[sldev] Texture caching/bandwidth/performance how to proceed

[Buckaroo Mu]

I really regret missing this meeting - had planned on it, went early & 
got a LM, and then got too busy at RL work to attend.

Further thoughts, after trapesing my way through the source - it 
/appears/ as though the cache code already supports reading TGA files 
from the cache - am I correct? Does the cache writing code support 
writing in TGA? If so, it's just a matter of telling the image fetching 
code to write out the decompressed version - although I get the feeling 
that this could be done much more gracefully with a larger rewrite. 
Alas, I lack the C++ skills for it. Damn LL for not coding in plain C 
(*snort).

Right now, I've bumped the limit on my texture cache size (a la the xml 
hack posted on another thread), but we're still dealing with the decode 
delay. I think that's the thing that really needs to be addressed, 
before further discussion about discard algorithms etc.

-Buckaroo