Re: Misc: BGBCC targeting RV64G, initial results...

I guess also possible is to make it such that a the required density (to
avoid split) is correlated to table size.
>
Say (table size, minimum density):
  128, 75%
  256, 80%
  512, 85%
  1024, 90%
  2048, 95%

Density is just a vague indicator.  I think it makes more sense to
define your heuristic based on specific (estimated) costs
(e.g. code(+table) size, or number of cycles).


        Stefan