Multi-clock SIMD computer and instruction-cache-enhancement thereof

Claim 2

A multi-clock single-instruction stream, multiple-data stream (SIMD) computer comprising: a global instruction broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; a multi-chip subsystem (MCS) regulated by second clock signal; and a processing element (PE) module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal: a second internal subsystem regulated by a fourth clock signal; and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals, wherein one of said third and fourth clock signals is a PE module clock signal oscillating at a PE module clock rate higher than said system clock, and wherein said second clock signal oscillates at a MCS clock rate higher than said system clock rate but no higher than said PE module clock rate.

Claim 3

A multi-clock SIMD computer comprising: a globel instruction broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; an MCS regulated by a second clock signal; and a PE module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal; a second internal subsystem regulated by a fourth clock signal; and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals, wherein one of said third and fourth clock signals is a PE module clock signal oscillating at a PE module clock rate higher than said system clock, wherein said second internal subsystem includes a processing element (PE) that is regulated by a PE clock signal, and wherein said plurality of clock signals includes said PE clock signal oscillating at a PE clock rate higher than said system clock rate but no higher than said PE module clock rate.

Claim 4

A multi-clock SIMD computer comprising: a globel instruction broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; an MCS regulated by a second clock signal; and a PE module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal; a second internal subsystem regulated by a fourth clock signal; and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals, wherein one of said third and fourth clock signals is a PE module clock signal oscillating at a PE module clock rate higher than said system clock, wherein said second internal subsystem includes a processing element (PE) that is regulated by a PE clock signal, and wherein said plurality of clock signals includes said PE clock signal oscillating at a PE clock rate higher than said system clock rate but no higher than said PE module clock rate.

Claim 5

A multi-clock SIMD computer comprising: a global instruction broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; an MCS regulated by a second clock signal; and a PE module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal; a second internal subsystem regulated by a fourth clock signal; and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals, wherein said first clock signal oscillates at a rate different from one of said second, third, and fourth clock signals, and wherein said plurality of clock signals includes one clock signal oscillating in-phase with said system clock signal.

Claim 6

A multi-clock SIMD computer comprising: a global instruction broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; an MCS regulated by a second clock signal; and a PE module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal; a second internal subsystem regulated by a fourth clock signal; and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals, wherein said first clock signal oscillates at a rate different from one of said second, third, and fourth clock signals, and wherein said plurality of clock signals includes one clock signal oscillating in-phase with said system clock signal.

Claim 7

A multi-clock SIMD computer comprising: a global instruction, broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; an MCS regulated by a second clock signal; and a PE module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal; a second internal subsystem regulated by a fourth clock signal: and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals. Wherein one of said third and fourth clock signals is a PE module clock signal oscillating at a PE module clock rate higher than said system clock, and wherein a first one of said plurality of clock signals may be restarted during any cycle of said PE module clock signal.

Claim 8

A multi-clock SIMD computer comprising: a global instruction, broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; an MCS regulated by a second clock signal; and a PE module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal; a second internal subsystem regulated by a fourth clock signal: and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals. Wherein one of said third and fourth clock signals is a PE module clock signal oscillating at a PE module clock rate higher than said system clock, and wherein a first one of said plurality of clock signals may be restarted during any cycle of said PE module clock signal.

Claim 9

A multi-clock SIMD computer according to claim 4, wherein a second one of said plurality of clock signals generated by said multi-clock generating circuit is restartable in synchrony with said system clock signal.

Claim 10

A multi-clock SlMD computer comprising: a global instructing broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; an MCS regulated by a second clock signal; and a PE module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal; a second internal subsystem regulated by a fourth clock signal; and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals, wherein one of said third and fourth clock signals is a PE module clock signal oscillating at a PE module clock rate higher than said system clock; wherein said global instruction broadcast subsystem includes: a system controller for providing a first sequence of instructions to said PE module; and a global instruction broadcast network, coupled to said system controller, for broadcasting said first sequence of instructions from said system controller to said PE module at said system clock rate; and wherein said first internal subsystem is regulated by said PE module clock signal and includes a sub-broadcast-interval instruction sequencer, coupled to said global instruction broadcast network, for supplying said first sequence of instructions to said second internal subsystem at said PE module clock rate.

Claim 11

A multi-clock SlMD computer comprising: a global instructing broadcast subsystem regulated by a system clock signal oscillating at a system clock rate; an MCS regulated by a second clock signal; and a PE module coupled to said global instruction broadcast subsystem and to said MCS, said PE module implemented on a single chip and including: a first internal subsystem regulated by a third clock signal; a second internal subsystem regulated by a fourth clock signal; and a multi-clock generating circuit, responsive to said system clock signal, for generating a plurality of clock signals including said second, third, and fourth clock signals, wherein one of said third and fourth clock signals is a PE module clock signal oscillating at a PE module clock rate higher than said system clock; wherein said global instruction broadcast subsystem includes: a system controller for providing a first sequence of instructions to said PE module; and a global instruction broadcast network, coupled to said system controller, for broadcasting said first sequence of instructions from said system controller to said PE module at said system clock rate; and wherein said first internal subsystem is regulated by said PE module clock signal and includes a sub-broadcast-interval instruction sequencer, coupled to said global instruction broadcast network, for supplying said first sequence of instructions to said second internal subsystem at said PE module clock rate.

Claim 12

A multi-clock SIMD computer according to claim 6, wherein said first internal subsystem further comprises a memory for storing said first sequence of instructions provided by said global instruction broadcast subsystem.

Claim 13

A multi-clock SIMD computer according to claim 7, Wherein said first internal subsystem further comprises a memory controller for providing addresses and read/Write control signals to said memory responsive to the global instruction broadcast subsystem.

Claim 14

A multi-clock SIMD computer according to claim 8, wherein said memory controller is responsive to an instruction issued by said global instruction broadcast subsystem which directs said memory controller to write said first sequence of instructions subsequently broadcast over said global instruction broadcast network into said memory, and wherein said memory controller writes said first sequence of instructions into said memory at said system clock rate.

Claim 15

A multi-clock SIMD computer according to claim 8, wherein said memory controller is responsive to an instruction issued by said global instruction broadcast subsystem which directs said memory controller to cease writing said first sequence of instructions received from said global instruction broadcast network into said memory.

Claim 16

A multi-clock SIMD computer according to claim 8, wherein said memory controller is responsive to an instruction issued by said global instruction broadcast subsystem which directs said memory controller to retrieve said first sequence of instruction stored in said memory, wherein said memory controller retrieves said first sequence of instructions from said memory at said PE module clock rate.

Claim 17

A multi-clock SIMD computer according to claim 9, wherein said memory controller is responsive to a second instruction issued by said global instruction broadcast subsystem which directs said memory controller to write a predetermined number of instructions broadcast over said global instruction broadcast network into said memory.

Claim 18

A multi-clock SIMD computer according to claim 10, wherein said memory controller recognizes a second instruction issued by said global instruction broadcast subsystem to be a delimiter for said first sequence of instructions being written into said memory.

Claim 19

A multi-clock SlMD computer according to claim 11, wherein said memory controller is responsive to a second instruction issued by the global instruction broadcast subsystem which directs said memory controller to retrieve a predetermined number of instructions from said memory.

Claim 20

A multi-clock SIMD computer according to claim 11, wherein said memory controller is responsive to a second instruction issued by said global instruction broadcast subsystem which directs said memory controller to retrieve said first sequence of instructions from a predetermined address in said memory.

Claim 21

A multi-clock SIMD computer according to claim 11, wherein said memory controller is responsive to a second instruction issued by said global instruction broadcast subsystem which directs said memory controller to retrieve a predetermined number of iterations of said first sequence of instructions stored in said memory.

Claim 22

A multi-clock SIMD computer according to claim 15, wherein: said memory includes said first sequence of instructions and a second sequence of instructions stored therein; and said memory controller is responsive to an instruction retrieved from said first sequence of instructions which directs said memory controller to jump to a predetermined address in said memory to retrieve said second sequence of instructions.

Claim 23

A multi-clock SIMD computer according to claim 8, wherein: said memory comprises a first access port and a second access port; and wherein said memory controller concurrently reads a first sequence of instructions from said first port of said memory while writing a second sequence of instructions to said second port of said memory,

Claim 24

A multi-clock SIMD computer according to claim 11 wherein: said memory controller is responsive to a second instruction issued by said global instruction broadcast subsystem which directs said memory controller to retrieve said first sequence of instructions from a predetermined address in said memory; and said memory controller is responsive to a third instruction issued by said global instruction broadcast subsystem which directs said memory controller to retrieve a predetermined number of iterations of said first sequence of instructions stored in said memory.

Claim 25

A processing element (PE) module for use in a multi-clock, single-instruction stream, multiple-data stream (SIMD) computer having a system controller, a global instruction broadcast network, and an inter-PE communication network (ICN), and being regulated by a system clock signal oscillating at a system clock rate, said PE module comprising: a plurality of processing elements all regulated by a PE module clock signal; a multi-chip subsystem (MCS) access controller, coupled to said plurality of processing elements, for selectively coupling said plurality of processing elements to said ICN, wherein said MCS access controller is regulated by said PE module clock signal; a local controller coupled to said global instruction broadcast network, said local controller including: an instruction decoder circuit, responsive to an instruction from said global instruction broadcast network, for generating a decoded instruction and an ICN control signal, wherein said instruction decoder circuit is regulated by said PE module clock signal; an ICN control re-timer circuit, coupled to said instruction decoder circuit, for receiving said ICN control signal from said instruction decoder circuit and retransmitting Said ICN control signal to said MCS access controller, wherein said ICN control signal re-timer circuit is regulated by an ICN clock signal; and a multi-clock generator circuit, responsive to said system clock signal, for generating a plurality of clock signals, wherein said plurality of clock signals includes said PE module clock signal oscillating at a PE module clock rate and said ICN clock signal oscillating at a ICN clock rate, wherein said PE module clock rate and said ICN clock rate are no lower than said system clock rate; and a local instruction broadcast network, coupled to said local controller, said plurality of processing elements, and said MCS access controller, for transferring said decoded instruction from said local controller to said plurality of processing elements and said ICN control signal from said local controller to said MCS access controller.

Claim 26

A processing element (PE) module for use in a multi-clock, single-instruction stream, multiple-data stream (SIMD) computer having a system controller, a global instruction broadcast network, and an inter-PE communication network (ICN), and being regulated by a system clock signal oscillating at a system clock rate, said PE module comprising: a plurality of processing elements all regulated by a PE module clock signal; a multi-chip subsystem (MCS) access controller, coupled to said plurality of processing elements, for selectively coupling said plurality of processing elements to said ICN, wherein said MCS access controller is regulated by said PE module clock signal; a local controller coupled to said global instruction broadcast network, said local controller including: an instruction decoder circuit, responsive to an instruction from said global instruction broadcast network, for generating a decoded instruction and an ICN control signal, wherein said instruction decoder circuit is regulated by said PE module clock signal; an ICN control re-timer circuit, coupled to said instruction decoder circuit, for receiving said ICN control signal from said instruction decoder circuit and retransmitting Said ICN control signal to said MCS access controller, wherein said ICN control signal re-timer circuit is regulated by an ICN clock signal; and a multi-clock generator circuit, responsive to said system clock signal, for generating a plurality of clock signals, wherein said plurality of clock signals includes said PE module clock signal oscillating at a PE module clock rate and said ICN clock signal oscillating at a ICN clock rate, wherein said PE module clock rate and said ICN clock rate are no lower than said system clock rate; and a local instruction broadcast network, coupled to said local controller, said plurality of processing elements, and said MCS access controller, for transferring said decoded instruction from said local controller to said plurality of processing elements and said ICN control signal from said local controller to said MCS access controller.

Claim 27

A PE module as set forth in claim 20, wherein said instruction decoder circuit further comprises: a receive-and-hold circuit that receives said instruction from said global instruction broadcast signal and holds said instruction until a new instruction is broadcast on the global instruction broadcast network; a sub-broadcast interval instruction sequencer, coupled to said receive-and-hold circuit, for receiving said instruction in said receive-and-hold circuit and re-transmitting said instruction, wherein said sub-broadcast interval instruction sequencer is regulated by said PE module clock signal; and an instruction decoder, coupled to said sub-broadcast interval instruction sequencer, that generates a decoded instruction and an ICN control signal from said instruction.

Claim 28

A PE module as set forth in claim 21, wherein said ICN control signal generated by said instruction decoder circuit comprises an ICN control word, and wherein said ICN control re-timer circuit comprises an ICN control register for storing said ICN control word.

Claim 29

A PE module as set forth in claim 21, wherein said instruction decoder circuit further comprises: a memory for storing a sequence of instructions transmitted over said .global instruction broadcast network; and a memory controller, coupled to said memory and to said receive-and-hold circuit, for retrieving said sequence of instructions stored in said memory, wherein said memory controller retrieves said sequence at said PE module clock rate responsive to an instruction held in said receive-and-hold circuit.

Claim 30

A multi-clock SlMD computer according to claim 17 wherein: said memory controller is responsive to a third instruction issued by said global instruction broadcast subsystem which directs said memory controller to retrieve a predetermined number of iterations of said first sequence of instructions stored in said memory; and said memory controller is responsive to a fourth instruction issued by said global instruction broadcast subsystem which directs said memory controller to retrieve a predetermined number of iterations of said second sequence of instructions stored in said memory.