So in comp architecture today I learned that when using a data cache and a table look aside buffer to fetch instructions and data you could actually have to go to disk twice if you miss in the TBL and the cache. How crazy is that, like 100,000 cycles for one instruction... so much stalling.
