I was searching through AMD's docs and but couldn't find anything for:
1. K-8 & K-10 internal registers, governing IMC settings, various frequencies and multipliers etc.
I was reading through memtest86+ code ( RAM test utility ) and I think it could be done _much_ better if one had access to such data.
Also, one could stress CPU cores while accessing RAM and tweak freq multipliers, RMA timings etc to either asess OC borders or stability of the machine.
2. chipset data ( especially GX790 with sideport, FX790 and SB750 ).
I seek to improve various ways of doing things in Linux and FX790 with inbuilt GPU and sideport seems very good foundation for many tricks.
For example, RAID-6 P and Q syndrome calculation rutines in kernel seem very optimised, but even so they can be slow, especially when two disks are down and everything has to be fully reconstructed.
Employing GPU shaders for such taks seems like a very good idea.
Also, console as well as graphic X11 terminal ops could be at least partially done in GPU as well as some encryption/decryption algorithms, so one could have encrypted HDD disks/partitons/files without much CPU overhead , reduced bandwidth or extra latency...
It seems to me that with a bit of right code RS790 could fly...
3. Discrete GPU chipset data.
Same as for point 2, albeight a bit more computing oriented. nVidia's CUDA seems to be fine, but it's not immediately obvious to me how one would use it in kernel code and how could it be integrated into display functions seamlessly. AMD had CTM program for a while and after it vanished I kinda hoped that specs are to be "opensourced" so CTM would be kid of pointless from that point on...
Is there some way to get to such data without sacrificing eldest daughter and a goat ?