[Next] [Previous] [Top]
Computer Species for Scientific Programming
- Superscalar workstation: not quite big and fast enough for all
- Cluster of workstations (COW): too expensive to scale up to MPP size
- Shared-memory parallel vector processor (PVP) [C90]: Great!
- Static RAM means even gather/scatter operations are fast
- Fairly large memory
- Excellent optimizing compilers
- Ten years development of good vectorizing compilers after introduction of first Cray
- Just keep making them bigger and faster! (And cheaper!)
- Symmetric Multiprocessor (SMP) with superscalar CPUs: ???
- Dynamic RAM means gather/scatter operations are slow
- Fairly large memory
- Starting on ten years of development of good caching compilers
- Distributed-memory massively parallel processor (MPP): "You must love it." [Apologies to George Orwell]
- Really large memory
- Starting now on developing any parallel compilers (HPF)
- Hierarchical machines (SMPs + fast network):??????
Scientific Computing: Where does MPP Fit In? - (lecture given 6/14/95)
[Next] [Previous] [Top]
Generated with CERN WebMaker