Upper Bound Sebastian

[poettig]

Hi,

maybe the problem to come up with reasonable upper bounds for the block weights isn't that complicated.
If we look at the overall problem and the 2-phase approach, then we will first use the bin
packing algorithm to distribute the first couple of MSA partitions. Whenever it then decides
to split an MSA partition, it knows the remaining capacities of the CPUs it wants to assign
the splitted partition to. Thus, if we model sites without repeats as hypernodes incident to a
single-node hyperedge, then we actually get the upper bounds on the block weights from the
bin packing algorithm and we just could use these.

What do you think?

Best
Sebastian

[lukashuebner]

Prof said, "sounds good" -> closed