Maspar Computer Corporation

Maspar Computer Company

David M Bulfer

John R Nickolls

Won Sub Kim

John Zapisek

W Thomas Blank

William T Blank

Mark P McKee

Hannes Figel

John M Long

A massively parallel processor is provided with a plurality of clusters. Each cluster includes a plurality of processor elements ('PEs') and a cluster memory. Each PE of the cluster has associated with it an address register, a stage register, an error register, a PE enable flag, a memory flag, and a grant request flag. A cluster data bus and an error bus connects each of the stage registers and error registers of the cluster to the memory. The grant request flags of the cluster are interconnected by a polling network, which polls only one of the grant request flags at a time. In response to a signal on the polling network and the state of the associated memory flag, the grant request flag determines an I/O operation between the associated data register and the cluster memory over the cluster data bus. Both data and error bits are associated with respective processor elements. The sequential memory operations proceed in parallel with parallel processor operations. The sequential memory operations also may be queued. Addressing modes include direct and indirect. In direct address mode, a PE addresses its own address space by appending its PE number to a broadcast partial address. The broadcast partial address is furnished over a broadcast bus, and the PE number is furnished on a cluster address bus. In indirect address mode, a PE addresses either its own address space or that of other PEs in its cluster by locally calculating a partial address, then appending to it either its own PE number or that of another PE in its cluster. The full address is furnished over the cluster address bus.

Parallel processor memory transfer system using parallel transfers

To effect a block data transfer between a plurality of physical I/O devices coupled through interfaces to an I/O channel ('IOC') bus, a source logical device is established by programmably assigning to each of the physical device interfaces a logical device identifier, a leaf identifier determining when the physical device participates relative to the first data transfer in the block data transfer, a burst count specifying the number of consecutive transfers for which the physical device is responsible when its interleave period arrives, and an interleave factor identifying how often the physical device participates in the block data transfer. A destination logical device is similarly established. The source and logical devices are then activated to accomplish a block transfer of data between them. To permit different I/O processors to operate independently in making I/O requests, requests from each I/O processor are communicated to an IOC controller over another bus, which need not be a high performance bus, and are serviced to construct header packets in a transaction buffer identifying IOC transactions, including source and destination logical devices. When each packet is finished, the responsible I/O processor puts a pointer into a transaction queue, which is a FIFO register. Each IOC transaction is initiated as its corresponding pointer is popped from the transaction queue. Apparatus embodiments are disclosed as well.

Broadcasting headers to configure physical devices interfacing a data

Growing community of inventors

Mountain View, CA, United States of America

David M Bulfer