The current gas model is pretty simple, at some point (maybe after the memory model change pans out) we should run some benchmark on individual instructions with various configurations (i.e. plaintext vs ciphertext registers) to make the gas cost reflect the actual cost better.
The current gas model is pretty simple, at some point (maybe after the memory model change pans out) we should run some benchmark on individual instructions with various configurations (i.e. plaintext vs ciphertext registers) to make the gas cost reflect the actual cost better.