@jof I'd assume that they're mostly power decoupling. The jauntiness looks to be simply a function of the tight routing at work. But. Argh.
-
-
@starsandrobots@jof I'm mostly interested in the giant heuristic win of allowing component layout, which simplifies the toposort of traces. -
-
@starsandrobots@jof Yes, which allows for multiple learned routers to duke it out in optimizing the traces within the placements. -
@starsandrobots@jof (The CFD people use fuck-off supercomputers; with that horsepower, I know that we could do similarly powerful things.) -
New conversation -
-
-
@starsandrobots@jof If NP-hard was an actual barrier to effective solutions, I think three or four of my jobs wouldn't have existed.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@starsandrobots@xek@jof there's a good rant from@dmolnar how practical SAT heuristics make NP reductions into useful engineering. -
@starsandrobots@xek@jof@dmolnar we reduced to NP to prove "FOO is *hard*" but with practical SAT heuristics, reductions become solutions. -
@eqe@starsandrobots@xek@jof No kidding! Hard instances can still show up...but that is what http://smtcomp.org/ is for!
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.