US 6,253,292 B1Granted

Distributed shared memory multiprocessor system based on a unidirectional ring bus using a snooping scheme

OwnerSeong Tae Jhang, Chu Shik Jhon +1inventors

InventorsSeong Tae Jhang, Chu Shik Jhon, Hyung Ho Kim

Life of this patent

Estimating the term…

Loading patent…

Abstract

A distributed shared memory multiprocessor system based on a unidirectional ring bus using a snooping scheme comprises a group of processor nodes and a ring bus. The processor nodes are arranged in the form of a ring and one of the processor nodes generates a request signal for a data block, the remaining processor nodes snoop their own internal parts, and one of the remaining processor nodes provides the data block. The ring bus is used for connecting the processor nodes in the form of the ring and providing a path through which the request signal is broadcast to each of the remaining processor nodes and the data block is unicast to the processor node which has generated the request signal for the data block.

Claims13

Dependent claims

A distributed shared memory multiprocessor system comprising:

a group of processor nodes, wherein the processor nodes, each node including a storage, arc arranged in the form of a ring and one of the processor nodes generates a request signal for a data block, the remaining processor nodes snoop their own storages, and one of the remaining processor nodes provides the data block;
and
a ring bus for connecting the processor nodes in the form of the ring and providing a path through which the request signal is broadcast to each of the remaining processor nodes and the data block is provided to the processor node which has generated the request signal,
wherein each of the processor nodes further includes:
a plurality of processor modules for generating the request signal for the data block;
a node controller for providing the data block corresponding to the request signal, wherein the node controller provides the data block corresponding to the request signal stored at a local shared memory or the remote cache of the processor node if the data block corresponding to the request signal is stored at the local shared memory or the remote cache of the processor node, and the node controller transmits the request signal via the ring bus if the data block corresponding to the request signal is not stored at the local shared memory or the remote cache of the processor node;
provides the data block corresponding to the request signal to the processor module of the processor node which requires the data block if the data block corresponding to the request signal is provided from one of the remaining processor nodes;
and examines whether the data block corresponding to the request signal is stored at its remote cache or the local shared memory and provides the data block if the request signal is provided from one of the remaining processor nodes;
and
a ring interface for connecting the node controller to the ring bus,
wherein the storage contains:
a local shared memory for storing data blocks, wherein the data blocks are shared by each of the processor modules;
and
a remote cache for storing the data block corresponding to the request signal, and
wherein the local shared memory contains:
a data memory for storing contents of the data block;
a memory controller, connected to a local system bus, for providing the data block stored at the data memory to the local system bus if the request signal is provided from the local system bus;
and
a memory directory cache, connected to the node controller, for storing data on the state of the data block stored at the data memory.

16 more elements

A distributed shared memory multiprocessor system comprising:

a group of processor nodes, wherein the processor nodes, each node including a storage, are arranged in the form of a ring and one of the processor nodes generates a request signal for a data block, the remaining processor nodes snoop their own storages, and one of the remaining processor nodes provides the data block;
and
a ring bus for connecting the processor nodes in the form of the ring and providing a path through which the request signal is broadcast to each of the remaining processor nodes and the data block is provided to the processor node which has generated the request signal,
wherein each of the processor nodes further includes:
a plurality of processor modules for generating the request signal for the data block;
a node controller for providing the data block corresponding to the request signal, wherein the node controller provides the data block corresponding to the request signal stored at a local shared memory or the remote cache of the processor node if the data block corresponding to the request signal is stored at the local shared memory or the remote cache of the processor node, and the node controller transmits the request signal via the ring bus if the data block corresponding to the request signal is not stored at the local shared memory or the remote cache of the processor node;
provides the data block corresponding to the request signal to the processor module of the processor node which requires the data block if the data block corresponding to the request signal is provided from one of the remaining processor nodes;
and examines whether the data block corresponding to the request signal is stored at its remote cache or the local shared memory and provides the data block if the request signal is provided from one of the remaining processor nodes;
a ring interface for connecting the node controller to the ring bus;
and
a pending buffer for storing information on a data block which is transmitted via the ring bus if it is not known whether the processor node requiring the data block has received the data block or not,
wherein the storage contains:
a local shared memory for storing data blocks, wherein the data blocks are shared by each of the processor modules;
and
a remote cache for storing the data block corresponding to the request signal.

12 more elements

A distributed shared memory multiprocessor system comprising:

a group of processor nodes, wherein the processor nodes, each node including a storage, are arranged in the form of a ring and one of the processor nodes generates a request signal for a data block, the remaining processor nodes snoop their own storages, and one of the remaining processor nodes provides the data block;
and
a ring bus for connecting the processor nodes in the form of the ring and providing a path through which the request signal is broadcast to each of the remaining processor nodes and the data block is provided to the processor node which has generated the request signal,
wherein each of the processor nodes further includes:
a plurality of processor modules for generating the request signal for the data block;
a node controller for providing the data block corresponding to the request signal, wherein the node controller provides the data block corresponding to the request signal stored at a local shared memory or the remote cache of the processor node if the data block corresponding to the request signal is stored at the local shared memory or the remote cache of the processor node, and the node controller transmits the request signal via the ring bus if the data block corresponding to the request signal is not stored at the local shared memory or the remote cache of the processor node;
provides the data block corresponding to the request signal to the processor module of the processor node which requires the data block if the data block corresponding to the request signal is provided from one of the remaining processor nodes;
and examines whether the data block corresponding to the request signal is stored at its remote cache or the local shared memory and provides the data block if the request signal is provided from one of the remaining processor nodes, and
a ring interface for connecting the node controller to the ring bus,
wherein the storage contains:
a local shared memory for storing data blocks, wherein the data blocks are shared by each of the processor modules;
and
a remote cache for storing the data block corresponding to the request signal,
wherein the remote cache contains:
a remote data cache for storing contents of the data block;
and
a remote tag cache section for storing a tag address and a state of the data block stored at the remote data cache, and
wherein the remote tag cache section is provided with two or more independent remote tag cache.

15 more elements

A distributed shared memory multiprocessor system comprising:

a group of processor nodes, wherein the processor nodes, each node including a storage, are arranged in the form of a ring and one of the processor nodes generates a request signal for a data block, the remaining processor nodes snoop their own storages, and one of the remaining processor nodes provides the data block;
and
a ring bus for connecting the processor nodes in the form of the ring and providing a path through which the request signal is broadcast to each of the remaining processor nodes and the data block is provided to the processor node which has generated the request signal,
wherein each of the processor nodes further includes:
a plurality of processor modules for generating the request signal for the data block;
a node controller for providing the data block corresponding to the request signal, wherein the node controller provides the data block corresponding to the request signal stored at a local shared memory or the remote cache of the processor node if the data block corresponding to the request signal is stored at the local shared memory or the remote cache of the processor node, and the node controller transmits the request signal via the ring bus if the data block corresponding to the request signal is not stored at the local shared memory or the remote cache of the processor node;
provides the data block corresponding to the request signal to the processor module of the processor node which requires the data block if the data block corresponding to the request signal is provided from one of the remaining processor nodes;
and examines whether the data block corresponding to the request signal is stored at its remote cache or the local shared memory and provides the data block if the request signal is provided from one of the remaining processor nodes, and
a ring interface for connecting the node controller to the ring bus,
wherein the storage contains:
a local shared memory for storing data blocks, wherein the data blocks are shared by each of the processor modules;
and
a remote cache for storing the data block corresponding to the request signal, and
wherein the local shared memory contains:
a data memory for storing contents of the data block;
a memory controller, connected to the node controller, for providing the data block stored at the data memory to the node controller if the request signal is provided from the node controller;
and
a memory directory cache, corrected to the node controller, for storing data on the state of the data block stored at the data memory.

15 more elements

Dependencies are inferred from the claim text.