { live : js }
39mThe Pixel Factory
18mVideo @Scale Fall 2022 LIVE Q&A #3
18mEscaping flatland: a romance of data science in the browser
32mHow ZQSD Used the Technique in Esports
36mCanvas Pixel Transformation: Beauty comes slowly
29mParallel Beauty - with Isolates into the Mandelbrot set
30mSeeing the needle AND the haystack: single-datapoint selection for billion-poin...