Parallel processor system including a cache memory subsystem that has independently addressable local and remote data areas

Claim 2

A parallel processor network system, comprising: a plurality of nodes each including a processor, a distributed shared memory accessible by any of the processors, and a cache memory subsystem, each cache memory subsystem including a cache memory divided address-wise into respective local and remote data areas for holding local and remote data to be stored in one of the shared memories, respectively; wherein each of the cache systems further includes means for determining whether an access request transmitted by one of the processors is intended for a local node or for a remote node, and means for addressably accessing either the local data area or the remote data area depending upon whether the access request is intended for a local node or a remote node, respectively, wherein an access request for accessing a remote node only affects the remote data in the remote data area without affecting the local data in the local data area.

Claim 3

A parallel processor network system according to claim 1, wherein the access request contains a header designating the local data area or the remote data area of one of the nodes.

Claim 4

A parallel processor network system according to claim 1, wherein each cache memory subsystem further includes a cache memory controller and an address array connected thereto for selectively addressing a local data area or remote data area based upon the access request.

Claim 5

A parallel processor network system according to claim 1, wherein each cache memory subsystem includes an addressable access discriminating circuit for selecting a local data area or a remote data area based upon the access request.

Claim 6

A parallel processor network system, comprising: a plurality of nodes each including a processor, a shared memory accessible by any of the processors, and a cache system including independently addressable local and remote data caches for respectively holding local and remote data to be stored in one of the shared memories; wherein each cache system further includes means for determining whether an access request transmitted by one of the processors is intended for a local node or for a remote node, and means for addressably accessing either the local data area or the remote data area depending upon whether the access request is intended for a local node or a remote node, respectively, wherein an access request for accessing a remote node only affects the remote data in the remote data area without affecting the local data in the local data area.

Claim 7

A parallel processor network system, comprising: a plurality of nodes each including a processor, a shared memory accessible by any of the processors, and a cache system including independently addressable local and remote data caches for respectively holding local and remote data to be stored in one of the shared memories; wherein each cache system further includes means for determining whether an access request transmitted by one of the processors is intended for a local node or for a remote node, and means for addressably accessing either the local data area or the remote data area depending upon whether the access request is intended for a local node or a remote node, respectively, wherein an access request for accessing a remote node only affects the remote data in the remote data area without affecting the local data in the local data area.

Claim 8

A parallel processor network system according to claim 5, wherein the access request contains a header designating the local data area or the remote data area of one of the nodes.

Claim 9

A parallel processor network system according to claim 5, wherein each cache system further includes a cache memory controller and an address array connected thereto for selectively addressing a local data area or remote data area based upon the access request.

Claim 10

A parallel processor network system according to claim 5, wherein each cache system includes an addressable access discriminating circuit for addressably selecting a local data area or a remote data area based upon the access request.

Claim 11

A processor element for use in a parallel processor network system, comprising: a processor; a local-remote divided cache memory subsystem, including means for receiving a data transfer request from the processor, means for determining whether the data transfer request concerns local data or remote data, a local data area for holding local data, a remote data area for holding remote data, and means for addressably accessing the local data area or the remote data area from information contained in the data transfer request; a remote access request circuit for transmitting an access request to an accessed remote processor element responsive to a signal received from the local-remote divided cache memory subsystem, and for receiving data from the accessed remote processor element as a result of the transmitted access request; a remote access reply circuit for receiving an access request from a requesting remote processor element, and for transferring data to the requesting remote processor element as a result of the received access request; a shared memory accessible by said local-remote divided cache memory subsystem and by said remote access reply circuit; and a selector for selectively providing access to the shared memory to one of the local-remote divided cache memory subsystem and the remote access reply circuit; wherein a data transfer request for remote data only affects the remote data in the remote data area without affecting the local data in the local data area.

Claim 12

A processor element for use in a parallel processor network system, comprising: a processor; a local-remote divided cache memory subsystem, including means for receiving a data transfer request from the processor, means for determining whether the data transfer request concerns local data or remote data, a local data area for holding local data, a remote data area for holding remote data, and means for addressably accessing the local data area or the remote data area from information contained in the data transfer request; a remote access request circuit for transmitting an access request to an accessed remote processor element responsive to a signal received from the local-remote divided cache memory subsystem, and for receiving data from the accessed remote processor element as a result of the transmitted access request; a remote access reply circuit for receiving an access request from a requesting remote processor element, and for transferring data to the requesting remote processor element as a result of the received access request; a shared memory accessible by said local-remote divided cache memory subsystem and by said remote access reply circuit; and a selector for selectively providing access to the shared memory to one of the local-remote divided cache memory subsystem and the remote access reply circuit; wherein a data transfer request for remote data only affects the remote data in the remote data area without affecting the local data in the local data area.

Claim 13

A processor element as claimed in claim 9, wherein the means for addressably accessing the local data area or remote data area includes a cache memory controller and an address array, the address array accessing the local data area or remote data area in response to a cache control signal received from the cache memory controller; and wherein the local-remote divided cache memory subsystem further includes a local memory access circuit for receiving a shared memory access request signal from the cache memory controller and for accessing the shared memory via the selector in response to the shared memory access request signal, and a remote memory access circuit for receiving a remote memory access request signal from the cache memory controller and for transmitting the access request to the remote processor element in response to the remote memory access request signal.

Claim 14

A processor element as claimed in claim 10, wherein the means for determining whether the data transfer request concerns local data or remote data includes a mask for receiving an address from the data transfer request received from the processor and for logically processing part of the address to output a mask information signal, a processor element number register for holding a processor element number received from the processor that indicates the processor element to which data are to be transferred to or from, and a comparator for receiving the processor element number from the processor element number register and the masked information and for outputting a local-remote decision signal indicating whether the request concerns local data or remote data; and wherein the local-remote divided cache memory subsystem further includes an address merge circuit for receiving the local-remote decision signal and the address, and for outputting a merged address to the cache memory controller, said merged address being employed by the cache memory controller to addressably access the local data area or the remote data area.

Claim 15

A processor element as claimed in claim 11, wherein the local-remote decision signal is a single binary bit, and wherein the address merge circuit embeds the local-remote decision bit at a predefined bit location in the address to create the merged address, said bit location dividing the bits of the address into higher-order and lower-order bits, the local-remote decision bit and the lower-order bits together defining a local data area address, and the higher-order bits defining a remote data area address.

Claim 16

A processor element as claimed in claim 9, wherein the means for determining whether the data transfer request concerns local data or remote data includes a mask for receiving an address from the data transfer request received from the processor and for logically processing part of the address to produce and output a mask information signal, a processor element number register for holding a processor element number received from the processor that indicates the processor element to which data are to be transferred to or from, and a comparator for receiving the processor element number from the processor element number register and the masked information and for outputting a local-remote decision signal indicating whether the request concerns local data or remote data.

Claim 17

A processor element as claimed in claim 13, wherein the part of the address logically processed by the mask to produce the mask information signal is the processor element number held by the processor element number register.

Claim 18

A parallel processor network system, comprising: a plurality of processor elements mutually connected by a network, each processor element having: a processor; a local-remote divided cache memory subsystem, including means for receiving a data transfer request from the processor, means for determining whether the data transfer request concerns local data or remote data, a local data area for holding local data, a remote data area for holding remote data, and means for addressably accessing the local data area or the remote data area from information contained in the data transfer request; a remote access request circuit for transmitting an access request to an accessed remote processor element responsive to a signal received from the local-remote divided cache memory subsystem, and for receiving data from the accessed remote processor element as a result of the transmitted access request; a remote access reply circuit for receiving an access request from a requesting remote processor element, and for transferring data to the requesting remote processor element as a result of the received access request; a shared memory accessible by said local-remote divided cache memory subsystem and by said remote access reply circuit; and a selector for selectively providing access to the shared memory to one of the local-remote divided cache memory subsystem and the remote access reply circuit; wherein a data transfer request for remote data only affects the remote data in the remote data area without affecting the local data in the local data area.

Claim 19

A parallel processor network system, comprising: a plurality of processor elements mutually connected by a network, each processor element having: a processor; a local-remote divided cache memory subsystem, including means for receiving a data transfer request from the processor, means for determining whether the data transfer request concerns local data or remote data, a local data area for holding local data, a remote data area for holding remote data, and means for addressably accessing the local data area or the remote data area from information contained in the data transfer request; a remote access request circuit for transmitting an access request to an accessed remote processor element responsive to a signal received from the local-remote divided cache memory subsystem, and for receiving data from the accessed remote processor element as a result of the transmitted access request; a remote access reply circuit for receiving an access request from a requesting remote processor element, and for transferring data to the requesting remote processor element as a result of the received access request; a shared memory accessible by said local-remote divided cache memory subsystem and by said remote access reply circuit; and a selector for selectively providing access to the shared memory to one of the local-remote divided cache memory subsystem and the remote access reply circuit; wherein a data transfer request for remote data only affects the remote data in the remote data area without affecting the local data in the local data area.

Claim 20

A parallel processor network system as claimed in claim 15, wherein the means for addressably accessing the local data area or remote data area includes a cache memory controller and an address array, the address array accessing the local data area or remote data area in response to a cache control signal received from the cache memory controller; and wherein the local-remote divided cache memory subsystem further includes a local memory access circuit for receiving a shared memory access request signal from the cache memory controller and for accessing the shared memory via the selector in response to the shared memory access request signal, and a remote memory access circuit for receiving a remote memory access request signal from the cache memory controller and for transmitting the access request to the remote processor element in response to the remote memory access request signal.

Claim 21

A parallel processor network system as claimed in claim 16, wherein the means for determining whether the data transfer request concerns local data or remote data includes a mask for receiving an address from the data transfer request received from the processor and for logically processing part of the address to output a mask information signal, a processor element number register for holding a processor element number received from the processor that indicates the processor element to which data are to be transferred to or from, and a comparator for receiving the processor element number from the processor element number register and the masked information and for outputting a local-remote decision signal indicating whether the request concerns local data or remote data; and wherein the local-remote divided cache memory subsystem further includes an address merge circuit for receiving the local-remote decision signal and the address, and for outputting a merged address to the cache memory controller, said merged address being employed by the cache memory controller to addressably access the local data area or the remote data area.

Claim 22

A parallel processor network system as claimed in claim 17, wherein the local-remote decision signal is a single binary bit, and wherein the address merge circuit embeds the local-remote decision bit at a predefined bit location in the address to create the merged address, said bit location dividing the bits of the address into higher-order and lower-order bits, the local-remote decision bit and the lower-order bits together defining a local data area address, and the higher-order bits defining a remote data area address.

Claim 23

A parallel processor network system as claimed in claim 15, wherein the means for determining whether the data transfer request concerns local data or remote data includes a mask for receiving an address from the data transfer request received from the processor and for logically processing part of the address to produce and output a mask information signal, a processor element number register for holding a processor element number received from the processor that indicates the processor element to which data are to be transferred to or from, and a comparator for receiving the processor element number from the processor element number register and the masked information and for outputting a local-remote decision signal indicating whether the request concerns local data or remote data.

Claim 24

A parallel processor network system as claimed in claim 19, wherein the part of the address logically processed by the mask to produce the mask information signal is the processor element number held by the processor element number register.

Claim 25

A processor element for use in a parallel processor network system, comprising: a processor; address discriminating means for receiving a data transfer request from the processor, and for outputting a selection signal indicating whether the data transfer request concerns local data of the processor element or remote data of a remote processor element of the network system; a local data cache subsystem for holding local data; a remote data cache subsystem for holding remote data; and selecting means for selecting one of the local data cache subsystem and the remote data cache subsystem responsive to the selection signal; wherein when said selecting means selects the remote data cache subsystem, only the remote data are affected without affecting the local data in the local data cache subsystem.

Claim 26

A processor element for use in a parallel processor network system, comprising: a processor; address discriminating means for receiving a data transfer request from the processor, and for outputting a selection signal indicating whether the data transfer request concerns local data of the processor element or remote data of a remote processor element of the network system; a local data cache subsystem for holding local data; a remote data cache subsystem for holding remote data; and selecting means for selecting one of the local data cache subsystem and the remote data cache subsystem responsive to the selection signal; wherein when said selecting means selects the remote data cache subsystem, only the remote data are affected without affecting the local data in the local data cache subsystem.

Claim 27

A processor element as claimed in claim 21, further comprising: a remote access request circuit for transmitting an access request to an accessed remote processor element responsive to a signal received from the remote data cache subsystem, and for receiving data from the accessed remote processor element as a result of the transmitted access request; a remote access reply circuit for receiving an access request from a requesting remote processor element, and for transferring data to the requesting remote processor element as a result of the received access request; a shared memory accessible by said local data cache subsystem and by said remote access reply circuit; and a selector for selectively providing access to the shared memory to one of the local data cache subsystem and the remote access reply circuit.

Claim 28

A processor element as claimed in claim 21, wherein the address discriminating means includes a mask for receiving an address from the data transfer request received from the processor and for logically processing part of the address to produce and output a mask information signal, a processor element number register for holding a processor element number received from the processor that indicates the processor element to which data are to be transferred to or from, and a comparator for receiving the processor element number from the processor element number register and the masked information and for outputting the selection signal.

Claim 29

A processor element as claimed in claim 23, wherein the address discriminating circuit further includes: a local latch, activated by the selection signal when the selection signal indicates that the data transfer request concerns local data, for accessing the local data cache subsystem; and a remote latch, activated by the selection signal when the selection signal indicates that the data transfer request concerns remote data, for accessing the remote data cache subsystem.

Claim 30

A processor element as claimed in claim 21, wherein the local data cache subsystem includes a cache memory controller, an address array for accessing the local data area in response to a cache control signal received from the cache memory controller, and a local memory access circuit for receiving a memory access request signal from the cache memory controller and for accessing the shared memory in response thereto; and wherein the remote data cache subsystem includes a cache memory controller, an address array for accessing the remote data area in response to a cache control signal received from the cache memory controller, and a remote memory access circuit for receiving a remote memory access request signal from the cache memory controller and for transmitting the access request to the remote processor element in response thereto.

Claim 31

A processor element for use in a parallel processor network system, comprising: a processor; a local data cache subsystem, including a local cache memory for holding local data, and first address discriminating means for receiving a data transfer request from the processor and for addressing the local cache memory when the data transfer request concerns local data of the processor element; and a remote data cache subsystem, including a remote cache memory for holding remote data, and second address discriminating means for receiving a data transfer request from the processor and for addressing the remote cache memory when the data transfer request concerns remote data of a remote processor element of the network system; wherein, when the data transfer request concerns the remote data, only the remote data are affected without affecting the local data in the local cache memory.

Claim 32

A processor element for use in a parallel processor network system, comprising: a processor; a local data cache subsystem, including a local cache memory for holding local data, and first address discriminating means for receiving a data transfer request from the processor and for addressing the local cache memory when the data transfer request concerns local data of the processor element; and a remote data cache subsystem, including a remote cache memory for holding remote data, and second address discriminating means for receiving a data transfer request from the processor and for addressing the remote cache memory when the data transfer request concerns remote data of a remote processor element of the network system; wherein, when the data transfer request concerns the remote data, only the remote data are affected without affecting the local data in the local cache memory.

Claim 33

A processor element as claimed in claim 26, further comprising: a remote access request circuit for transmitting an access request to an accessed remote processor element responsive to a signal received from the remote data cache subsystem, and for receiving data from the accessed remote processor element as a result of the transmitted access request; a remote access reply circuit for receiving an access request from a requesting remote processor element, and for transferring data to the requesting remote processor element as a result of the received access request; a shared memory accessible by said local data cache subsystem and by said remote access reply circuit; and a selector for selectively providing access to the shared memory to one of the local data cache subsystem and the remote access reply circuit.

Claim 34

A processor element as claimed in claim 26, wherein each of the first and second address discriminating means includes a mask for receiving an address from the data transfer request received from the processor and for logically processing part of the address to produce and output a mask information signal, a processor element number register for holding a processor element number received from the processor that indicates the processor element to which data are to be transferred to or from, and a comparator for receiving the processor element number from the processor element number register and the masked information and for outputting a signal for addressing the respective cache memory.

Claim 35

A processor element as claimed in claim 26, wherein the first address discriminating means includes a first latch for receiving the data transfer request and for providing a signal for accessing the local cache memory; and wherein the second address discriminating means includes a second latch for receiving the data transfer request and for providing a signal for accessing the remote cache memory.

Claim 36

A processor element as claimed in claim 29, wherein the local data cache subsystem includes a cache memory controller, an address array for accessing the local cache memory in response to a cache control signal received from the cache memory controller, and a local memory access circuit for receiving a memory access request signal from the cache memory controller and for accessing the shared memory in response thereto; and wherein the remote data cache subsystem includes a cache memory controller, an address array for accessing the remote cache memory in response to a cache control signal received from the cache memory controller, and a remote memory access circuit for receiving a remote memory access request signal from the cache memory controller and for transmitting the access request to the remote processor element in response thereto.

Claim 37

A parallel processor network system, comprising: a plurality of processor elements mutually connected by a network, each processor element having: a processor; address discriminating means for receiving a data transfer request from the processor, and for outputting a selection signal indicating whether the data transfer request concerns local data of the processor element or remote data of a remote processor element of the network system; a local data cache subsystem for holding local data; a remote data cache subsystem for holding remote data; and selecting means for selecting one of the local data cache subsystem and the remote data cache subsystem responsive to the selection signal; wherein, when said selecting means selects the remote data cache subsystem, only the remote data are affected without affecting the local data in the local data cache subsystem.

Claim 38

A parallel processor network system, comprising: a plurality of processor elements mutually connected by a network, each processor element having: a processor; address discriminating means for receiving a data transfer request from the processor, and for outputting a selection signal indicating whether the data transfer request concerns local data of the processor element or remote data of a remote processor element of the network system; a local data cache subsystem for holding local data; a remote data cache subsystem for holding remote data; and selecting means for selecting one of the local data cache subsystem and the remote data cache subsystem responsive to the selection signal; wherein, when said selecting means selects the remote data cache subsystem, only the remote data are affected without affecting the local data in the local data cache subsystem.

Claim 39

A parallel processor network system as claimed in claim 31, each said processor element further comprising: a remote access request circuit for transmitting an access request to an accessed remote processor element responsive to a signal received from the remote data cache subsystem, and for receiving data from the accessed remote processor element as a result of the transmitted access request; a remote access reply circuit for receiving an access request from a requesting remote processor element, and for transferring data to the requesting remote processor element as a result of the received access request; a shared memory accessible by said local data cache subsystem and by said remote access reply circuit; and a selector for selectively providing access to the shared memory to one of the local data cache subsystem and the remote access reply circuit.

Claim 40

A parallel processor network system as claimed in claim 31, wherein the address discriminating means includes a mask for receiving an address from the processor and for logically processing part of the address to produce and output a mask information signal, a processor element number register for holding a processor element number received from the processor that indicates the processor element to which data are to be transferred to or from, and a comparator for receiving the processor element number from the processor element number register and the masked information and for outputting the selection signal.

Claim 41

A parallel processor network system as claimed in claim 33, wherein the address discriminating circuit further includes: a local latch, activated by the selection signal when the selection signal indicates that the data transfer request concerns local data, for accessing the local data cache subsystem; and a remote latch, activated by the selection signal when the selection signal indicates that the data transfer request concerns remote data, for accessing the remote data cache subsystem.

Claim 42

A parallel processor network system as claimed in claim 31, wherein the local data cache subsystem includes a cache memory controller, an address array for accessing the local data area in response to a cache control signal received from the cache memory controller, and a local memory access circuit for receiving a memory access request signal from the cache memory controller and for accessing the shared memory in response thereto; and wherein the remote data cache subsystem includes a cache memory controller, an address array for accessing the remote data area in response to a cache control signal received from the cache memory controller, and a remote memory access circuit for receiving a remote memory access request signal from the cache memory controller and for transmitting the access request to the remote processor element in response thereto.

Claim 43

A parallel processor network system, comprising: a plurality of processor elements mutually connected by a network, each processor element having: a processor; a local data cache subsystem, including a local cache memory for holding local data, and first address discriminating means for receiving a data transfer request from the processor and for addressing the local cache memory when the data transfer request concerns local data of the processor element; and a remote data cache subsystem, including a remote cache memory for holding remote data, and second address discriminating means for receiving a data transfer request from the processor and for addressing the remote cache memory when the data transfer request concerns remote data of a remote processor element of the network system; wherein, when the data transfer request concerns the remote data, only the remote data are affected without affecting the local data in the local cache memory.

Claim 44

A parallel processor network system, comprising: a plurality of processor elements mutually connected by a network, each processor element having: a processor; a local data cache subsystem, including a local cache memory for holding local data, and first address discriminating means for receiving a data transfer request from the processor and for addressing the local cache memory when the data transfer request concerns local data of the processor element; and a remote data cache subsystem, including a remote cache memory for holding remote data, and second address discriminating means for receiving a data transfer request from the processor and for addressing the remote cache memory when the data transfer request concerns remote data of a remote processor element of the network system; wherein, when the data transfer request concerns the remote data, only the remote data are affected without affecting the local data in the local cache memory.

Claim 45

A parallel processor network system as claimed in claim 36, each processor element further comprising: a remote access request circuit for transmitting an access request to an accessed remote processor element responsive to a signal received from the remote data cache subsystem, and for receiving data from the accessed remote processor element as a result of the transmitted access request; a remote access reply circuit for receiving an access request from a requesting remote processor element, and for transferring data to the requesting remote processor element as a result of the received access request; a shared memory accessible by said local data cache subsystem and by said remote access reply circuit; and a selector for selectively providing access to the shared memory to one of the local data cache subsystem and the remote access reply circuit.

Claim 46

A parallel processor network system as claimed in claim 36, wherein each of the first and second address discriminating means includes a mask for receiving an address from the data transfer request received from the processor and for logically processing part of the address to produce and output a mask information signal, a processor element number register for holding a processor element number received from the processor that indicates the processor element to which data are to be transferred to or from, and a comparator for receiving the processor element number from the processor element number register and the masked information and for outputting a signal for addressing the respective cache memory.

Claim 47

A parallel processor network system as claimed in claim 36, wherein the first address discriminating means includes a first latch for receiving the data transfer request and for providing a signal for accessing the local cache memory; and wherein the second address discriminating means includes a second latch for receiving the data transfer request and for providing a signal for accessing the remote cache memory.

Claim 48

A parallel processor network system as claimed in claim 39, wherein the local data cache subsystem includes a cache memory controller, an address array for accessing the local cache memory in response to a cache control signal received from the cache memory controller, and a local memory access circuit for receiving a memory access request signal from the cache memory controller and for accessing the shared memory in response thereto; and wherein the remote data cache subsystem includes a cache memory controller, an address array for accessing the remote cache memory in response to a cache control signal received from the cache memory controller, and a remote memory access circuit for receiving a remote memory access request signal from the cache memory controller and for transmitting the access request to the remote processor element in response thereto.