• 0 Posts
  • 20 Comments
Joined 1 year ago
cake
Cake day: June 11th, 2023

help-circle






  • It’s like saying ‘you might think this engine is broken since it can’t run on the water that it is filled with, but if you simply remove the water and replace it with petrol suddenly the engine is fixed.’

    The post seems to approach the paradox as if it meant to show that tolerance is inherently broken when in reality it just points out the possibility of problematic aspects if incorrectly applied, like in the above where it is obvious the engine itself was never broken. The paradox doesn’t disappear, it simply doesn’t apply to that particular application.

    The main idea from OPs post is often ascribed to Yonatan Zunger as some huge revelation, but really this idea has been about for quite some time as its not exactly hard to come up with. For example, K. R. Popper 1945, and E. M. Forster 1922 both wrote about this.












  • Game devs have many teams all with different jobs, for a big game like this you’d typically have multiple teams dedicated to optimization in different areas (and between them). The specific problem in this case was how the game was communicating with graphics drivers (among others), which for any graphics heavy game is very fundamental to performance optimization. The problems aren’t even an after-the-fact optimization sort of thing that teams should have to identify and follow-up on, batching jobs is standard practice when interacting with GPUs whether or not there’s a translation layer.

    When the devs of a core translation API between two supported graphics drivers that are commonplace in the gaming ecosystem have to write code to specifically fix issues with your application you’ve done something fundamentally wrong.