Jump to content

J–Machine

From Wikipedia, the free encyclopedia

The J–Machine (Jellybean-Machine) was a parallel computer designed by the MIT Concurrent VLSI Architecture group in conjunction with the Intel Corporation. The machine used "jellybean" parts—cheap and multitudinous commodity parts, each with a processor, memory, and a fast communication interface—and a novel network interface to implement fine grained parallel programs.[1]

History

[edit]

The J-machine project was started in 1988 based on work in Bill Dally's doctoral work at Caltech.[2]

The philosophy of the work was "processors are cheap and memory is expensive," the J in the project's title standing for jellybean which are small cheap candies. In order to make use of large numbers of processors, the machine featured a novel network interface using message passing.[3] This allowed a node to send a message to any other node within 2 microseconds.[4]

Three 1024-node J-machine systems have been built and are kept at MIT, Caltech and Argonne National Laboratory.[5]

[edit]

Notes

[edit]
  1. ^ Dally, William; Chang, Andrew; Chien, Andrew; Fiske, Stuart; Horwat, Waldemar; Keen, John; Lethin, Richard; Noakes, Michael; Nuth, Peter (1998). "The J-Machine: A Retrospective" (PDF). Retrieved 2009-06-17.
  2. ^ "J-Machine Project Page". web.mit.edu. Retrieved 2022-12-30.
  3. ^ Dally, William J.; Towles, Brian (2004). Principles and practices of interconnection networks. Morgan Kaufmann. pp. 102–109. ISBN 0-12-200751-4.
  4. ^ Hord, R. Michael (1993). "12. The J-Machine: A fine-grain concurrent computer". Parallel supercomputing in MIMD architectures. CRC Press. pp. 225–236. ISBN 0-8493-4417-4.
  5. ^ "The Jellybean Machine". CVA Group, Stanford University. Retrieved 2009-06-17.