Dual-scale topology optoelectronic matrix algebraic processing system

• US 5,321,639 A
• Filed: 03/02/1992
• Issued: 06/14/1994
• Est. Priority Date: 03/02/1992
• Status: Expired due to Fees

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Ford, Joseph E., Marsden, Gary C., Krishnamoorthy, Ashok V., Esener, Sadik C.
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): TRAMMELL, JAMES P
• Application Number:
  US07/846,277
• Time in Patent Office: 834 Days
• Field of Search: 364/807, 364/822, 364/837, 364/845, 364/601, 364/602, 364/713, 364/606, 395/25, 359/107, 359/108
• US Class Current: 708/7
• CPC Class Codes: G06E 1/02 operating upon the order or content of the data handled
• CPC Class Codes: G06N 3/0675 using electro-optical, acousto-optical or opto-electronic means

Abstract

A parallel architecture matrix algebraic processing system exhibits patterns of arrayed (i) light transmitters and (ii) light receivers that are identical, but at differing scales. Planar arrays of one or more optoelectronic processors--principally semiconductor chips or chip arrays--having both computational and light input/output capabilities optically communicate from one plane to the next through free-space space-invariant optical data distributions--principally lenses and computer-generated holograms--having both replication and distribution capabilities. Each optoelectronic processor, or OP, consists of a number of arrayed optoelectronic processing elements, or OPEs. The OPEs, in turn, typically consist of a number of optoelectronic sub-processing units are preferably electrically interconnected in a tree-based structure, preferably an H-tree. Leaf units include typically one light detector plus local memory, logic circuitry, and electrical input/output. Fanning units typically include local memory, logic circuitry, and electrical input/output. A root unit typically includes electrically-connected local memory, logic circuitry, electrical input/output, and a light transmitter. Vector results of algebraic computations and combinations are flexibly performable in the units of each OPE, and variously optically distributable to other OPEs in successive OPs. The versatile algebraic vector manipulations and vector distributions support primitive functions such as intrinsic and extrinsic vector outer products; operations such as vector-matrix multiplication; and complex systems such as neural networks, fuzzy logic and relational databases. A system of ≧10³ fully optically communicating OPEs achieves capacities of 10⁶ -10⁸ interconnects, and processing speeds of 10¹² interconnects/second.

Claims

1. An optoelectronic processing system comprising:

a plurality of optoelectronic processors each having

multiple pluralities of light detector means each of which pluralities is two-dimensionally arrayed in a first pattern, each plurality of light detector means for receiving optically-encoded input data

electrical circuit means for manipulating the received input data to produce result data, and

a plurality of light transmitters two-dimensionally arrayed in a second pattern for transmitting the result data as optically-encoded light; and

a plurality of processor-to-processor optical distribution means functionally interleaved between the plurality of optoelectronic processors, each simultaneously operative

for optically distributing in a third dimension the result data from the two-dimensionally second-arrayed plurality of light transmitters within a one processor to multiple two-dimensionally first-arrayed pluralities of light detectors within a next successive one of the plurality of optoelectronic processors;

wherein manipulation of optically-encoded input data transpires electrically within a processor while distribution of the result data by the plurality of processor-to-processor optical distribution means transpires optically.

2. The optoelectronic processing system according to claim 1 wherein each of the plurality of processor-to-processor optical distribution means is optically distributing the result data in parallel from the plurality of light transmitters within the one processor to each of the multiple pluralities of light detectors within the next successive one of the plurality of processors wherein optical distribution of the result vector is in parallel.

3. The optoelectronic matrix algebraic processing system according to claim 1

wherein the first pattern of each of the multiple arrayed pluralities of light detectors in one processor is geometrically similar to the second pattern of the plurality of light transmitters within a preceding one of the plurality of processors;

wherein each of the plurality of processors-to-processor optical distribution means is optically distributing the result data in parallel from each of the pluralities of light transmitters, arrayed in the second pattern, of one processor to each of the multiple pluralities of light detectors, arrayed in a geometrically similar first pattern, of a next successive one of the plurality of processors.

4. The optoelectronic processing system according to claim 1 wherein each of the plurality of processor-to-processor optical distribution means comprises:

space-invariant optics.

5. The optoelectronic processing system according to claim 4 wherein one of the plurality of space-invariant optics comprises:

demagnification means for reducing the size of the optically-encoded result data; and

replication means for replicating the reduced-sized optically-encoded result data and distributing the replicated result data in parallel to the multiple pluralities of light detectors of a next successive one of the plurality of processors.

6. The optoelectronic matrix algebraic processing system according to claim 5 wherein the demagnification means comprises:

a lens;

and wherein the replication means comprises;

a hologram.

7. The optoelectronic processing system according to claim 4 wherein one of the plurality of space-invariant optics comprises:

replication means for replicating the optically-encoded result data and distributing the replicated result data in parallel to the multiple pluralities of light detectors of a next successive one of the plurality of processors.

8. The optoelectronic matrix algebraic processing system according to claim 7 wherein the replication means comprises:

lenslets.

9. The optoelectronic matrix algebraic processing system according to claim 7 wherein the replication means comprises:

lenslets.

73. The matrix-algebraic optoelectronic processing system according to claim ,4 wherein at least one of the first-to-second-matrix optical distribution means and the second-to-first-matrix optical distribution means comprises:

a space-invariant optical system.

10. An optoelectronic matrix algebraic processing system comprising:

an optoelectronic processor means

for receiving, in each of multiple two-dimensionally arrayed pluralities of light detectors an optically-encoded input data vector,

for storing in a local memory a data matrix,

for electrically algebraically manipulating in electrical circuitry the received input data vector in consideration of the stored data matrix to produce a result vector, and

for transmitting, in a plurality of two-dimensionally arrayed light transmitters, the result vector as optically-encoded light; and

an optical distribution means

for optically distributing the result data vector from the plurality of light transmitters to each of the multiple arrayed pluralities of light detectors of the processor means as the input data vector;

wherein a recycling of the output vector by the optical distribution means back to the processor means as the input vector permits recursive electronic algebraic manipulation of the input vector by the processor means. processor means as the input data vector;

11. The optoelectronic matrix algebraic processing system according to claim 10 wherein the optical distribution means comprises:

space-invariant optics.

12. The optoelectronic matrix algebraic processing system according to claim 11 wherein the space-invariant optics comprises:

demagnification means for reducing the size of the optically-encoded result data vector; and

replication means for replicating the reduced-sized optically-encoded result data vector and distributing the replicated result vector in parallel back to the processor means as the inout vector.

13. The optoelectronic matrix algebraic processing system according to claim 12 wherein the demagnification means comprises:

a lens;

and wherein the replication means comprises;

a hologram.

14. The optoelectronic matrix algebraic processing system according to claim 11 wherein the space-invariant optics comprises:

replication means for replicating the reduced-sized optically-encoded result data vector and distributing the replicated result vector in parallel back to the processor means as the input vector.

15. An optoelectronic method of matrix algebraic processing comprising:

storing one or more data matrices in local memories of each of a plurality of optoelectronic processors,

receiving one or more optically-encoded input data vectors in each of multiple arrayed pluralities of light detectors that are within each of the optoelectronic processors,

electrically algebraically manipulating in electrical circuitry within each optoelectronic processor the one or more input data vectors received at that optoelectronic processor in consideration of the one or more data matrices stored at that optoelectronic processor so as to produce at least one result vector,

transmitting the at least one result vector off each optoelectronic processor as optically-encoded light; and

optically distributing, in each of a plurality of processor-to-processor optical distribution means that are functionally interleaved between the plurality of optoelectronic processors, the result data vector of a one processor to the multiple arrayed pluralities of light detectors of one or more of the processors;

wherein algebraic manipulations producing the result vector are performed electrically while communication of the result vector is performed optically;

wherein there is a distribution of each result data vector arising at one processor to one or more processors.

16. The optoelectronic method of matrix algebraic processing according to claim 15 wherein the optically distributing is space-invariant.

17. The optoelectronic method of matrix algebraic processing according to claim 15

wherein the optical distributing of one optical distribution means communicates the result data vector in a feedback loop back to the same optoelectronic processor from which it arose;

therein to permit recursive performance said same optoelectronic processor of the electrical algebraic manipulating.

18. The optoelectronic method of matrix algebraic processing according to claim 15

wherein the optical distributing of one optical distribution means communicates the result data vector from one processor to a next processor as a first receipt by the next processor of the result vector;

therein to successive performance in successive optoelectronic processors of the electrical algebraic manipulating.

19. The optoelectronic method of matrix algebraic processing according to claim 15

wherein the optically distributing of one of the optical distribution means communicates the result data vector from one processor to a plurality of processors.

20. The optoelectronic method of matrix algebraic processing according to claim 15

wherein the optically distributing of a plurality of the optical distribution means communicates the result data vector from a plurality of processors to one processor.

21. The optoelectronic method of matrix algebraic processing according to claim 15 further comprising:

optically receiving at an optoelectronic processor the one or more data matrices via multiple arrayed pluralities of light detectors that are within the optoelectronic processors; and

storing the optically-received data matrices in the local memories of the receiving optoelectronic processor.

22. The optoelectronic method of matrix algebraic processing according to claim 15 further comprising:

electrically receiving at an optoelectronic processor the one or more data matrices; and

storing the electrically-received data matrices in the local memories of the receiving optoelectronic processor.

23. A optically-mapping light-communicating optoelectronic vector-matrix algebraic processing system for algebraically manipulating a received source vector stepwise in accordance with each of a plurality of stored matrices to produce a result vector, the vector-matrix algebraic processing system comprising:

a plurality of optoelectronic processors each having a plurality of optoelectronic processing elements, each optoelectronic processing element having

a plurality of light detector means, collectively for receiving one or more optically-encoded input data vectors,

local memory means for storing a portion of one or more data matrices,

electrical computational means for electrically performing an algebraic operation on the input vector data received by the plurality of light detector means in consideration of the portion of the matrix data stored within the local memory means to produce a portion of a result data vector, and

a light transmitting means for transmitting the result data vector portion as a portion of an optically-encoded output data vector,

wherein the local memory means of all the plurality of optoelectronic processing elements of an optoelectronic processor collectively store at least one complete data matrix,

wherein the collective electrical computational means of all the plurality of optoelectronic processing elements of an optoelectronic processor collectively produce at least one complete result data vector, and

wherein the collective light transmitting means of all the plurality of optoelectronic processing elements collectively transmit at least one complete optically-encoded output data vector;

a source data vector distribution means for distributing at least one optically-encoded source data vector to the plurality of light detectors of at least one of the optoelectronic processing elements of at least one of the plurality of processors for use therein as an optically-encoded input data vector; and

a plurality of processor-to-processor optical distribution means, each

for distributing an optically-encoded output data vector of an optoelectronic processor to a plurality of light detectors within a plurality of optoelectronic processing elements of a next one of the plurality of optoelectronic processors for use therein as an input data vector,

wherein a result vector of a one optoelectronic processor is optically received as an input vector by one or more next optoelectronic processors, and so on, one processor to the next;

wherein each optoelectronic processor stepwise in turn electrically performs an algebraic operation on a received vector so that, ultimately, a final optically-encoded result vector is produced by successive algebraic manipulations of an original, source, vector in accordance with those matrices that are held within the local memories of the plurality of optoelectronic processing elements of each of the plurality of optoelectronic processors.

24. The optically-mapping optoelectronic vector-matrix algebraic processing system according to claim 23

wherein the plurality of light detector means of each of the optoelectronic processing elements of each of the optoelectronic processors are arrayed in a pattern associated with each such plurality of optoelectronic processing elements;

and

wherein the optoelectronic processing elements of each of the optoelectronic processors are arrayed in a pattern associated with each such plurality of optoelectronic processors;

wherein because the optoelectronic elements are arrayed in patterns, and because the light transmitting means are within the optoelectronic processing elements, then the light transmitting means are also arrayed in the patterns of the optoelectronic processing elements within which they are contained.

25. The optically-mapping optoelectronic vector-matrix algebraic processing system according to claim 24

wherein the pattern of the light transmitters of the plurality of optoelectronic processing elements of an optoelectronic processor is replicated in a next successive optoelectronic processor in its pattern of its light detectors that are within each of its optoelectronic processing elements, but at a smaller scale.

26. The optically-mapping optoelectronic vector-matrix algebraic processing system according to claim 24

wherein the pattern of the light transmitters of the plurality of optoelectronic processing elements of an optoelectronic processor is replicated in a next successive optoelectronic processor in its pattern of its optoelectronic processing elements.

27. The optoelectronic vector-matrix algebraic processing system according to claim 23

wherein the processor-to-processor optical distribution means is space-invariant.

28. The optoelectronic vector-matrix algebraic processing system according to claim 27 wherein the space-invariant processor-to-processor optical distribution means comprises:

means for distributing each bit of each output data vector so that it is received by a corresponding one of the plurality of light detectors within each of the plurality of optoelectronic processing elements of the next processor;

wherein this means is arbitrarily called, in consideration of the matrix that is stored within the local memories of the plurality of optoelectronic processing elements, a vertical optical distribution.

29. The optoelectronic- vector-matrix algebraic processing system according to claim 27 wherein the space-invariant processor-to-processor optical distribution means comprises:

means for distributing each bit of each output data vector so that it is received by all of the plurality of light detectors that are within each corresponding one of the plurality of optoelectronic processing elements of the next processor;

wherein this means is arbitrarily called, in consideration of the matrix that is stored within the local memories of the plurality of optoelectronic processing elements, a horizontal optical distribution.

30. The optoelectronic vector-matrix algebraic processing system according to claim 23 wherein the source vector distribution means comprises:

optical means for distributing the source vector to the plurality of light detectors of the plurality of optoelectronic processing elements of a first one of the plurality of optoelectronic processors for use by this first optoelectronic processor as the input data vector.

31. The optoelectronic vector-matrix algebraic processing system according to claim 23 wherein the source vector distribution means comprises:

electrical means for distributing the source vector to the electrical computational means of the plurality of optoelectronic processing elements of a first one of the plurality of optoelectronic processors for use by this first optoelectronic processor as the input data vector.

32. A single-processor optoelectronic matrix-algebraic processing system for receiving and for algebraically manipulating an optically-encoded external data vector, the system comprising:

one optoelectronic processor comprising;

a plurality of optoelectronic processing elements first-arrayed in a plane in a first pattern, each element comprising;

a plurality of light detector means, second-arrayed in the plane in a second pattern similar to the first pattern, for receiving the optically-encoded external data vector from spatially off the plane,

memory means for storing a portion of a data matrix,

electrical circuit means for electrically algebraically manipulating the external data vector received by the light detector means in consideration of the portion of the data matrix stored by the memory means to produce a portion of a result vector, and

a light transmitter means for transmitting the portion of the result vector spatially off the plane as optically-encoded light,

wherein the collective electrical circuit means of the collective optoelectronic processing elements produce, and the collective light transmitter means of the collective optoelectronic processing elements transmit, the entire result vector; and

a free-space light transfer means for communicating the result vector from the collective light transmitting means of the collective optoelectronic processing elements of the optoelectronic processor in a closed loop back to the light detector means of the plurality of optoelectronic processing elements of the optoelectronic processor;

therein to feed back in a loop back via optical communication a result vector from the optoelectronic processor to itself.

33. The system according to claim 32 further wherein the optoelectronic processor further comprises:

means for optically receiving the data matrix into the optoelectronic processor; and

means for storing the optically-received data matrix in the local memories.

34. The system according to claim 32 further comprising:

means for electrically receiving the data matrix into the optoelectronic processor; and

means for storing the electrically-received data matrix in the local memories.

35. The system according to claim 32 wherein the free-space light transfer means is space-invariant.

36. The system according to claim 35 wherein the free-space space-invariant light transfer means comprises:

optical distribution means for distributing the result vector so that it is received by a corresponding one of the plurality of light detectors within each of the plurality of optoelectronic processing elements of the optoelectronic processor;

wherein this optical distribution is arbitrarily called, in consideration of the matrix that is stored within the plurality of local memories of the optoelectronic processing elements of the optoelectronic processor, a vertical optical distribution.

37. The system according to claim 36 wherein the vertical optical distribution means comprises:

optical demagnification means for demagnifying the first pattern of optically-encoded light that is transmitted from the collective transmitting means as are within the collective plurality of opto-electronic processing elements of the optoelectronic processor to the size of the second pattern, and

optical replication means for replicating the demagnified optically-encoded light so that it is received into the detector means within each of the plurality of optoelectronic processor elements of the optoelectronic processor;

therein to achieve optical connection from the collective transmitting means as are within the collective plurality of opto-electronic processing elements of the optoelectronic processor to the detector means within each of the plurality of optoelectronic processor elements of the optoelectronic processor.

38. The system according to claim 35 wherein the free-space space-invariant light transfer means comprises:

optical distribution means for distributing the result vector so that portion of the result vector transmitted by the light transmitting means of an optoelectronic processing element of the optoelectronic processor is received by all of the plurality of light detectors that are within the same optoelectronic processing element of the optoelectronic processor;

wherein this optical distribution is arbitrarily called, in consideration of the matrix that is stored within the plurality of local memories of the optoelectronic processing elements of the optoelectronic processor, a horizontal optical distribution.

39. The system according to claim 38 wherein the horizontal optical distribution means comprises:

optical replication means for replicating the optically-encoded light as arises from each transmitter within each of plurality of optoelectronic processing elements of the optoelectronic processor and for distributing such light to the detector means of that same identical one of the plurality of optoelectronic processing elements of the optoelectronic processor;

therein to achieve optical connection from the transmitters within each optoelectronic processing element of the optoelectronic processor to the detector means of the same optoelectronic processing element of the optoelectronic processor.

44. The system according to claim 37 or claim 43 wherein the optical demagnification means comprises:

a lens.

45. The system according to claim 37 or claim 43 wherein the optical replication means comprises:

a hologram.

47. The system according to claim 39 or claim 46 wherein the optical replication means comprises:

lenslets

48. The system according to claim 32 or claim 40 wherein at least one light transmitting means of one optoelectronic processor comprises:

a first chip;

and wherein at least one electrical circuit means of one optoelectronic processor comprises;

a second chip electrically connected to the first chip.

49. The system according to claim 32 or claim 40 wherein at least one light transmitting means of one optoelectronic processor comprises:

a circuit upon a first chip;

and wherein at least one electrical circuit means of one optoelectronic processor comprises;

a circuit upon a second chip electrically connected to the first chip by being bonded face-to-face with the first chip in a flip-chip arrangement.

50. The system according to claim 32 or claim 40 wherein at least one light detector means comprises:

a circuit upon a first chip fabricated in silicon technology; and

wherein at least one light transmitting means comprises;

a circuit upon a second chip fabricated in PLZT technology and electrically connected to the first chip.

51. The system according to claim 32 or claim 40 wherein at least one light detector means comprises:

a circuit upon a first chip fabricated in silicon technology;

and wherein at least one light transmitting means comprises;

a circuit upon a second chip fabricated in PLZT technology and electrically connected to the first chip by being bonded face-to-face with the first chip in a flip-chip arrangement.

52. The system according to claim 32 or 40 wherein at least one light transmitting means comprises:

a circuit upon a chip;

and wherein at least one the light detecting means comprises;

a circuit upon the chip;

and wherein at least one electrical circuit means comprises;

a circuit upon the chip;

wherein at least one light transmitting means, at least one light detecting means, and at least one electrical circuit means are fabricated as a monolithic structure upon a single chip.

53. The system according to claim 32 or claim 40 wherein at least one plurality of optoelectronic processing elements in at least one processor comprises:

N optoelectronic processing elements arrayed in a square two-dimensional first array of dimension √

N elements by √

N elements.

54. The system according to claim 32 or claim 40 wherein at least one of the plurality of optoelectronic processing elements in at least one processor comprises:

an electrically-connected tree-based interconnection of the plurality of light detector means, the memory means, the electrical circuit means, and the light transmitting means.

55. The system according to claim 32 or claim 40 wherein at least one of the plurality of optoelectronic processing elements in at least one processor comprises:

an electrically-connected binary tree-based interconnection of the plurality of light detector means, the memory means, the electrical circuit means, and the light transmitting means.

56. The system according to claim 32 or claim 40 wherein at least one of the plurality of optoelectronic processing elements in at least one processor comprises:

an electrically-connected binary H-tree-based interconnection of the plurality of light detector means, the memory means, the electrical circuit means, and the light transmitting means.

57. The system according to claim 32 or claim 40 wherein at least one of the plurality of optoelectronic processing elements in at least one processor comprises:

fan-in processing unit means for electrically summing the outputs of the plurality of light detector means and presenting the sum to the light transmitting means.

40. A plural-processor optoelectronic matrix-algebraic processing system for receiving and for algebraically manipulating an optically-encoded external data vector, the system comprising:

a first optoelectronic processor in a first-processor plane comprising;

a plurality of optoelectronic processing elements arrayed in the first-processor plane in a first pattern, each element comprising;

a plurality of light detector means, arrayed in the first-processor plane in a second pattern, for receiving an optically-encoded data vector from spatially off the first-processor plane,

memory means for storing a portion of a data matrix,

electrical circuit means for electrically algebraically manipulating the data vector received by the light detector means in consideration of the portion of the data matrix stored by the memory means to produce a portion of a first result vector, and

a light transmitting means for transmitting the portion of the first result vector spatially off the first-processor plane as optically-encoded light,

wherein the collective electrical circuit means of the collective optoelectronic processing elements of the first optoelectronic processor produce, and the collective light transmitting means of the collective optoelectronic processing elements of the first optoelectronic processor transmit, the entire first result vector;

a second optoelectronic processor in a second-processor plane comprising;

a plurality of optoelectronic processing elements, arrayed in the second-processor plane in a third pattern similar to one of the first and the second patterns, each element comprising;

a plurality of light detector means, arrayed in the second-processor plane in a fourth pattern, similar to one of the first and second patterns, for receiving an optically-encoded data vector from spatially off the second-processor plane,

memory means for storing a portion of a data matrix,

electrical circuit means for electrically algebraically manipulating the data vector received by the light detector means in consideration of the portion of the data matrix stored by the memory means to produce a portion of a second result vector, and

a light transmitting means for transmitting the portion of the second result vector spatially off the second-processor plane as optically-encoded light,

wherein the collective electrical circuit means as are within the collective processing elements produce, and the collective light transmitting means as are within the collective optoelectronic processing elements transmit, the entire second result vector;

a first free-space light transfer means for communicating the first result vector from the light transmitting means of the collective optoelectronic processing elements of the first processor to the light detector means within the plurality of optoelectronic processing elements of one of (i) the first and the second processors, and (ii) the second processor;

a second free-space light transfer means for communicating the second result vector from the light transmitting means of the collective optoelectronic processing elements of the second processor to the light detector means within the plurality of optoelectronic processing elements of one of (i) the second and the first processors, and (ii) the first processor.

41. The optoelectronic processing system according to claim 40 further comprising:

and external data matrix distribution means for optically communicating an optically-encoded external data matrix to the plurality of optoelectronic processing elements of at least one of the first and the second optoelectronic processors.

42. The system according to claim 40 wherein the first free-space light transfer means is space-invariant.

43. The system according to claim 42 wherein the space-invariant first free-space light transfer means comprises:

optical demagnification means for demagnifying the first pattern of optically-encoded light that is transmitted from the collective light transmitting means as are within the collective plurality of opto-electronic processing elements of the first processor to the size of the third pattern; and

optical replication means for replicating the demagnified optically-encoded light in the first pattern, and for distributing the demagnified and replicated first-pattern light to the detector means within the plurality of optoelectronic processor elements within the second processor, which detector means within the second processor are arrayed in the third pattern;

wherein the third pattern is similar to the first pattern, but of a reduced size;

therein to achieve optical connection from the light transmitting means of the first processor to each of the plurality of detector means as are within all the optoelectronic processing elements of the second processor.

46. The system according to claim 42 wherein the space-invariant free-space light transfer means comprises:

optical replication means for replicating the second pattern of optically-encoded light that is transmitted from the collective light transmitting means as are within the collective plurality of opto-electronic processing elements of the first processor, and for distributing the replicated second-pattern light to the plurality of light detector means within one of the plurality of optoelectronic processor elements within the second processor, which light detector means within one optoelectronic processing element of the second processor are arrayed in the fourth pattern;

wherein the fourth pattern is similar to the second pattern;

therein to achieve optical connection from the light transmitting means of the first processor to each of the plurality of light detector means as are within one of the optoelectronic processing elements of the second processor.

58. A matrix algebraic optoelectronic processing system operating on external data vectors, the system comprising:

an optoelectronic processor comprising a plurality of M processing elements arranged in a substantially planar having M rows, each processing element comprising and electrically interconnecting in a tree structure

N leaf units corresponding to the matrix elements within a given row, each leaf unit comprising an electrically-connected

plurality of light detectors,

local memory,

logic circuitry, and

electronic input/output,

the leaf units electrically connected to a root node unit of the tree, the root node unit comprising an electrically-connected

local memory,

logic circuitry,

electronic input/output, and

optical transmitter;

wherein the processing element supports an electrical distribution of data via its tree structure between the plurality of leaf units and the root unit; and

a first, vertical, optical distribution for introducing to the M processing elements a first data vector originating outside the matrix algebraic processor, the introducing being from a location off the plane of the M planar-arrayed processing elements, the introducing transmitting the first external vector data element x_j to the jth leaf unit of each processing element by action of supplying information-encoded light to a first one of the plurality of light detectors of this jth leaf unit;

wherein local computation is performable at each processing element'"'"'s leaf units on (i) data received via the first optical distribution and the electrical distribution, and (ii) data from the leaf unit'"'"'s local memory;

wherein during a fan-in of the external data vector a computation, distributed among leaf units and the root unit, is performed within each processing element during passage of the data from the leaf units to the root unit;

wherein computation upon the external data vector produces a result at the root unit;

wherein the computational result is transmittable from the root unit'"'"'s optical transmitter as an optical output;

wherein the optoelectronic processing system is capable of performing matrix vector arithmetic and symbolic manipulations.

59. The matrix algebraic optoelectronic processing system according to claim 58 wherein the optoelectronic processors tree-structured processing elements further comprise between the leaf units and the root node unit:

intermediate fanning-unit nodes of the tree, each fanning-unit node comprising electrically-connected

local memory,

logic circuitry, and

electronic input/output,

wherein local computation is also performable at each processing element'"'"'s fanning-unit nodes on (i) data received via the electrical distribution from the leaf nodes, and (ii) data from the fanning-unit'"'"'s local memory;

wherein during a fan-in of the external data vector the computation is distributed among the fanning-unit nodes as well as the leaf units and the root unit element during the passage of the data from the leaf units to the root unit;

wherein the optoelectronic processing system is capable of performing generalized matrix vector arithmetic and symbolic manipulations with tree-based processing.

60. The matrix algebraic optoelectronic processing system according to claim 59

wherein the electrical distribution of data within each tree-structured processing element is bi-directional, being from and to the plurality of leaf units to and from the root unit via the fanning units;

and wherein the system further comprises;

a second, horizontal, optical distribution for introducing to the M processing elements a second data vector originating outside the matrix algebraic processor, the introducing being from a location off the plane of the M planar-arrayed processing elements, the introducing transmitting the second external vector data element y_i to each leaf unit of the ith processing element by action of supplying information-encoded light to a second one of the plurality of light detectors of each leaf unit;

wherein the optoelectronic processing system is capable of performing intrinsic and extrinsic vector outer product arithmetic and symbolic manipulations.

61. Two matrix algebraic optoelectronic processors according to claim 58 physically spaced parallel and operating in tandem as a system to process the matrix algebraic problem of M×

N size, the tandem matrix algebraic processing system comprising;

a first space-invariant optical system for performing the first optical distribution from a first to a second of the spaced-parallel optoelectronic processors by action of transmitting the first external vector data element x_j to the jth leaf unit of each processing element by action of supplying information-encoded light to a first one of the two light detectors of the leaf unit;

a second space-invariant optical system for performing the second optical distribution from the first to the second of the spaced-parallel optoelectronic processors by action of transmitting the second external vector data element y_i to each leaf unit of the ith processing element by action of supplying information-encoded light to a second one of the two light detectors of the leaf unit;

a third space-invariant optical system for performing the first optical distribution from the second to the first of the spaced-parallel optoelectronic processors by action of transmitting the first external vector data element x_j to the jth leaf unit of each processing element by action of supplying information-encoded light to a first one of the two light detectors of the leaf unit; and

a fourth space-invariant optical system for performing the second optical distribution from the second to the first of the spaced-parallel optoelectronic processors by action of transmitting the second external vector data element y_i to each leaf unit of the ith processing element by action of supplying information-encoded light to a second one of the two light detectors of the leaf unit.

62. The matrix algebraic optoelectronic processing system according to claim 58 further comprising:

a space-invariant optical system for introducing to the M processing elements a plurality M of memory data vectors originating outside the matrix algebraic processing system, one data vector per processing element;

wherein the introducing is to the local memories of the plurality N leaf units, the fanning units, and the root unit of the M processing elements.

63. The matrix algebraic optoelectronic processing system according to claim 58 further comprising:

a second electrical distribution for introducing to the M processing elements a plurality M of memory data vectors originating outside the processing system, one data vector per processing element;

wherein the introducing is to the local memories of the plurality N leaf units, the fanning units, and the root unit of the M processing elements.

64. An optoelectronic matrix algebraic processing system operating on external data vectors, the system comprising:

a plurality L of optoelectronic processor OP_k where k equals 1 through L, each optoelectronic processor OP_k comprising;

a plurality of M_k arrayed optoelectronic processing elements OPE_m, m equals 1 through M_k, each optoelectronic processing element OPE_k comprising and electrically connecting in a tree structure

N_k leaf units LU_n, n equals 1 to N_k, each comprising an electrically-connected

plurality of light detectors,

local memory,

logic circuitry, and

electrical input/output,

electrically connected to

a root node unit of the tree, the root node unit comprising an electrically-connected

local memory,

logic circuitry,

electrical input/output, and

optical transmitter;

wherein the optoelectronic processing element OPE_k supports an electrical distribution of data via its tree structure between the plurality of leaf units and the root unit; and

a plurality of vertical optical distributions VOD_v, v equals 1 though V, each for the purpose of distributing a data vector transmitted from an associated optoelectronic processor OP_t(v) to an associated optoelectronic processor OP_r(v), where the data vector portion transmitted by the optoelectronic processing element OPE_i of the transmitting optoelectronic processor OP_t(v) is received by one of the plurality of light detectors within the leaf unit TU_i within each optoelectronic processing element OPE_m, m equals 1 though M_r(v), within the receiving optoelectronic processor OP_r(v), with the restriction that M_t(v) equals N_r(v) ; and

a plurality of horizontal optical distributions HOD_h, h equals 1 though H, each for the purpose of distributing a data vector transmitted from an associated optoelectronic processor OP_t(h) to an associated optoelectronic processor OP_r(h), where the data vector portion transmitted by the optoelectronic processing element OPE_i of the transmitting optoelectronic processor OP_t(h) is received by one of the plurality of light detectors within each leaf unit LU_i within each optoelectronic processing element OPE_j, within the receiving optoelectronic processor OP_r(h), with the restriction that M_t(h) equals M_r(h) ;

wherein local computation is performable at each processing element'"'"'s leaf unit on (i) data received via any of the optical distributions and the electrical distribution, and (ii) data from the leaf unit'"'"'s local memory;

wherein during a fan-in of the external data vector a computation, distributed among leaf units and the root unit, is performed within each processing element during passage of the data from the leaf units to the root unit;

wherein computation on data vectors received through optical distributions produces a result at the root unit;

wherein the computational result is transmittable from the root unit'"'"'s optical transmitter as an optical output;

wherein the optoelectronic processing system is capable of performing matrix vector arithmetic and symbolic, algebraic, manipulations.

65. A tandem-architecture free-space-optically-communicating optoelectronic matrix-algebraic processing system comprising:

two planar optoelectronic processors, each for (i) accepting optically-encoded information received from off-plane, (ii) electrically processing the received information, and (iii) optically transmitting the electrically-processed information off-plane;

a first free-space optical system for coupling the optical off-plane transmission of a first processor to a second processor; and

a second free-space optical system for coupling the optical off-plane transmission of the second processor to the first processor.

66. The system according to claim 65 wherein each of the planar optoelectronic processors comprises:

a grid array of M×

N leaf units each having an light detector;

M root units each having an optical transmitter; and

N electrical fanning means connected in a tree-based structure between M leaf elements and a corresponding root unit, for processing and for electrically communicating optically-encoded information received at the leaf units to a corresponding one of the N root units;

wherein the processed information received at the root unit is transmitted off-plane by the optical transmitter.

67. The system according to claim 65

wherein the grid array of M×

N leaf units of the first processor encodes an M×

N data matrix;

wherein the grid array of M×

N leaf elements of the second processors encodes the transpose of the data matrix;

wherein the first space-invariant optical system couples the M transmitters of the M root units of the first processor to a selected row of the N×

M matrix of leaf elements of the second processor; and

wherein the second space-invariant optical system couples the N transmitters of the N root units of the second processor to a selected column of the M×

N matrix of leaf elements of the first processor.

68. The system according to claim 65

wherein the two planar optoelectronic processors are on distinct planes.

69. The system according to claim 65

wherein the two planar optoelectronic processors are coplanar.

70. An optically-mapping optoelectronic vector-matrix algebraic processing system for algebraically manipulating an N₁ -bit input vector stepwise in accordance with L stored matrices Y_k, each matrix Y_k being of M_k rows by N_k columns where k equals 1 through L, in order to produce a M_k -bit output vector, the vector-matrix algebraic processing system comprising:

one or more, L, optoelectronic processors, each optoelectronic processor OP_k, k equals 1 through L, comprising;

a plurality of M_k arrayed optoelectronic processing elements 1, 2, . . . M, each optoelectronic processing element OPE_m, m equals 1 through M, comprising;

a plurality of N_k light detectors, each light detector LD_n, n equals 1 through N, for detecting a corresponding data bit X_n, n equals 1 through N_k, of an optically-encoded N_k -bit input vector X,

a plurality of at least N_k local memories, each local memory LM_n, n equals 1 though N_k, for storing a data bit Y_mn of a row m of a matrix Y_k, m equaling the number of the OPE_m in processor OP_k, in which processor OP_k the local memory LM_mn is located while n equals 1 through N_k,

an electrical computational means for electrically performing an algebraic operation on the N_k -bit input vector X detected by the N_k light detectors of processor OP_k in consideration of the row m of the stored matrix Y_k held within the at least N_k local memories to produce a result bit Z_i of a result vector Z, and

a light transmitter for transmitting the result bit Z_i of the result vector Z as an optically-encoded output signal;

wherein the collective N_k local memories of the collective M_k optoelectronic processing elements of each optoelectronic processor OP_k thus hold a matrix Y_k of size M_k rows ×

N_k columns,

wherein the collective M_k light transmitters of the collective M_k optoelectronic processing elements of each optoelectronic processor OP_k thus transmit a result vector Z_j of M_k bits;

wherein the algebraic operation performed by the collective L optoelectronic processors is thus the stepwise manipulation of an N₁ -bit input vector X by a successive matrices Y_k each of M_k rows by N_k columns, k equals 1 through L, with the additional restriction that N_k is equal to M_k-1 ; and

a processor-to-processor optical distribution means for optically distributing a first result vector Z₁ from the N₁ light transmitters of an first optoelectronic processor OP₁ to the N₁ light detectors of each of the M₂ optoelectronic processing elements of a next optoelectronic processor OP₂ and so on, processor-to-processor as each in turn performs an algebraic operation, the distributing being so that each bit Z₁ of the first result vector Z₁ of the first optoelectronic processor OP₁ is transmitted to the ith light detector of all M₁ rows of processing elements of the next optoelectronic processor OP₂ and so on;

wherein a last optically-encoded result vector Z_L is derived as an algebraic manipulation of the original optically-encoded input vector X by a series of matrices Y_k, k equals 1 through L, as are held within each of the plurality L of optoelectronic processors.

71. An optically-mapping optoelectronic tandem vector-vector algebraic processing system

for algebraically manipulating an N-bit vector X in accordance with an N-bit vector Y to produce a N×

N-bit result matrix Z, and, in tandem,

for algebraically manipulating the N-bit vector Y in accordance with the N-bit vector X to separately produce the N×

N-bit result matrix Z,

wherein one reason for separately-producing two same-result matrices Z is that these matrices are located separately, and are useful in intrinsic vector outer product computations, the vector-vector algebraic processing system for producing two result matrices in tandem comprising;

a first optoelectronic processor 1 having a plurality of N arrayed optoelectronic processing elements 1, 2, . . . N, each processing element i, i equals 1 through N, comprising

a plurality of N light detectors, each for detecting a data bit X_j of an optically-encoded input vector X, j equals 1 through N,

a plurality of at least N local memories at least one of which is for storing a data bit Y₁ of a stored vector Y,

an electrical computational means for electrically performing an algebraic operation on the input vector X detected by the N light detectors in consideration of the stored data vector Y held within the N local memories to produce a result bit Z_i of a result vector Z,

a light transmitter for transmitting a received bit-signal as an optically-encoded output signal,

a bi-directional electrical distribution means (i) for, at a first time, communicating the data bit Y;

of the vector Y as is stored in at least one of the N local memories to the light transmitter in order that it may be transmitted as an output signal, and, (ii) for, at a second time, distributing the result bit Z_i to the N local memories for storage therein as a new result data bit Z_ij of the result matrix Z, and

wherein the collective local memories of the collective N optoelectronic processing elements thus initially hold a vector Y of size N bits;

wherein, at the first time, the collective light transmitters of the collective N optoelectronic processing elements transmit this vector Y of N bits;

wherein the algebraic operation performed by the collective optoelectronic processing elements of the first optoelectronic processor is thus the manipulation of an N-bit input vector X by a N-bit stored vector Y to produce a N×

N-bit result matrix Z which is stored in the collective local memories of the first optoelectronic processor;

a second optoelectronic processor 2 having a plurality of N arrayed optoelectronic processing elements 1, 2, . . . N, each processing element i, i equals 1 through N, comprising

a plurality of N light detectors, each for detecting a data bit Y_j of an optically-encoded input vector Y, j equals 1 through N,

a plurality of at least N local memories at least one of which is for storing a data bit X_i of a stored vector X,

an electrical computational means for electrically performing an algebraic operation on the input vector Y detected by the N light detectors in consideration of the stored data vector X held within the N local memories to produce a result bit Z_i of a result vector Z,

a light transmitter for transmitting a received bit-signal as an optically-encoded output signal,

a bi-directional electrical distribution means (i) for, at a first time, communicating the data bit X_i of the vector X as is stored in at least one of the N local memories to the light transmitter in order that it may be transmitted as an output signal, and, (ii) for, at a second time, distributing the result bit Z₁ to the N local memories for storage therein as a new result data bit Z_ij of the result matrix Z, and

wherein the collective local memories of the collective N optoelectronic processing elements thus initially hold a vector X of size N bits;

wherein, at the first time, the collective light transmitters of the collective N optoelectronic processing elements transmit this vector X of N bits;

wherein the algebraic operation performed by the collective optoelectronic processing elements of the second optoelectronic processor is thus the manipulation of an N-bit input vector Y by a N-bit stored vector X to produce a N×

N-bit result matrix Z which is stored in the collective local memories of the first optoelectronic processor;

a first-processor-to-second-processor optical distribution means for optically distributing the data vector Y from the N light transmitters of the first processor to each of the N light detectors of each of the N processing elements of the second processor, this distributing being so that each bit Y_i of the data vector Y of the first processor is transmitted to the ith light detector of all N rows of processing elements of the second processor; and

a second-processor-to-first-processor optical distribution means for optically distributing the data vector X from the N light transmitters of the second processor to each of the N light detectors of each of the N processing elements of the first processor, this distributing being so that each bit X_i of the data vector X of the second processor is transmitted to the ith light detector of all N rows of processing elements of the first processor; and

wherein a result matrix Z is derived in the first processor as an algebraic manipulation of the original optically-encoded input vector X by a the vector Y that was held within local memories of the plurality of N arrayed optoelectronic processing elements;

wherein the same result matrix Z is derived in the second processor as an algebraic manipulation of the original optically-encoded input vector Y by a the vector X that was held within local memories of the plurality of N arrayed optoelectronic processing elements.

72. An optoelectronic matrix-algebraic processing system operating on external data vectors in order to solve a matrix algebraic problem of M×

N size, the system comprising;

two optoelectronic processors 1,2 each having M optoelectronic processing elements arrayed in N rows, each processing element PE_i, i equals 1 through M, comprising

N light detectors, each for detecting a data bit X_j of an optically-encoded input vector X, j equals 1 through M,

N local memories each for storing a data bit Y_ij of a stored vector Y, j equals 1 through N,

an electrical computational means for electrically performing a matrix algebraic operation on the input vector X detected by the N light detectors and on the stored data vector Y held within the N local memories to produce a result bit Z;

of a result vector Z, and

an optical transmitter for transmitting the result bit Z_i of the result vector Z off the plane of the arrayed processing elements;

wherein the collective optical transmitter of the M optoelectronic processors transmit an M-bit result vector Z;

a first-processor-to-second-processor optical distribution means for distributing a first result vector Z₁ from the M light transmitters of a first processor to the M light detectors of each of the N optoelectronic processing elements of a second processor, the distributing being so that each bit Z_1i of the first data vector Z₁ of the first processor is transmitted to the ith light detector of all N rows of processing elements of the second processor;

a second-processor-to-first-processor optical distribution means for distributing a first result vector Z from the N light transmitters of the second processor to the N light detectors of the first processor, the distributing being so that each bit Z_2i of the second data vector Z₂ of the second processor is transmitted to the jth light detector of the ith row of processing elements of the first processor.

74. A dual-topology H-tree optoelectronic matrix-algebraic processing system comprising:

two planar optoelectronic processors, each comprising

one or more processing elements in an H-tree topology, each processing element comprising

an N×

M array of leaf node light detectors receiving optically-encoded information from off-plane, electrically connected by equal-length first electrical paths to

intermediate-node electrical circuits for electrically processing the received information, the electrical circuits being electrically connected to one another in a tree by second electrical paths that are of substantially the same, equal-length, length as are the first electrical paths, and finally electrically connecting by third electrical paths that are again substantially the same, equal-length, length as are the first electrical paths to

a plurality of N light transmitters for optically transmitting the processed information off-plane; and

a first optical system for mapping the optical off-plane broadcast of the N light transmitters of a first processor onto the M light detectors of a second processor; and

a second optical system for mapping the optical off-plane broadcast of the N light transmitters of the second processor onto the M light detectors of the first processor;

wherein the H-tree topology of the processing elements supports that all the light detectors and electrical circuits and optical transmitter are electrically connected by paths of substantially the same length;

wherein a duality of the topology arises because the optoelectronic processors are two in number.

75. A method of optically communicating information between two-dimensional planes of an optoelectronic processor having a plurality of successive two-dimensional planes, and

a plurality of optoelectronic processing elements, each which optoelectronic processing element has (i) an energy transmitter, by which transmitter information is encoded and transmitted off plane in third dimension, and (ii) a plurality of energy receivers, by each of which receivers encoded information is optically received from off plane in the third dimension, the method comprising:

first-arraying the optoelectronic processing elements in at least one of the plurality of planes in a first pattern at a first scale;

wherein because the optoelectronic processing elements are first-arrayed in the first pattern at the first scale then so also are the energy transmitters arrayed in the same first pattern at the same first scale;

second-arraying the plurality of energy receivers that are within each of the optoelectronic elements of a next successive one of the plurality of planes to be in a second pattern at a second scale which is smaller than the first scale;

demagnifying via optics a first pattern of energy, which first energy pattern is encoded and transmitted from the collective energy transmitters as are within the collective plurality of optoelectronic processing elements as are first-arrayed within the at least one plane, to the size of the second pattern, and

replicating via optics the demagnified encoded energy pattern so that it is received into the energy receivers that are within each of the plurality of optoelectronic processing elements of the next subsequent plane;

therein to achieve optical connection from the collective energy transmitters as are within the collective plurality of optoelectronic processing elements of the one plane to the energy receivers that are within each of the plurality of optoelectronic processing elements of the next subsequent plane.

76. An optoelectronic processor comprising:

a plurality of successive two-dimensional planes each of which has and contains

a plurality of optoelectronic processing elements, two-dimensionally first-arrayed in a first pattern at a first scale, each of which optoelectronic processing elements has and contains

an energy transmitter, by which transmitter information is encoded and transmitted off plane in third dimension, and

a plurality of energy receivers, by each of which receivers encoded information is optically received from off plane in the third dimension, arrayed in a second pattern at a second scale which is smaller than the first scale

wherein because the optoelectronic processing elements are first-arrayed in the first pattern at the first scale then so also are the energy transmitters of the plural optoelectronic processing elements arrayed in the same first pattern at the same first scale;

optical demagnification means for demagnifying a first pattern of energy, which first energy pattern is encoded and transmitted from the collective energy transmitters as are within the collective plurality of optoelectronic processing elements as are first-arrayed within a plane, to the size of the second pattern; and

optical replication means for replicating the demagnified encoded energy pattern so that it is received into the energy receivers that are within each of the first-arrayed plurality of optoelectronic processing elements of a next subsequent plane;

wherein the optical demagnification means and the optical replication means jointly optically communicatively connect the collective energy transmitters as are within the collective plurality of optoelectronic processing elements of the one plane to the energy receivers that are within each of the plurality of optoelectronic processing elements of the next subsequent plane.

77. An optoelectronic processor comprising:

a plurality of successive two-dimensional planes each of which has and contains

a plane-associated number of optoelectronic processing elements, two-dimensionally first-arrayed in an plane-associated first pattern at a plane-associated first scale, each of which optoelectronic processing elements has and contains

an energy transmitter, by which transmitter information is encoded and transmitted off plane in third dimension, and

a plurality of energy receivers, by each of which receivers encoded information is optically received from off plane in the third dimension, arrayed in a plane-associated second pattern at a plane-associated second scale which is smaller than the first scale,

wherein because the optoelectronic processing elements are first-arrayed in the plane-associated first pattern at the plane-associated first scale then so also are the energy transmitters of the plural optoelectronic processing elements arrayed in the same plane-associated first pattern at the same first scale,

wherein any of (i) the numbers of optoelectronic processing elements, (ii) the plane-associated first pattern, (iii) the plane-associated first scale, (iv) the plane-associated second pattern, and (v) the plane-associated second scale, may vary from one plate to the next, and at least some do so very;

wherein (i) the numbers of optoelectronic processing elements varies as between at least two successive planes out of the plurality thereof; and

optical distribution means for optically distributing a first pattern of energy encoded and transmitted from the collective energy transmitters that are within a collective plurality of optoelectronic processing elements as are first-arrayed within a one plane to the energy receivers that are within each of the first-arrayed plurality of optoelectronic processing elements of a next subsequent plane;

wherein, between the at least two successive planes that do not contain equal numbers of optoelectronic processing elements, the optical distribution means suffices to optically distribute energy from some collective number of energy transmitters as are within the collective plurality of optoelectronic processing elements of the one plane to another, different, numbers of energy receivers that are within each of the plurality of optoelectronic processing elements of the next subsequent plane;

wherein optical communication between planes accords both for fan-out from smaller to larger numbers of optoelectronic processing elements, and for fan-in from larger to smaller number of optoelectronic processing elements.