Topic: Would changing the memory req of scrypt destroy compatibility? (Read 393 times)
I *think* the parameter is hardcoded in cgminer. The modification is small, but it has to be done.
I've read on this forum more than once that scrypt actually uses less memory than recommended in the original scrypt documentation -- and that the implications of this are higher ASIC vulnerability and a greater GPU multiplier effect. I'm mainly concerned about the first, and so my questions are:
* How difficult is it to change (in source) the memory parameter of scrypt, which I understand to be labeled as "n"?
* Would this break compatibility with all scrypt programs, such as CPUminer and CGminer?
* How difficult is it to change (in source) the memory parameter of scrypt, which I understand to be labeled as "n"?
* Would this break compatibility with all scrypt programs, such as CPUminer and CGminer?
Jump to: