I have to patch TypeOfMem() too. Perhaps that is the flaky part.
A synthetic test loop showed it being 2-3x faster for chipmem. I didn't test the slowdown for fastmem.
UPDATE: Still buggy. Thinking.
UPDATE2: Major facepalm moment. This is absolutely not working. All FreeMem() calls will add the chipmem back to fastmem! To make it work it has to either split the original fastmem into two regions with the fast-chip part taken out, or you have to take over all the memory functions and integrate it into the core. Can we just have some new hw with faster chipmem plixxors?
Last edited by NorthWay; 01 August 2017 at 08:24.
Reason: update2
|