[sldev] Cache speed experiment & results...

[Buckaroo Mu]

Dahlia Trimble wrote:
> If the goal were to improve seek time in a cache, how about keeping 
> the cache on one of those readyboost-capable usb flash drives? They 
> can have seek times as low as 1 ms or better, although the transfer 
> rate and write times arent all that fast.
>  
> Then again it's probably moot if the bottleneck is the decoding :/
>
Bad idea. Flash has limited write-cycles, you don't want to use it as a 
cache - you'll burn it out faster than you could believe.Besides, if 
seek times were the bottleneck, the RAMdisk experiment would have shown 
noticeable improvement. The bottleneck is definitely in the decoding. 
That's why I'm encouraging a test of caching raw data rather than 
re-compressed.