01:00 <sufdone> OR aka intersection means a bitwise ior done to the elements/bit-vector fields, the underlying implementation arithmetic is as seen rather sub and add in the library. integer itself can be mapped on 32 bit value through exponention range indexing, that means you add offset to one value as range index which will be summed like in ior comments, so integer that has all bits in is represented as 32 contiguous values aka 1+2+3..to 32. max value is 528

01:01 <sufdone> an arithmetic hence as maximum 32 answer sets

01:01 <sufdone> as/has

01:02 <sufdone> so such compilation and data packing is planned

01:02 <sufdone> it would solve all performance related issues if paddle is pushed a little more

01:41 <sufdone> range indexing means the difference is used to index into transition values, the API is very clear, and the container indexes show where it is taken, if the match is in as result of ior meaning either of the range bits have 1, it's taken from container 1 else the sum is made to new container, who user needs to free later.

01:53 <sufdone> just reading the code helps, so default operation is equality likely that means a match of junctive AND.

01:55 <sufdone> So there new graphics spec can be quite easily exposed, i am soon working on the prototype

02:37 <sufdone> very advanced but thin library, written very well by programmers. So the floating point format is just about doing correct "or" and "and" ops on correct ranges

02:40 <sufdone> I am amazed by such work my own, very very clever stuff

02:45 <sufdone> i talked about the netflix compression stuff, and people had solved it already on the go, but i have still a month to program before getting onto prototype my own too.

02:47 <sufdone> 100 percent success is likely guaranteed, as that work is not alot in amount

03:26 <sufdone> W can get so bad ass performance out of this library with so many methods, but it would make sense to make some medium implementation to community and leave the best performance to military etc.

03:26 <sufdone> so i already know which is the most performant compression and code, but i expose one step lighter

04:19 <sufdone> So the floating point presentation is maximum mantissa, is 1+2+3+4+5+6+7+8 and remainder is 1+2+3...to 23 and the operation is bitwise "or" on mantissa bits, then "and" and "or" on the remainder bits as pass 2, 1st i,e mantissa field remains as or twice called so, but remainder bits have or on first bits 8 bits from 23, then "and" on all bits 1to23

04:21 <sufdone> negation pass they also already have, that means that the value is sign negated

22:34 <kroll> enjoy http://qwj6kqpc7crsofkexdpmoltkjpb4tqh5xcgpkw762au4ycd7mgdmx7yd.onion