[ut.general] cider seminar on Simulated Annealing VLSI Placement on an MIMD Multiprocessor

szymans@utecfa.UUCP (Ted Szymanski) (01/28/86)

                Fast, High Quality VLSI Placement
                    on an MIMD Multiprocessor
                        By Jonathan Rose
                           Room GB 221
                           Time: 12:05
                  Date: Friday, Jan. 31st, 1986
                            Abstract
High quality automatic VLSI  layout,  especially  placement,  has
been notoriously difficult to achieve.  Recent advances in layout
algorithms, notably Simulated Annealing, have  attained  signifi-
cant  improvements  in  quality  but  at  the cost of an enormous
amount of computing time.  This seminar discusses recent work to-
wards  obtaining the same quality as Simulated Annealing but in a
shorter time using a Multiple Instruction stream - Multiple  Data
stream (MIMD) multiprocessor.
The objectives of this machine will be discussed and a quick  re-
view of Simulated Annealing will be given.  Some experiences with
the algorithm and a characterization of its search space will  be
presented.   Several  approaches  to parallel Simulated Annealing
will be described, and the  results  obtained  to  date  will  be
given.