Writing a blog post on Data Oriented Design. Any skeptics want to share their worries/criticisms? Looking to address some of the most common ones.
-
Näytä tämä ketju
-
Vastauksena käyttäjälle @vengefularia
My concern: DoD is by design quite low level and can be a leaky abstraction (people baking in knowledge about their target hw into assumptions about data). On a single project this might be beneficial, but is dangerous if becomes “standardized” and dogmatic.
1 vastaus 0 uudelleentwiittausta 11 tykkäystä -
Vastauksena käyttäjille @BartWronsk ja @vengefularia
It can also lead to a similar *reversal* of logic/reasoning - instead of analyzing and understanding the data, retro-fitting it to their HW or systems (like ECS which I am personally not a big fan of - it’s a great tool, but not for everything!).
3 vastausta 0 uudelleentwiittausta 1 tykkäys -
Vastauksena käyttäjille @BartWronsk ja @vengefularia
Final concern: overall it’s a low level tool;I think we should move to higher level constructs and design methods in future (graph,declarative,functional,dependency based) and let new tools and compilers help with fitting those to target hw. (Examples: Tensorflow, Halide lang)
2 vastausta 0 uudelleentwiittausta 4 tykkäystä -
Vastauksena käyttäjille @BartWronsk ja @vengefularia
By this I mean that DoD algorithms are difficult to refactor/reorder/reorganize, especially if you want to change a few stages. They have a very specific transformation structure of data baked into its whole design; which is designed by heuristics/educated guess, hard to modify
2 vastausta 0 uudelleentwiittausta 3 tykkäystä
In all my experience with DoD, the exact opposite is true. DoD algorithms are easier to refactor/reorder. I started leaning into DoD because it made refactoring easier and my code more modular.
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.