Image compression coprocessor with data flow control and multiple processing units

Claim 2

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein said control unit means contains an instruction memory for holding at least first and second programs loaded from an external host, and means for causing the execution of instructions from said first program prior to completion of said second program.

Claim 3

The image compression coprocessor of claim 1 wherein each of said special purpose processing means have different, special purpose hardware.

Claim 4

The image compression coprocessor of claim 1 wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; wherein each of said data tokens includes from one to a plurality of blocks of data or data vectors; wherein each of said data tokens further includes a token descriptor identifying said token as a data token and indicating the number of blocks of data attached.

Claim 5

The image compression coprocessor of claim 1 wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; wherein said instructions are transferred to said plurality of processing means in packets separately from data or control signals for said processing means.

Claim 6

The image compression coprocessor of claim 1 wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; wherein a plurality of said processing means include: instruction register means, coupled to said bus, for holding one of said instructions; buffer means, coupled to said bus, for holding said at least one of said data tokens; processing logic means, coupled to said buffer means and said instruction register means, for performing said subset of steps; and state machine means, coupled to said instruction register means and said buffer means, for controlling the interfacing of said instruction register and buffer with said bus.

Claim 7

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a host interface port for coupling to a master host processor; and run length processor means, connected between said bus and said host interface port, for converting data between said token format and a run length format for communicating with said master host processor.

Claim 8

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a host interface port for coupling to a master host processor; and run length processor means, connected between said bus and said host interface port, for converting data between said token format and a run length format for communicating with said master host processor.

Claim 9

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a host interface port for coupling to a master host processor; and token interface means, coupled between said bus and said host interface port, for allowing said host processor to insert tokens directly and to view tokens on said bus.

Claim 10

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a host interface port for coupling to a master host processor; and token interface means, coupled between said bus and said host interface port, for allowing said host processor to insert tokens directly and to view tokens on said bus.

Claim 11

The image compression coprocessor of claim 1 further comprising: a video interface for coupling said bus to external video memory; and a processor interface for coupling said bus to an external host processor.

Claim 12

The image compression coprocessor of claim 1 wherein said control unit means further comprises bus arbitration means, coupled to said bus, for arbitrating use of said bus between said control unit means and said plurality of processing means.

Claim 13

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a semaphore register coupled to said control unit means; a plurality of semaphore instructions in said data flow program for modifying a count in said semaphore register upon the accessing of a token, comparing a value in said semaphore register with a value in a token and preventing a new data token from being accessed if said semaphore value exceeds a maximum value, and modifying said count upon a data token leaving said control unit.

Claim 14

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a semaphore register coupled to said control unit means; a plurality of semaphore instructions in said data flow program for modifying a count in said semaphore register upon the accessing of a token, comparing a value in said semaphore register with a value in a token and preventing a new data token from being accessed if said semaphore value exceeds a maximum value, and modifying said count upon a data token leaving said control unit.

Claim 15

The image compression coprocessor of claim 10 wherein a first semaphore instruction tests and decrements said count before each instruction requiring a data token, and a second semaphore instruction increments said count after each data token leaves said control unit.

Claim 16

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a semaphore register coupled to said control unit means; a plurality of instructions in said data flow program having a control field with a number; and means for comparing said number in said control field in an instruction with a count in said semaphore register and passing said instruction through said control unit means for processing when said number matches said count.

Claim 17

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a semaphore register coupled to said control unit means; a plurality of instructions in said data flow program having a control field with a number; and means for comparing said number in said control field in an instruction with a count in said semaphore register and passing said instruction through said control unit means for processing when said number matches said count.

Claim 18

The image compression coprocessor of claim 12 wherein said number in said control field is one number in a sequence, and further comprising means for modifying said count in said semaphore register such that only one instruction is allowed to pass at a time.

Claim 19

The coprocessor of claim 1 wherein said control unit comprises: an internal bus interface coupled to said internal bus; an enabled instruction queue means, coupled to said internal bus interface, for holding instructions to be transferred to said special purpose processing means; a data token memory coupled to said internal bus interface; and update unit means, coupled to said enabled instruction queue means, for providing instructions to said enabled instruction queue means.

Claim 20

The coprocessor of claim 14 wherein said update unit means comprises: first means for determining if an instruction is in said enabled instruction queue means; second means for determining if data tokens associated with said instruction are in said data token memory; third means for determining if a one of said special purpose processing means required by said instruction is busy; means for providing said instruction to said enabled instruction queue means responsive to said first, second and third means for determining.

Claim 21

The coprocessor of claim 14 further comprising: fourth means for determining if said enabled instruction queue means is full.

Claim 22

The coprocessor of claim 14 wherein said update unit means further comprises: a semaphore register; and means, coupled to said semaphore register, for modifying a count in said semaphore register upon accessing a data token in said data token memory, testing a semaphore value in said semaphore register and preventing a new data token from being accessed if said semaphore value exceeds a maximum value, and modifying said count upon a data token leaving said control unit.

Claim 23

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a first register, wherein said program includes a CRTOKEN instruction which can be executed only once in said program, said CRTOKEN instruction being allowed to execute only if said first register has a first value, said CRTOKEN instruction causing said first register to assume a value other than said first value after execution, said coprocessor including reset means for changing said first register to said first value.

Claim 24

An image compression coprocessor integrated on a single semiconductor chip comprising: control unit means for operating said coprocessor in accordance with a stored program; an internal bus coupled to said control unit means; and a plurality of special purpose processing means, each coupled to said bus, for performing a subset of a group of steps in an image compression/decompression process; wherein each of said special purpose processing means have different, special purpose hardware; wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a first register, wherein said program includes a CRTOKEN instruction which can be executed only once in said program, said CRTOKEN instruction being allowed to execute only if said first register has a first value, said CRTOKEN instruction causing said first register to assume a value other than said first value after execution, said coprocessor including reset means for changing said first register to said first value.

Claim 25

The coprocessor of claim 1 further comprising: wherein said stored program is a data flow program, and said control unit means transfers instructions and data tokens to said plurality of special purpose processing means over said bus; a block allocation memory for storing blocks of data for said data tokens; a token address memory for storing pointers to said blocks of memory, each of said pointers corresponding to a data token; and control means, coupled to said token address memory, for copying a block of data from one data token to another data token by modifying said pointers.

Claim 26

The coprocessor of claim 19 further comprising a header memory for storing descriptors for said data tokens, said token address memory storing pointers to both a descriptor in said header memory and a block of data in said block allocation memory for each data token.

Claim 27

The coprocessor of claim 15 further comprising fourth means for determining if a functional unit in an external processor is busy.

Claim 28

The coprocessor of claim 21 wherein one of said special purpose processing means is an auxiliary interface unit for interfacing with external processing units, and wherein said fourth means comprises a status table in said auxiliary interface unit.

Claim 29

The coprocessor of claim 22 wherein said status table is a status register having a location for each functional unit, and further comprising: means for setting a first status bit in said status register upon the transmission of an instruction designating a first functional unit in said external processor; and means for clearing said first status bit upon the reception of a result packet from said first functional unit in said external processor.

Claim 30

The coprocessor of claim 17 wherein said internal bus interface includes a scalar processor unit means for copying an operand of a semaphore instruction to a result token after said modifying a count and testing a semaphore value in said update unit means.

Claim 31

The processor of claim 1 wherein a plurality of said processing means include: an instruction register, coupled to said bus, for holding one of said instructions; a buffer, coupled to said bus, for holding at least one of said data tokens; processing logic, coupled to said buffer and said instruction register, for performing said subset of steps; and a state machine, coupled to said instruction register and said buffer, for controlling the interfacing of said instruction register and buffer with said bus.

Claim 32

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a host interface port for coupling to a master host processor; run length processor means, connected between said bus and said host interface port, for converting data between said token format and a run length format for communicating with said master host processor.

Claim 33

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a host interface port for coupling to a master host processor; run length processor means, connected between said bus and said host interface port, for converting data between said token format and a run length format for communicating with said master host processor.

Claim 34

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a host interface port for coupling to a master host processor; and token interface means, coupled between said bus and said host interface port, for allowing said host processor to insert tokens directly and to view tokens on said bus.

Claim 35

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a host interface port for coupling to a master host processor; and token interface means, coupled between said bus and said host interface port, for allowing said host processor to insert tokens directly and to view tokens on said bus.

Claim 36

The processor of claim 1 further comprising an auxiliary unit, coupled between said global bus and an auxiliary interface, for coupling an external auxiliary processor to said global bus.

Claim 37

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a semaphore register coupled to said control unit; a plurality of semaphore instructions in said data flow program for modifying a count in said semaphore register upon the accessing of a token, comparing a value in said semaphore register with a value in a token and preventing a new data token from being accessed if said semaphore value exceeds a maximum value, and modifying said count upon a data token leaving said control unit.

Claim 38

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a semaphore register coupled to said control unit; a plurality of semaphore instructions in said data flow program for modifying a count in said semaphore register upon the accessing of a token, comparing a value in said semaphore register with a value in a token and preventing a new data token from being accessed if said semaphore value exceeds a maximum value, and modifying said count upon a data token leaving said control unit.

Claim 39

The processor of claim 29 wherein a first semaphore instruction tests and decrements said count before each instruction requiring a data token, and a second semaphore instruction increments said count after each data token leaves said control unit.

Claim 40

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a semaphore register coupled to said control unit means; a plurality of instructions in said data flow program having a control field with a number; and means for comparing said number in said control field in an instruction with a count in said semaphore register and passing said instruction through said control unit means for processing when said number matches said count.

Claim 41

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a semaphore register coupled to said control unit means; a plurality of instructions in said data flow program having a control field with a number; and means for comparing said number in said control field in an instruction with a count in said semaphore register and passing said instruction through said control unit means for processing when said number matches said count.

Claim 42

The processor of claim 31 wherein said number in said control field is one number in a sequence, and further comprising means for modifying said count in said semaphore register such that only one instruction is allowed to pass at a time.

Claim 43

A processor integrated on a single semiconductor chip comprising: a control unit; a memory coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; an internal bus interface coupled to said internal bus; an enabled instruction queue means, coupled to said internal bus interface, for holding instructions to be transferred to said processing circuits; a data token memory coupled to said internal bus interface; and an update unit, coupled to said enabled instruction queue, for providing instructions to said enabled instruction queue; a semaphore register; and means, coupled to said semaphore register, for modifying a count in said semaphore register upon accessing a data token in said data token memory, testing a semaphore value in said semaphore register and preventing a new data token from being accessed if said semaphore value exceeds a maximum value, and modifying said count upon a data token leaving said control unit.

Claim 44

A processor integrated on a single semiconductor chip comprising: a control unit; a memory coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps in a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; an internal bus interface coupled to said internal bus; an enabled instruction queue means, coupled to said internal bus interface, for holding instructions to be transferred to said processing circuits; a data token memory coupled to said internal bus interface; and an update unit, coupled to said enabled instruction queue, for providing instructions to said enabled instruction queue; a semaphore register; and means, coupled to said semaphore register, for modifying a count in said semaphore register upon accessing a data token in said data token memory, testing a semaphore value in said semaphore register and preventing a new data token from being accessed if said semaphore value exceeds a maximum value, and modifying said count upon a data token leaving said control unit.

Claim 45

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps an a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a first register, wherein said program includes a CRTOKEN instruction which can be executed only once in said program, said CRTOKEN instruction being allowed to execute only if said first register has a first value, said CRTOKEN instruction causing said first register to assume a value other than said first value after execution, said processor including reset means for changing said first register to said first value.

Claim 46

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal bus coupled to said control unit; and a plurality of processing circuits, each coupled to said bus, for performing a subset of a group of steps an a process performed by said processor; wherein said stored program is a data flow program, and said control unit transfers instructions and data tokens to said plurality of processing circuits over said bus; wherein said data tokens are data vectors; a first register, wherein said program includes a CRTOKEN instruction which can be executed only once in said program, said CRTOKEN instruction being allowed to execute only if said first register has a first value, said CRTOKEN instruction causing said first register to assume a value other than said first value after execution, said processor including reset means for changing said first register to said first value.

Claim 47

The processor of claim 33 wherein said internal bus interface includes a scalar processor unit for copying an operand of a semaphore instruction to a result token after said modifying a count and testing a semaphore value in said update unit.

Claim 48

The processor of claim 29 further comprising: a data token memory coupled to said global bus and accessible by each of said processing circuits.

Claim 49

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal global bus coupled to said control unit; a plurality of processing circuits, each coupled to said global bus, for performing a subset of a group of steps in a process performed by said processor; and a data token memory coupled to said global bus and accessible by each of said processing circuits; wherein said stored program is a data flow program, and said control unit asynchronously transfers instructions and data tokens to said plurality of processing circuits over said global bus; wherein said data tokens are data vectors.

Claim 50

A processor integrated on a single semiconductor chip comprising: a control unit; a memory, coupled to said control unit, having a stored program; an internal global bus coupled to said control unit; a plurality of processing circuits, each coupled to said global bus, for performing a subset of a group of steps in a process performed by said processor; and a data token memory coupled to said global bus and accessible by each of said processing circuits; wherein said stored program is a data flow program, and said control unit asynchronously transfers instructions and data tokens to said plurality of processing circuits over said global bus; wherein said data tokens are data vectors.