US 5,603,046 AGrant

Method for Complex Data Movement in a Multi-Processor Data Processing System

Issue Date:1997-02-11

•15 Claims

•120 Drawing Sheets

Abstract

A method for complex data movement in a multi-processor data processing system. In one embodiment, the multi-processor data processing system (10) includes an array (12) of data processors (50-65), a plurality of edge interface circuits (14,16), and a bus interface controller (22). In an alternate embodiment, multiprocessor data processing system (210) includes an array (212) of data processors (250-258), a plurality of edge interface circuits (214-217), and a bus interface controller (222). The data processing systems (10,210) are capable of performing complex data movement patterns between the processors (50-65,250-258) and the corresponding edge interface circuits (14, 16, 214-217), such as a transpose pattern, a ping-pong pattern, and a checkerboard pattern.

Metadata

Assignee

Motorola Inc.

Inventors

Michael F. Wiles
Meltin Bell
Michael G. Gallup
L. Rodney Goke
Jack R. Davis
Erik L. Welty

Application Information

Application Number:US 5062578

Filing Date:1995-07-24

Priority Date:1993-11-02

Art Unit:232

Classifications

IPC:

G06F 1300

Field of Search:

395800;DIG. 1;200.15;840;841;851;853

Patent Drawings (120 sheets)

Description

Reference to Related Applications

The present application is related to the following U.S. patent applications:

Application Ser. No. 08/144,378, filed Nov. 2, 1993, now U.S. Pat. No. 5,548,777, titled "A MULTIPROCESSOR DATA PROCESSING SYSTEM", by Jack R. Davis et al.; and

Application Ser. No. 08/144,381, filed Nov. 2, 1993, now abandoned, titled "METHOD AND APPARATUS FOR PROVIDING BUS PROTOCOL SIMULATION", by Erik L. Welty et al.

Field of the Invention

The present invention relates in general to a data processing system, and more particularly to a method for complex data movement in a multi-processor data processing system.

Background of the Invention

Fuzzy logic, neural networks, and other parallel, array oriented applications are becoming very popular and important in data processing. Most digital data processing systems today have not been designed with fuzzy logic, neural networks, and other parallel, array oriented applications specifically in mind. Thus there are considerable performance and cost benefits to be gained in designing digital data processing systems which are especially adapted and designed to meet the requirements of fuzzy logic, neural networks, and other parallel, array oriented applications.

Summary of the Invention

The previously mentioned needs are fulfilled and other advantages achieved with the present invention. In one form, the present invention is a method for transferring data in a data processing system.

In one embodiment, the data processing system has a plurality of processors, interface circuitry, and switch circuitry. The plurality of processors is arranged in an array having a first row, an Nth row, a first column, and an Mth column. The interface circuitry is coupled to the Nth row of the array, and the switch circuit is coupled to the interface circuitry and to the Mth column of the array.

The method includes the step of storing a first data value in a first one of the plurality of processors. The first one of the plurality of processors is located in the Nth row of the array and in the first column of the array. The method includes the step of storing a second data value in a second one of the plurality of processors. The second one of the plurality of processors is located in the Nth row of the array and in the Mth column of the array.

The method also includes the step of transferring the first data value from the first one of the plurality of processors to the interface circuitry, and the step of transferring the second data value from the second one of the plurality of processors to the interface circuitry. The method includes the step of transferring the first data value from the interface circuitry to the switch circuit, and the step of transferring the second data value from the interface circuitry to the switch circuit. The method includes the step of transferring the first data value from the switch circuit to a third one of the plurality of processors. The third one of the plurality of processors is located in the first row of the array and in the Mth column of the array. The method also includes the step of transferring the second data value from the switch circuit to the second one of the plurality of processors.

In an alternate embodiment, the data processing system has a plurality of processors, first interface circuitry, second interface circuitry, and a data storage element. The plurality of processors is arranged in an array having a first array port and a second array port. The first interface circuitry is coupled to the first array port. The second interface circuitry is coupled to the second array port. The data storage element is coupled to the first interface circuitry and to the second interface circuitry.

The method includes the step of storing a first data value in one of the data storage element and the array. The method includes the step of storing a second data value in one of the data storage element and the array. The method includes the step of storing a third data value in one of the data storage element and the array. The method includes the step of storing a fourth data value in one of the data storage element and the array.

The method also includes the step of transferring the first data value between the data storage element and the first interface circuitry, while concurrently transferring the second data value between the array and the second interface circuitry. And, the method includes the step of transferring the third data value between the data storage element and the second interface circuitry, while concurrently transferring the fourth data value between the array and the first interface circuitry.

In yet another alternate embodiment, the data processing system has a plurality of processors, a first interface circuit, and a second interface circuit. The plurality of processors is arranged in an array having a first row, an Nth row, a first column, and an Mth column. The first interface circuit is coupled to the Nth row of the array, and the second interface circuit is coupled to the Mth column of the array. A first one of the plurality of processors is located in the first row and the first column. A second one of the plurality of processors is located in the first row and the Mth column. A third one of the plurality of processors is located in the Nth row and the first column. And, a fourth one of the plurality of processors is located in the Nth row and the Mth column. Each of the plurality of processors has a control register.

The method includes the step of storing a first control value in the control register of the first one of the plurality of processors, and the step of storing the first control value in the control register of the fourth one of the plurality of processors. The method includes the step of storing a second control value in the control register of the second one of the plurality of processors, and the step of storing the second control value in the control register of the third one of the plurality of processors. The method includes the steps of selecting the first row, selecting the Nth row, selecting the first column, and selecting the Mth column. The method includes the step of transferring data to the first interface circuit from the first one of the plurality of processors and from the fourth one of the plurality of processors, while concurrently transferring data to the second interface circuit from the second one of the plurality of processors and from the third one of the plurality of processors.

The present invention will be understood by one skilled in the art from the detailed description below in conjunction with the accompanying drawings.

Brief Description of the Drawings

FIG. 1 illustrates, in block diagram form, a data processing system 10 in accordance with one embodiment of the present invention;

FIG. 2 illustrates, in block diagram form, an array of processors (AEs) of FIG. 1 in accordance with one embodiment of the present invention;

FIG. 3 illustrates, in block diagram form, an edge interface of FIG. 1 in accordance with one embodiment of the present invention;

FIG. 4 illustrates, in block diagram form, bus protocol simulation logic (BPS) of FIG. 3 in accordance with one embodiment of the present invention;

FIG. 5 illustrates, in block diagram form, a portion of data processing system 10 of FIG. 1;

FIG. 6 illustrates, in block diagram form, a bus interface controller (BIC) of FIG. 1 in accordance with one embodiment of the present invention;

FIG. 7 illustrates, in block diagram form, a portion of the bus interface controller (BIC), the control bus, and the data bus of FIG. 6 in accordance with one embodiment of the present invention;

FIG. 8 illustrates, in block diagram form, a system memory of FIG. 1 in accordance with one embodiment of the present invention;

FIG. 9 illustrates, in block diagram form, a data processing system 210 in accordance with one embodiment of the present invention;

FIG. 10 illustrates, in block diagram form, an array of processors (AEs) of FIG. 9 in accordance with one embodiment of the present invention;

FIG. 11 illustrates, in block diagram form, an edge interface of FIG. 9 in accordance with one embodiment of the present invention;

FIG. 12-1 illustrates, in block diagram form, bus protocol simulation logic (BPS) of FIG. 11 in accordance with one embodiment of the present invention;

FIG. 12-2 illustrates, in block diagram form, a portion of memory 290 of FIG. 12-1 in accordance with one embodiment of the present invention;

FIG. 12-3 illustrates, in block diagram form, a transpose data movement pattern in accordance with one embodiment of the present invention;

FIG. 12-4 illustrates, in block diagram form, a ping-pong data movement pattern in accordance with one embodiment of the present invention;

FIG. 12-5 illustrates, in block diagram form, a checkerboard data movement pattern in accordance with one embodiment of the present invention;

FIGS. 13 through 30 illustrate, in schematic diagram form, a data processing system 10 of FIG. 1 in accordance with one embodiment of the present invention;

FIGS. 31A-31G illustrate, in tabular form, a parts list for the schematic diagram of data processing system 10 of FIGS. 13 through 30 in accordance with one embodiment of the present invention;

FIGS. 32 through 49-2 illustrate, in schematic diagram form, a data processing system 210 of FIG. 9 in accordance with one embodiment of the present invention; and

FIGS. 50A-50G illustrate, in tabular form, a parts list for the schematic diagram of data processing system 210 of FIGS. 32 through 49-2 in accordance with one embodiment of the present invention.

Description of the Preferred Embodiments

The term "bus" will be used to refer to a plurality of signals or conductors which may be used to transfer one or more various types of information, such as data, addresses, control, or status. The terms "assert" and "negate" will be used when referring to the rendering of a signal, status bit, or similar apparatus into its logically true or logically false state, respectively. If the logically true state is a logic level one, the logically false state will be a logic level zero. And if the logically true state is a logic level zero, the logically false state will be a logic level one. The symbol "%" preceding a number indicates that the number is represented in its binary or base two form. The symbol "$" preceding a number indicates that the number is represented in its hexadecimal or base sixteen form.

FIGS. 1 through 8 illustrate a high level block diagram of one embodiment of data processing system 10. For more detail, refer to the complete board level schematic of data processing system 10 illustrated in FIGS. 13 through 31-2. FIGS. 9 through 12-2 illustrate a high level block diagram of one embodiment of data processing system 210. For more detail, refer to the complete board level schematic of data processing system 210 illustrated in FIGS. 32 through 50-2.

FIG. 1 illustrates one embodiment of a data processing system 10. Data processing system 10 has an Association Engine (AE) array 12, an edge interface circuit 14, an edge interface circuit 16, a switch circuit 18, a daughter board electrical connector 20, a system memory circuit 24, a host computer 28 coupled to a bus interface controller (BIC) circuit 22 by way of serial interface 40, and a host computer 30 coupled to a small computer system interface (SCSI) circuit 26 by way of SCSI connector 42.

The north and west ports of the AE array 12 are coupled to a control bus 36, an address bus 34, and a daughter board connector 20. Edge interface 14 is bi-directionally coupled to the south port of AE array 12. Edge interface 16 is bi-directionally coupled to the east port of AE array 12. Edge interface 14 is bi-directionally coupled to data bus 32 and to control bus 36, and is coupled to receive information from address bus 34. Edge interface 16 is bi-directionally coupled to data bus 32 and to control bus 36, and is coupled to receive information from address bus 34. Daughter board connector 20 is bi-directionally coupled to daughter board bus (DB bus) 38 and to control bus 36.

Bus interface controller 22 is bi-directionally coupled to data bus 32, address bus 34, and control bus 36. System memory 24 is bi-directionally coupled to data bus 32, and is coupled to receive information from address bus 34 and control bus 36. The SCSI circuit 26 is bi-directionally coupled to data bus 32, address bus 34, and control bus 36. Switch 18 receives control information from control bus 36, receives data from edge interface 14, and provides information to edge interface 16.

Host computer 30 includes a hard disk drive with a SCSI interface (not shown). In an alternate embodiment of the present invention, bus interface controller 22 and SCSI circuit 26 may be coupled to the same host computer 30. In yet other embodiments of the present invention, only one of host computer 28 and host computer 30 may be present.

In one embodiment of the present invention, bus interface controller 22 is a data processor which is capable of executing software instructions. The software instructions are used to control the transfer of information across the data bus 32, the address bus 34, and the control bus 36. Instructions and/or data may be downloaded to the bus interface controller 22 from the host computer 28 by way of serial interface 40. Data may then be transferred back to the host computer 28 from the bus interface controller 22 by way of the serial interface 40. The system memory 24 may be used to store software instructions and/or data. The SCSI circuit 26 may be used to transfer software instructions and/or data from the host computer 30 by way of SCSI connector 42. Data may then be transferred back to the host computer 30 from the SCSI circuit 26 by way of the SCSI connector 42.

Daughter board connector 20 may be used to input and output data to and from a daughter board (not shown). The configuration of the optional daughter board will be determined by the particular application in which data processing system 10 is being used. Edge interfaces 14 and 16 provide the ability to interface two different bus types operating at potentially different frequencies. The AE array 12 requires a particular bus protocol in order to communicate information between the AE array 12 and the edge interfaces 14 and 16. The data bus 32, the address bus 34, and the control bus 36 require a different protocol from the AE array 12 in order to communicate information between the busses 32, 34, and 36 and the edge interfaces 14 and 16. The edge interfaces 14 and 16 therefore must be able to communicate using the two different bus protocols.

FIG. 2 illustrates one embodiment of an array of processors (AEs) 12 of FIG. 1. Each Association Engine integrated circuit 50-65 is merely one embodiment of a data processor which may be used to form array 12. Other data processors may be used to form array 12 in FIG. 1. A detailed description of the structure and operation of the Association Engine data processor is described in a related, copending application entitled "A Data Processing System And Method Thereof", invented by Michael G. Gallup et al., having Ser. No. 08/040,779, filed Mar. 31, 1993, and assigned to the assignee hereof, which is expressly incorporated by this reference.

The combined north ports of AE processors 50-53 form a north array port of AE array 12. The combined south ports of AE processors 62-65 form a south array port of AE array 12. The combined west ports of AE processors 50, 54, 58, and 62 form a west array port of AE array 12. The combined east ports of AE processors 53, 57, 61, and 65 form an east array port of AE array 12. Although the illustrated embodiment of AE array 12 illustrates a four column by four row array of AE processors 50-65, an alternate embodiment of the present invention may use any number of rows and any number of columns.

Note that each AE processor 50-65 provides a busy signal, BUSY (not shown), and an interrupt signal, INTR (not shown), to the edge interfaces 14 and 16. The busy signal and the interrupt signal are not wire-ORed together. Also, each AE processor 50-65 has an enable pin, EN (not shown), all of which are wired-ORed together and coupled to edge interfaces 14 and 16. Each AE processor 50-65 has an access type pin, OP (not shown), all of which are wired-ORed together and coupled to edge interfaces 14 and 16. Each AE processor 50-65 has a read/write control pin, R/W (not shown), all of which are wired-ORed together and coupled to edge interfaces 14 and 16. And each AE processor 50-65 has a run/stop pin, R/S (not shown), all of which are wired-ORed together and coupled to edge interfaces 14 and 16.

FIG. 3 illustrates one embodiment of edge interface circuits 14 and 16 of FIG. 1. Edge interface circuits 14 and 16 each have bus protocol simulation logic circuits 70-73, address generator circuit 74, duration counter circuit 76, buffer circuit 78-81, and high speed address bus 82, which are coupled as illustrated in FIG. 3. Note that the term "X" in signal names "XCI" and "XCO" is used as a variable. For edge interface 14, which is coupled to the south array port, "XCI" represents "SCI", and for edge interface 16, which is coupled to the east array port, "XCI" represent "ECI". Likewise, for edge interface 14, which is coupled to the south array port, "XCO" represents "SCO", and for edge interface 16, which is coupled to the east array port, "XCO" represent "ECO". Also note that in one embodiment of the present invention, the "XCI" and "XCO" signals are active low.

FIG. 4 illustrates one embodiment of bus protocol simulation logic circuits 70-73 of FIG. 3. Bus protocol simulation logic circuits 70-73 each have a memory circuit 90, a programmable logic device circuit (PLD) 92, a host address buffer circuit 94, a host data transceiver circuit 96, and an AE transceiver circuit 98, which are coupled as illustrated in FIG. 4. Note that in one embodiment of the present invention, the memory 90 is a byte wide memory storing data bits[0:7], which is the same width as the data path between each AE processor 50-65. Note that each bus protocol simulation logic circuit 70-73 is coupled to one port of a corresponding one of the AE processors 50-65.

FIG. 5 illustrates one embodiment of switch 18 of FIG. 1. Switch 18 has a plurality of transpose buffer circuits 100-103, which are coupled to edge interface 14 and edge interface 16.

FIG. 6 illustrates one embodiment of bus interface controller 22 of FIG. 1. Bus interface controller 22 has a programmable logic circuit 110, a control processor 112, a decoder array 114, and a plurality of registers, including other registers 116, busy status register 118, busy mask register 120, and comparator 122, which are coupled as illustrated in FIG. 6.

FIG. 7 illustrates one embodiment of a portion of bus interface controller 22 of FIG. 6. Busy mask register 120 stores a plurality of busy mask bits[0:15]. Each AE processor 50-65 provides one busy signal to comparator 122. If the busy mask bit for the corresponding AE processor 50-65 is negated, then the busy signal for that AE processor 50-65 is excluded from the comparison. Comparator 122 compares the non-excluded busy signals and asserts the global done output signal when all non-excluded busy signals have been negated, indicating that all of the non-excluded AE processors have completed operation.

FIG. 8 illustrates one embodiment of system memory 24 of FIG. 1. In one embodiment, system memory 24 is implemented using an EEPROM (Electrically Erasable Programmable Read Only Memory) 126, and a SRAM (Static Random Access Memory). Alternate embodiments of the present invention may use any type of memory for system memory 24.

FIGS. 9 through 12-2 illustrate a high level block diagram of one embodiment of data processing system 210. For more detail, refer to the complete board level schematic of data processing system 210 illustrated in FIGS. 32 through 50-2.

FIG. 9 illustrates one embodiment of a data processing system 210. Data processing system 210 has an Association Engine (AE) array 212, an edge interface circuit 214, an edge interface circuit 215, an edge interface circuit 216, an edge interface circuit 217, a host computer interface logic circuit 244, a host computer connector 246, and a host computer 248.

Edge interface 214 is bi-directionally coupled to the south port of AE array 212. Edge interface 215 is bi-directionally coupled to the north port of AE array 212. Edge interface 216 is bi-directionally coupled to the east port of AE array 212. And, edge interface 217 is bi-directionally coupled to the west port of AE array 212. Edge interface 214 is bi-directionally coupled to data bus 232 and to control bus 236, and is coupled to receive information from address bus 234. Edge interface 215 is bi-directionally coupled to data bus 232 and to control bus 236, and is coupled to receive information from address bus 234. Edge interface 216 is bi-directionally coupled to data bus 232 and to control bus 236, and is coupled to receive information from address bus 234. Edge interface 217 is bi-directionally coupled to data bus 232 and to control bus 236, and is coupled to receive information from address bus 234.

Host computer interface logic 244 is bi-directionally coupled to data bus 232 and control bus 236, and is coupled to provide information to address bus 234. Host computer interface logic 244 is bi-directionally coupled to host computer connector 246. Host computer connector 246 is bi-directionally coupled to host computer 248.

FIG. 10 illustrates one embodiment of an array of processors (AEs) 212 of FIG. 9. Each Association Engine integrated circuit 250-258 is merely one embodiment of a data processor which may be used to form array 212. Other data processors may be used to form array 212 in FIG. 9. Although the illustrated embodiment of AE array 212 illustrates a three column by three row array of AE processors 250-258, an alternate embodiment of the present invention may use any number of rows and any number of columns.

The combined north ports of AE processors 250-252 form a north array port of AE array 212. The combined south ports of AE processors 256-258 form a south array port of AE array 212. The combined west ports of AE processors 250, 253, and 256 form a west array port of AE array 212. The combined east ports of AE processors 252, 255, and 258 form an east array port of AE array 212.

FIG. 11 illustrates one embodiment of edge interface circuits 214-217 of FIG. 9. Edge interface circuits 214-217 each have bus protocol simulation logic circuits 270-272, address generator circuit 274, duration counter circuit 276, buffer circuit 278-280, and high speed address bus 282, which are coupled as illustrated in FIG. 11. Note that the term "X" in signal names "XCI" and "XCO" is used as a variable. For example, for edge interface 214, which is coupled to the south array port, "XCI" represents "SCI", and "XCO" represents "SCO".

FIG. 12-1 illustrates one embodiment of bus protocol simulation logic circuits 270-272 of FIG. 11. Bus protocol simulation logic circuits 270-272 each have a memory circuit 290, a programmable logic device circuit (PLD) 292, a host address buffer circuit 294, a host data transceiver circuit 296, and an AE transceiver circuit 928, which are coupled as illustrated in FIG. 12-1. Note that each bus protocol simulation logic circuit 70-73 is coupled to one port of a corresponding one of the AE processors 50-65.

Note that in one embodiment of the present invention, the memory 290 is a 9-bit wide memory storing data bits[0:8], which is one bit more than the width of the data path between each AE processor 250-258. The extra bit is provided to programmable logic device 292. The programmable logic device 292 uses the extra bit as a control bit which is used to determine whether to assert or negate the "XCI" signal provided to the "XCI" input integrated circuit pin of the corresponding AE processor 250-258. Each location in memory 290 has a control portion 291 and a data portion 293. Although the control portion 291 in the illustrated embodiment of the present invention is one bit, alternate embodiments of the present invention may have a plurality of bits in the control portion 291. And although the data portion 293 is illustrated in FIG. 12-1 as having eight bits, alternate embodiments of the present invention may use any number of bits.

FIG. 12-2 illustrates one embodiment of a portion of memory 290 of FIG. 12-1. It is important to remember that data processing system 210 includes a plurality of memories 290. Data bit-8, labeled "D8", stores a single control bit which is provided to programmable logic device 292. In one embodiment of the present invention, programmable logic device 292 determines the proper timing and the proper conditions to allow the control bit D8 to be provided to the "XCI" input of the corresponding AE processor 250-258. When the proper conditions have been met, the programmable logic device 292 does not affect the value of the control bit D8 which is transferred to the corresponding AE processor 250-258.

Due to the bus protocol and timing required by the AE processors 250-258, the control bit D8 in address location "N" is associated with the data bits D0-D7 in address location "N+1". For example, the control bit %0 in address location $X00 is associated with the data value $21 in address location $X01. When the control bit D8 in address location "N" is asserted, the data bits D0-D7 in the address location "N+1" are transferred, at the proper time, from memory 290 to the data input integrated circuit pins of the corresponding AE processor 250-258 by way of AE transceiver 298. Note that the term "X" in addresses $X00 through $X09 is used as a variable indicating that the address space $X00 through $X09 may be located at any convenient location in a memory map.

As an example, in one embodiment of the present invention, %0 is the negated state of the control bit D8, %1 is the asserted state of the control bit D8, memory 290 is coupled to AE processor 250, and memory 290 stores the values as indicated in FIG. 12-2. The data value $20 in address location $X00 is not used. Control bit D8 from address location $X00 is transferred to programmable logic device 292. Because the control bit is %0, no transfer is initiated by the programmable logic device 292 and the data value $21 in address location $X01 is not used. Next, control bit D8 from address location $X01 is transferred to programmable logic device 292. Because the control bit is %0, no transfer is initiated by the programmable logic device 292 and the data value $22 in address location $X02 is not used.

Next, control bit D8 from address location $X02 is transferred to programmable logic device 292. Because the control bit is %1, a transfer is initiated by the programmable logic device 292. On a first bus cycle, the programmable logic device 292 asserts the XCI signal provided to the AE processor 250. On the second bus cycle, the data value $23 from address location $X03 is transferred to AE processor 250. Control bits within each AE processor 250, 253, and 256 determine the movement of the data value $23 through the processor array 212. Each AE processor, if it receives the data value $23, has the choice of either storing or not storing the data value. In addition, each AE processor, if it receives the data value $23, has the choice of whether or not to pass the data value $23 on, and if so, in which direction.

Note that one bus cycle is required to transfer a data value from one AE port to the opposite AE port. For example, if AE processor 250 receives the data value $23 during the second bus cycle, then AE processor 250, if configured appropriately, can provide the data value $23 at the south port during the third bus cycle. And if AE processor 253 receives the data value $23 during the third bus cycle, then AE processor 253, if configured appropriately, can provide the data value $23 at the south port during the fourth bus cycle.

Also during the second bus cycle, control bit D8 from address location $X03 is transferred to programmable logic device 292. Because the control bit is %1, a transfer is initiated by the programmable logic device 292. On the second bus cycle, the programmable logic device 292 asserts the XCI signal provided to the AE processor 250. On the third bus cycle, the data value $24 from address location $X04 is transferred to AE processor 250. Control bits within each AE processor 250, 253, and 256 determine the movement of the data value $24 through the processor array 212. Each AE processor, if it receives the data value $24, has the choice of either storing or not storing the data value. In addition, each AE processor, if it receives the data value $24, has the choice of whether or not to pass the data value $24 on, and if so, in which direction.

Again, one bus cycle is required to transfer a data value from one AE port to another AE port. For example, if AE processor 250 receives the data value $24 during the third bus cycle, then AE processor 250, if configured appropriately, can provide the data value $24 at the south port during the fourth bus cycle. And if AE processor 253 receives the data value $24 during the fourth bus cycle, then AE processor 253, if configured appropriately, can provide the data value $24 at the south port during the fifth bus cycle. Note that by storing the asserted value "1" of the control bit D8 in consecutive address locations, data can be streamed into the AE processors (e.g. AE processors 250, 253, and 256) which are in the same row or column as corresponds to the memory 290. And by storing the negated value "0" of the control bit D8 in an address location, a break can be made in the stream of data be transferred to the AE processors (e.g. AE processors 250, 253, and 256) which are in the same row or column as corresponds to the memory 290.

Various methods for transferring data in data processing system 10 will now be discussed. Note that in some embodiments of the present invention, the tap and switch circuitry (not shown) in each AE processor 50-65 may be programmed to modify the data movement among the AE processors 50-65 in the AE array 12.

FIG. 12-3 illustrates a transpose data movement pattern using data processing system 10. Although the transpose data movement pattern is illustrated using data processing system 10, the transpose data movement pattern may be used to transfer data on other data processing systems. The transpose data movement pattern is especially useful for performing matrix computations which require a transpose operation. In one embodiment of the present invention, the matrix operations may involve one or more matrices of data stored in the Coefficient Memory Arrays (CMAs) (not shown) of the AE processors 50-65. In the example illustrated in FIG. 12-3, data is transferred out of the south array port of AE array 12 and into the east array port of AE array 12.

An example of the transpose data movement pattern will now be discussed. A data value is stored in each of AE processors 62-65. The data value stored in AE processor 62 is transferred to and stored in a corresponding memory 90, the data value stored in AE processor 63 is transferred to and stored in a corresponding memory 90, the data value stored in AE processor 64 is transferred to and stored in a corresponding memory 90, and the data value stored in AE processor 65 is transferred to and stored in a corresponding memory 90. The corresponding memories 90 are located in interface circuit 14 (see FIGS. 3 and 4). In one embodiment of the present invention, each time that the four data values from each of the processor 62-65 are transferred, they are transferred in parallel.

From the corresponding memories 90, the four data values are transferred to the transpose buffers 100-103 in switch 18 (see FIG. 5). The data value from AE processor 62 is transferred to transpose buffer 100, from transpose buffer 100 to interface circuit 16, and from interface circuit 16 to AE processor 53 where it is stored. The data value from AE processor 63 is transferred to transpose buffer 101, from transpose buffer 101 to interface circuit 16, and from interface circuit 16 to AE processor 57 where it is stored. The data value from AE processor 64 is transferred to transpose buffer 102, from transpose buffer 102 to interface circuit 16, and from interface circuit 16 to AE processor 61 where it is stored. The data value from AE processor 65 is transferred to transpose buffer 103, from transpose buffer 103 to interface circuit 16, and from interface circuit 16 to AE processor 65 where it is stored.

In alternate embodiments of the present invention, the data values may be stored in more, fewer, and different circuits as they are transferred. The bus interface controller 22 (see FIG. 1) is used to select and control the transpose data movement pattern by way of control bus 36.

FIG. 12-4 illustrates a ping-pong data movement pattern using data processing system 10. Although the ping-pong data movement pattern is illustrated using data processing system 10, the ping-pong data movement pattern may be used to transfer data on other data processing systems. The ping-pong data movement pattern is especially useful for quickly transferring a significant amount of data between the AE array 12 and the host processor 30.

An example of the ping-pong data movement pattern will now be discussed. A plurality of data values is stored in host computer 30. In one embodiment of the present invention, a large number of data values are stored on a hard disk drive of computer 30. The SCSI circuit 26 controls the transfer of data from the host computer to the memories 90 in edge interfaces 14 and 16.

A first set of data values are transferred from host computer 30 to the memories 90 in edge interface 14 by way of SCSI connector 42, SCSI circuit 26, and data bus 32. Then while the AE processors 50-65 are receiving and optionally storing the first set of data values from edge interface 14, a second set of data values are transferred from host computer 30 to the memories 90 in edge interface 16 by way of SCSI connector 42, SCSI circuit 26, and data bus 32. Then while the AE processors 50-65 are receiving and optionally storing the second set of data values from edge interface 16, a third set of data values are transferred from host computer 30 to the memories 90 in edge interface 14 by way of SCSI connector 42, SCSI circuit 26, and data bus 32. Note that a set of data values may include one or more bytes of data. Note also that each memory 90 in edge interface 14 may receive data at the same time in parallel, and may likewise provide data at the same time in parallel.

Thus using the ping-pong data movement pattern, the SCSI circuit 26 alternates between transferring data to/from edge interface 14 and to/from edge interface 16. Likewise, the AE array 12 alternates between transferring data to/from edge interface 14 and to/from edge interface 16. However, when the SCSI circuit is transferring data to/from edge interface 14, the AE array 12 is transferring data to/from edge interface circuit 16; and when the SCSI circuit is transferring data to/from edge interface 16, the AE array 12 is transferring data to/from edge interface circuit 14. The bus interface controller 22 (see FIG. 1) is used to select the ping-pong data movement pattern; however, the SCSI circuit 26 controls the actual data transfer once the ping-pong data movement pattern has been selected and initiated by the bus interface controller 22.

FIG. 12-5 illustrates a checkerboard data movement pattern using data processing system 10. Although the checkerboard data movement pattern is illustrated using data processing system 10, the checkerboard data movement pattern may be used to transfer data on other data processing systems. The checkerboard data movement pattern is especially useful in allowing both edge interfaces 14 and 16 to burst data to the AE array 12. In other words, Stream Accesses, in Host Mode, can be made simultaneously in the south-to-north direction, and in the east-to-west direction.

An example of the checkerboard data movement pattern will now be discussed. The bus interface controller 22 (see FIG. 1) is used to select and control the checkerboard data movement pattern by way of control bus 36. The bus interface controller 22 must cause the proper control values to be stored in the appropriate control bits of each AE processor 50-65 in order to permit the checkerboard data movement pattern to be used. The bus interface controller 22 selects all rows and all columns in AE array 12 by asserting both the row integrated circuit pin ROW and the column integrated circuit pin COL of each AE processor 50-65. The bus interface controller 22 also configures each AE processor 50-65 to be in Host Stream Mode. In addition, the bus interface controller 22 must cause the appropriate values to be stored in the Host Stream Select Port bits (HSP[1:0]) in each AE processor 50-65

Note that in Host Stream Mode, the AE processors 50-65 can receive and store data, can pass on data without storing, and can act as the original source of data. The Host Stream Select Port bits (HSP[1:0]) (not shown) in each AE processor 50-65 determine the movement of the data through the processor array 212. For write accesses, each AE processor receives a data value, and based upon the value of the HSP control bits, has the choice of either storing or not storing the data value before passing on the data value. For read accesses, the HSP control bits in each AE processor 50-65 determine whether that particular processor acts as an original source of the data value, or merely passes on the data value. Note that the checkerboard data movement pattern allows data to be transferred between AE processors 50-65 in the west-to-east direction and the north-to-south direction concurrently.

In the example illustrated in FIG. 12-5, the AE processors which are cross-hatched, namely 51, 53, 54, 56, 59, 61, 62, and 64 all have a %10 stored in the Host Stream Select Port bits (i.e. HSP control bits) and thus transfer data out the east port. On the other hand, the AE processors which are not cross-hatched, namely 50, 52, 55, 57, 58, 60, 63, and 65, all have a %01 stored in the HSP control bits and thus transfer data out the south port.

As an example, for north-to-south data movement, a data value sourced by AE processor 50 would be transferred to AE processor 54, then AE processor 58, then AE processor 62, and then finally to interface circuit 14 (i.e. memory 90 coupled to AE processor 62). A data value sourced by AE processor 55 would be transferred to AE processor 59, then AE processor 63, and then finally to interface circuit 14 (i.e. memory 90 coupled to AE processor 63). For west-to-east data movement, a data value sourced by AE processor 51 would be transferred to AE processor 52, then AE processor 53, and then finally to interface circuit 16 (i.e. memory 90 coupled to AE processor 53). A data value sourced by AE processor 54 would be transferred to AE processor 55, then AE processor 56, then AE processor 57, and then finally to interface circuit 16 (i.e. memory 90 coupled to AE processor 57). The above example is an example of a fine-grain checkerboard data movement pattern.

The present invention may also be used to perform a coarse-grain checkerboard data movement pattern. As an example, AE processors 50-51, 54-55, 60-61, and 64-65 may all have a %10 stored in the Host Stream Select Port bits (i.e. HSP control bits) and thus may transfer data out the east port. On the other hand, the AE processors 52-53, 56-59, and 62-63 may all have a %01 stored in the HSP control bits and may thus transfer data out the south port.

For west-to-east data movement, a data value sourced by AE processor 50 would be transferred to AE processor 51, then AE processor 52, then AE processor 53, and then finally to interface circuit 16 (i.e. memory 90 coupled to AE processor 53). A data value sourced by AE processor 51 would be transferred to AE processor 52, then AE processor 53, and then finally to interface circuit 16 (i.e. memory 90 coupled to AE processor 53). For north-to-south data movement, a data value sourced by AE processor 52 would be transferred to AE processor 56, then AE processor 60, then AE processor 64, and then finally to interface circuit 14 (i.e. memory 90 coupled to AE processor 64). A data value sourced by AE processor 56 would be transferred to AE processor 60, then AE processor 64, and then finally to interface circuit 14 (i.e. memory 90 coupled to AE processor 64).

The checkerboard data movement pattern thus allows data to be transferred from the AE array 12 to interface circuit 14 at the same time that data is being transferred from the AE array 12 to interface circuit 16. Note that the Host Stream Select Port bits (i.e. HSP control bits) in each AE processor 50-65 also determine whether the AE processor 50-65 sources and passes data, or merely passes data. In one embodiment of the present invention, if the HSP control bits are either %01 or %10, the AE processor 50-65 both sources and passes on data. However, if the HSP control bits are either %00 or %11, the AE processor 50-65 does not source data, but merely passes on data.

Note that more than one AE processor 50-65 in a column and more than one AE processor 50-65 in a row may source data. For example, in the fine-grain checkerboard data movement pattern illustrated in FIG. 12-5, AE processors 52 and 60 both have a %01 stored in the HSP control bits and thus source data out the south port. AE processor 52 sources data first, while AE processors 56, 60, and 64 merely pass on the data sourced by AE processor 52. Then when AE processor 52 has finished sending data, AE processor 60 sources data, while AE processor 64 merely pass on the data sourced by AE processor 60.

While the present invention has been illustrated and described with reference to specific embodiments, further modifications and improvements will occur to those skilled in the art. It is to be understood, therefore, that this invention is not limited to the particular forms illustrated and that it is intended in the appended claims to cover all modifications that do not depart from the spirit and scope of this invention.

Claims

We claim:

1. A method for transferring data in a data processing system, the data processing system comprising a plurality of processors, interface circuitry external to the plurality of processors, and switch circuitry external to the plurality of processors, the plurality of processors being arranged in an array having a first row, an Nth row, a first column, and an Mth column, the interface circuitry being coupled to each processor in the Nth row of the array, and the switch circuitry being coupled to the interface circuitry and to each processor in the Mth column of the array, the method comprising the steps of: storing a first data value in a first one of the plurality of processors, the first one of the plurality of processors being located in the Nth row of the array and in the first column of the array; storing a second data value in a second one of the plurality of processors, the second one of the plurality of processors being located in the Nth row of the array and in the Mth column of the array; transferring the first data value from the first one of the plurality of processors to the interface circuitry; transferring the second data value from the second one of the plurality of processors to the interface circuitry; transferring the first data value from the interface circuitry to the switch circuitry; transferring the second data value from the interface circuitry to the switch circuitry; transferring the first data value from the switch circuitry to a third one of the plurality of processors, the third one of the plurality of processors being located in the first row of the array and in the Mth column of the array; and transferring the second data value from the switch circuitry to the second one of the plurality of processors, wherein if the switch circuitry is used to transfer data to at least one of the plurality of processors located in the first column of the array, the data must be transferred to the at least one of the plurality of processors located in the first column of the array by passing through at least one of the plurality of processors located in the Mth column of the array.

2. A method as in claim 1, further comprising the step of: selecting a transpose data movement mode.

3. A method as in claim 1, further comprising the steps of: storing the first data value in the third one of the plurality of processors; and repeating said step of storing the second data value in the second one of the plurality of processors.

4. A method as in claim 1, wherein the array also has a second row, a third row, a second column, and a third column, the method further comprising the steps of: storing a third data value in a fourth one of the plurality of processors, the fourth one of the plurality of processors being located in the Nth row of the array and in the second column of the array; storing a fourth data value in a fifth one of the plurality of processors, the fifth one of the plurality of processors being located in the Nth row of the array and in the third column of the array; transferring the third data value from the fourth one of the plurality of processors to the interface circuitry; transferring the fourth data value from the fifth one of the plurality of processors to the interface circuitry; transferring the third data value from the interface circuitry to the switch circuitry; transferring the fourth data value from the interface circuitry to the switch circuitry; transferring the third data value from the switch circuitry to a sixth one of the plurality of processors, the sixth one of the plurality of processors being located in the second row of the array and in the Mth column of the array; and transferring the fourth data value from the switch circuitry to a seventh one of the plurality of processors, the seventh one of the plurality of processors being located in the third row of the array and in the Mth column of the array.

5. A method as in claim 1, wherein said step of transferring the first data value from the first one of the plurality of processors is performed concurrently with said step of transferring the second data value from the second one of the plurality of processors.

6. A method as in claim 5, wherein said step of transferring the first data value from the interface circuitry is performed concurrently with said step of transferring the second data value from the interface circuitry.

7. A method as in claim 6, wherein said step of transferring the first data value from the switch circuitry is performed concurrently with said step of transferring the second data value from the switch circuitry.

8. A method for transferring data in a data processing system, the data processing system comprising a plurality of processors, first interface circuitry, second interface circuitry, and a data storage element, the plurality of processors being arranged in an array having a first array port and a second array port, the first interface circuitry being coupled to the first array port, the second interface circuitry being coupled to the second array port, and the data storage element being coupled to the first interface circuitry and to the second interface circuitry, the method comprising the steps of: storing a first data value in one of the data storage element or the array; storing a second data value in one of the data storage element or the array; storing a third data value in one of the data storage element or the array; storing a fourth data value in one of the data storage element or the array; transferring the first data value between the data storage element and the first interface circuitry, while concurrently transferring the second data value between the array and the second interface circuitry; and transferring the third data value between the data storage element and the second interface circuitry, while concurrently transferring the fourth data value between the array and the first interface circuitry, wherein data transfer between the data storage element and the first interface circuitry is bi-directional, wherein data transfer between the data storage element and the second interface circuitry is bi-directional, wherein data transfer between the array and the second interface circuitry is bi-directional, and wherein data transfer between the array and the first interface circuitry is bi-directional.

9. A method as in claim 8, further comprising the step of: selecting a data movement mode, wherein said step of selecting comprises the step of programming at least one register control bit in each of the plurality of processors.

10. A method as in claim 8, wherein the first data value includes a first plurality of data bytes, wherein the second data value includes a second plurality of data bytes, wherein the third data value includes a third plurality of data bytes, and wherein the fourth data value includes a fourth plurality of data bytes.

11. A method for transferring data in a data processing system, the data processing system comprising a plurality of processors, a first interface circuit, and a second interface circuit, the plurality of processors being arranged in an array having a first row, an Nth row, a first column, and an Mth column, the first interface circuit being coupled to the Nth row of the array, and the second interface circuit being coupled to the Mth column of the array, a first one of the plurality of processors being located in the first row and the first column, a second one of the plurality of processors being located in the first row and the Mth column, a third one of the plurality of processors being located in the Nth row and the first column, and a fourth one of the plurality of processors being located in the Nth row and the Mth column, each of the plurality of processors having a control register, the method comprising the steps of: storing a first control value in the control register of the first one of the plurality of processors; storing the first control value in the control register of the fourth one of the plurality of processors; storing a second control value in the control register of the second one of the plurality of processors; storing the second control value in the control register of the third one of the plurality of processors; selecting the first row; selecting the Nth row; selecting the first column; selecting the Mth column; and transferring data to the first interface circuit which is sourced from the first one of the plurality of processors and which is sourced from the fourth one of the plurality of processors, while concurrently transferring data to the second interface circuit which is sourced from the second one of the plurality of processors and which is sourced from the third one of the plurality of processors, wherein data sourced by the first one of the plurality of processors is transferred to the first interface circuit by way of the third one of the plurality of processors, and wherein data sourced by the third one of the plurality of processors is transferred to the second interface circuit by way of the fourth one of the plurality of processors, and wherein said step of transferring is controlled, at least in part, by the first and second control values stored in the first, second, third, and fourth ones of the plurality of processors.

12. A method as in claim 11, wherein said step of transferring data further comprises the steps of: sourcing a first data value from the first one of the plurality of processors; receiving the first data value at the third one of the plurality of processors; and providing the first data value from the third one of the plurality of processors to the first interface circuit.

13. A method as in claim 12, wherein said step of transferring data further comprises the steps of: sourcing a second data value from the second one of the plurality of processors to the second interface circuit.

14. A method as in claim 13, wherein said step of transferring data further comprises the steps of: sourcing a third data value from the third one of the plurality of processors; receiving the third data value at the fourth one of the plurality of processors; and providing the third data value from the fourth one of the plurality of processors to the second interface circuit.

15. A method as in claim 14, wherein said step of transferring data further comprises the steps of: sourcing a fourth data value from the fourth one of the plurality of processors to the first interface circuit.

Patent Citations (37)

Patent	Date	Inventor
US3287703	1966-11-01	Slotnick
US3796992	1974-03-01	Nakamura et al.
US4463445	1984-07-01	Grimes
US4470112	1984-09-01	Dimmick
US4488218	1984-12-01	Grimes
US4523310	1985-06-01	Brown et al.
US4811201	1989-03-01	Rau et al.
US4908751	1990-03-01	Smith
US4925312	1990-05-01	Onaga et al.
US4974146	1990-11-01	Works et al.
US5067095	1991-11-01	Peterson et al.
US5073867	1991-12-01	Murphy et al.
US5083285	1992-01-01	Shima et al.
US5086405	1992-02-01	Chung et al.
US5140670	1992-08-01	Chua et al.
US5146420	1992-09-01	Vassiliadis et al.
US5148515	1992-09-01	Vassiliadis et al.
US5151874	1992-09-01	Jeong et al.
US5151971	1992-09-01	Jousselin et al.
US5152000	1992-09-01	Hills
US5155389	1992-10-01	Furtek
US5155699	1992-10-01	Chung et al.
US5157785	1992-10-01	Jackson et al.
US5165009	1992-11-01	Watanabe et al.
US5165010	1992-11-01	Masuda et al.
US5167008	1992-11-01	Engeler
US5168573	1992-12-01	Fossum et al.
US5175858	1992-12-01	Hammerstrom
US5182794	1993-01-01	Gasperi et al.
US5239654	1993-08-01	Ing-Simmons et al.
US5274818	1993-12-01	Vasilevsky et al.
US5321639	1994-06-01	Krishnamoorthy et al.
US5355453	1994-10-01	Row et al.
US5408676	1995-04-01	Mori
US5416854	1995-05-01	Fukuda et al.
US5421019	1995-05-01	Holsztynski et al.
US5450604	1995-09-01	Davies

Non-Patent Literature (41)

"Parallel Super Computing in SIMD Architectures", Hord, 1990 pp. 5-69.
"Multiple DSP Enviroment for Intelligent Signal Processing", SASS et al, 1987, IEEE vol. 75, No. 9 Sep. 1987.
"Data Flow Computers and VLSI Computations" Computer Architecture & Parallel Processing, Chap 10, pp. 752-811 Huang et al., 1984.
"Neural Networks Primer Part I" published in A1 Expert in Dec. 1987 and written by Maureen Caudill, pp. 46 through 52.
"Neural Networks Primer Part II" published in A1 Expert in Feb. 1988 and written by Maureen Caudill, pp. 55 through 61.
"Neural Networks Primer Part III" published in A1 Expert in Jun. 1988 and written by Maureen Caudill, pp. 53 through 59.
"Neural Networks Primer Part IV" published in A1 Expert in Aug. 1988 and written by Maureen Caudill, pp. 61 through 67.
"Neural Networks Primer Part V" published in A1 Expert in Nov. 1988 and written by Maureen Caudill, pp. 57 through 65.
"Neural Networks Primer Part VI" published in A1 Expert in Feb. 1989 and written by Maureen Caudill, pp. 61 through 67.
"Neural Networks Primer Part VII" published in A1 Expert in May 1989 and written by Maureen Caudill, pp. 51 through 58.
"Neural Networks Primer Part VIII" published in A1 Expert in Aug. 1989 and written by Maureen Caudill, pp. 61 through 67.
"Fast Spheres, Shadows, Textures, Transparencies, and Image Enhancements in Pixel Planes" by H. Fuchs et al. and published in Computer Graphics, vol. 19, No. 3, Jul. 1985, pp. 111-120.
"Pixel-Planes: Building a VLSI-Based Graphic System" by J. Poulton et al. and published in the proceedings of the 1985 Chapel Hill Conference on VLSI, pp. 35-60.
"Pixel-Planes 5: A Heterogeneous Multiprocessor Graphics System Using Processor-Enhanced Memories" by Fuchs et al. and published in Computer Graphics, vol. 23, No. 3, Jul. 1989, pp. 79-88.
"Parallel Processing in Pixel-Planes, a VLSI logic-enhanced memory for raster graphics" by Fuchs et al. published in the proceedings of ICCD' 85 held in Oct., 1985, pp. 193-197.
"Building a 512.times.512 Pixel-Planes System" by J. Poulton et al. and published in Advanced Research in VLSI, Proceedings of the 1987 Stanford Conference, pp. 57-71.
"Coarse-gain & fine-grain parallelism in the next generation Pixel-planes graphic sys." by Fuchs et al. and published in Parallel Processing for Computer Vision and Display, pp. 241-253.
"Pixel Planes: A VLSI-Oriented Design for 3-D Raster Graphics" by H. Fuchs et al. and publ. in the proc. of the 7th Canadian Man-Computer Comm. Conference, pp. 343-347.
"The Torus Routing Chip" published in Journal of Distributed Computing, vol. 1, No. 3, 1986, and written by W. Dally et al. pp. 1-17.
"A Microprocessor-based Hypercube Supercomputer" written by J. Hayes et al. and published in IEEE MICRO in Oct. 1986, pp. 6-17.
"ILLIAC IV Software and Application Programming" written by David J. Kuck and published in IEEE Transactions on Computers, vol. C-17, No. 8, Aug. 1968, pp. 758-770.
"An Introduction to the ILLIAC IV Computer" written by D. McIntyre and published in Datamation, Apr. 1970, pp. 60-67.
"The ILLIAC IV Computer" written by G. Barnes et al. and published in IEEE Transactions on Computers, vol. C-17, No. 8, Aug. 1968, pp. 746-757.
The ILLIAC IV The First Supercomputer written by R. Michael Hord and publishing by Computer Science Press, pp. 1-69.
MC68000 8-/16-/32- Bit Microprocessor User's Manual, Eighth Edition, pp. 4-1 through 4-4; 4-8 through 4-12.
MC68020 32-Bit Microprocessor User's Manual, Fourth Edition, pp. 3-12 through 3-23.
Introduction to Computer Architecture written by Harold S. Stone et al. and published by Science Research Associates, Inc. in 1975, pp. 326 through 355.
"A VLSI Architecture for High-Performance, Low-Cost, On-chip Learning" by D. Hammerstrom for Adaptive Solutions, Inc., Feb. 28, 1990, pp. II-537 through II-544.
"CNAPS-1064 Preliminary Data CNAPS-1064 Digital Neural Processor" published by Adaptive Solutions, Inc. pp. 1-8.
DSP56000/DSP56001 Digital Signal Processor User's Manual, Rev. 1, published by Motorola, Inc. pp. 2-9 through 2-14, 5-1 through 5-21, 7-8 through 7-18.
"M-Structures: Ext. a Parallel, Non-strict, Functional Lang. with State" by Barth et al., Comp. Struct. Group Memo 327 (MIT), Mar. 18, 1991, pp. 1-21.
"A Pipelined, Shared Resource MIMD Computer" by B. Smith et al. and published in the Proceedings of the 1978 International Conference on Parellel Processing, pp. 6-8.
M68000 Family Programmer's Reference Manual published by Motorola, Inc. in 1989, pp. 2-71 through 2-78.
"The DSP is being reconfigured" by Chappell Brown and published in Electronic Engineering Times, Monday, Mar. 22, 1993, Issue 738, p. 29.
DSP56000/56001 Digital Signal Processor User's Manual published by Motorola, Inc. pp. 2-4 and 2-5, 4-6 and 4-7.
MC68340 Integrated Processor User's Manual published by Motorola, Inc. in 1990, pp. 6-1 through 6-22.
Transputer Architecture Technical Overview published by INMOS in Sep. 1985.
Product Description of the IMS T414 Transputer published by INMOS in Sep. 1985.
Product Description of the IMS T212 Transputer published by INMOS in Sep. 1985.
Proceedings from the INMOS Transputer Seminar tour conducted in 1986, published in Apr. 1986.
"Control Data STAR-100 Processor Design" written by R. G. Hintz et al. and published in the Innovative Architecture Digest of Papers for COMPCOM 72 in 1972, pp. 1 through 4.

Abstract

Patent Drawings (120 sheets)