3 ms·

I suspect it may be: https://www.youtube.com/watch?v=5I4YPkVU7mY https://www.youtube.com/watch?v=5I4YPkVU7mY I'm interested in this because these linearized tr

I suspect it may be: https://www.youtube.com/watch?v=5I4YPkVU7mY https://www.youtube.com/watch?v=5I4YPkVU7mY I'm interested in this because these linearized tree representations lend themselves well to processing on GPU. We use them in Vello for representing clip and blend groups, with quite deep nesting level allowed. By contrast, the traditional pointer chasing approach to ASTs limits parallelism and has poor utilization of memory bandwidth.

sctb 3y ago
Thanks for the link! I just found the one I was thinking of, which is older but perhaps goes into more detail: https://www.youtube.com/watch?v=gcUWTa16Jc0&list=PLDU0iEj6f8duXzmgnlGX4hMHJUMYh4rJq&index=3 https://www.youtube.com/watch?v=gcUWTa16Jc0&list=PLDU0iEj6f8....
mhh__ 3y ago
Agreed but also this feels like a way to spending hundreds of engineer-years for a 7.3% speedup on the common case.
- sudosysgen 3y ago
  Even a 0.5% speedup is worth hundreds of engineer-years if it's in the common case for a popular language.