karl3ļ¼ writeme.com wrote: > so let's outline this unstable idea again > > - start with a region > > often we look for the next hole starting at the first read. > -> this gives a potential region of cached items > -> then a potential partially cached item > -> then a potential region of uncached items > -> then a potential partially cached item again > > lots of short reads may have only 2 and 4 and these would be roughly identical > there is possibly existing code for 3, and 1 is simple.
well, what would a parallel approach look like ? :/ uhhhhhhhhhhhhh 1549 maybe some aren't sure of pursuing a parallel approach if karl is still trying to stabilize the serial approach and could possibly help with that. uncertain
