International Business Machines Corporation

Other

Arnold Blum

Fritz Irro

Manfred Schmidt

Hermann Schulze-Schoelling

Johann Hajdu

Stephan R Richter

Manfred Perske

Hellmuth R Geng

Guenther Sonntag

Horst Von Der Heyden

Gottfried Andreas Goldrian

Marian Briska

Helmut Schettler

Knut K Najmann

Udo Wille

Guenter Knauft

Herbert Chilinski

Dieter Bazlen

Dietrich W Bock

Siegfried Neuber

Helmut Schaal

Claus Erich Mohr

Leopold Reichl

Frank Gerth

Wolfgang Kumpf

Rolf Berger

Bernd Spaeth

The invention concerns the transfer of data information in a multiprocessor computer system. If data information has to be transferred between two processor units then the associated control information is made available on a connection bus and the data information is transferred afterwards via a switch unit from the first to the second processor unit. If however data information is to be transferred from a sending processor unit to all other processor units (broadcast transfer) then not only is control information transferred via the connection bus but subsequent data information too; a transfer of data information via the switch unit does not occur in this case. In this manner it is possible to reduce the outlay for the switch unit in terms of circuitry and programming.

Routing control information via a bus selectively controls whether data

In a low temperature generation process, compressed gas (3) at high pressure enters from an inlet nozzle (1) and alternately expands into one of two resonator tubes (A,B) of an expansion engine. The expanding gas excites standing acoustic waves (4) in the resonator tubes (A,B). The acoustic energy of the waves (4) is converted into electrical energy by acoustic/electric power converters and is led away outside the 'cold area' of the tubes. The expansion engine contains one or more resonator tubes with a common inlet nozzle (1). Each resonator tube has an acoustic/electric power converter (5) and an exhaust port (6). Together with a compressor, a heat exchanger and a heat sink, the expansion engine provides a very effective cooling system which may be used for the cooling of small electronic devices like chips or modules.

Low temperature generation process and expansion engine

A system for mounting VLSI devices on a substrate is disclosed, offering a high contact density. Each package consists of a semiconductor device having protruding elongated contact pin (2) on its surface and a wiring substrate having a cavity (3) on its surface. The cavities are filed with a conductive material (7) of a low melting point composition and sealed with a thin non-conductive foil (4). During packaging, the contact pins are made to penetrate the foil, and extend into the conductive alloy, thus making electrical contact therewith. To ease the penetration of the foil, the contact pins could be set into oscillating motion by means of an ultrasonic generator.

Removable VLSI assembly

The compressor cascade comprises a plurality of tandem-connected membrane pumps, each of the pumps having a plurality of stroke chambers whose volumes decrease in the direction of the fluid flow through the pumps. Each chamber has several parallel-connected input/output channels for interconnecting the individual membrane pumps and a check valve in each input/output channel for forcing the fluid in a specified direction. By electrostatic attraction forces, the membranes in the pumps are energized synchronously to resonance oscillations of the same frequency and deflection, building up the necessary operating pressure as the fluid is moved from the stroke chamber of one membrane pump into the smaller volume stroke chamber of the next succeeding membrane pump. The movement of the fluid through the membrane pumps of the compressor cascade leads to its compression, and the pressure at the end of the cascade is related to the reduction in volume of each succeeding stroke chamber.

Cascade compressor

An integrated (silicon based) packaging/wiring structure provides for VLSI chips 4 to be placed within openings of somewhat larger size in a semiconductor interconnection wafer (IW, 2) supported by a carrier 1. The interconnection wafer 2 includes multilevel (ML) wiring planes and incorporated circuit components integrated in a less demanding technology as compared to the VLSI chips 4. Silicon contact chips 5 with conductive surface layers 22, 23 placed over the chip/IW plane provide for the required interconnections by means of needle-like structures 24 inserted in corresponding via holes. The needles are better suited to withstand shear strain than are conventional C-4 (Controlled Collapse Chip Connection) joints. Consequently a much higher number of chip pads can be provided. Power supply is effected via rather large-dimensioned conductive planes, e.g. in the form of Cu rails 20, running within the carrier 1 and surfacing stud-like (at 21) in the peripheral region of said openings in the interconnection wafer 2 for further distribution via the contact chip 5. The wiring system can be supplemented, if required, with an additional wiring wafer 6.

Integrated wiring system for VLSI

The invention concerns arrangements and methods for error testing and diagnosing processors (e.g., 9; FIG. 2), whose logic subsystems (20) are interconnected by storage elements (23, 24) which in the error test and diagnostic mode are connected in the form of shift register means for the shift clock controlled application of test data and for receiving result data, and which comprise means (58) for comparing the actual result data with desired result data, said means setting an error indicator (59) for initiating further actions in the case of a mismatch. For testing the correct implementation of operations and operational secondary functions, a signature generator circuit (30) is provided comprising a test accumulator (51, 52, . . . , 5m) for accumulating the test and result data from the storage elements (23, 24) and a test clock generator and counter (28) for controlling the accumulation, as well as a test memory (29) providing test programs consisting of test data, desired result data and instructions to be tested of the processor instruction set. The signature generator circuit ( 30) is connected to an interface register (11, 12, . . . , 1m) and/or a system bus (8) of the processor, the stages of the interface register being included in the shift register means consisting of the storage elements (23, 24) at positions 21, 22, . . . , nm.

Built-in parallel testing circuit for use in a processor

In a bus-oriented computer system, the decision as to which unit is to receive the bus takes account of the current status of the bus system and the respective operation to be performed on the bus. For that purpose, status information of the connected units, the bus command to be executed and the address of the requested unit are fed to the allocation logic (arbiter) on separate or commonly used lines, thus avoiding idle times during the use of the bus. By evaluating the bus command, the allocation priority can be dynamically changes in order to suppress bus accesses that are bound to fail from the start.

Method and apparatus for bus arbitration in a data processing system

The detailed testing of processors, manufactured according to very large scale integration principles, which also extends to secondary functions of operations, such as the setting or non-setting of particular state indicators, necessitates the transfer of large quantities of test data between the processor and the tester, for which purpose no distinction is drawn between external and integrated testers. For testing such structures, the known LSSD method is frequently used wherein the storage elements of the logic subsystems are combined in the form of shift register chains for testing. To permit a fast exchange of test data on the system bus, connecting the processor to a tester, interface register stages are also included in the shift register chain which has a garland-shaped structure and whose beginning and end are connected by a controllable switch during testing. During testing, the entry of test data and the emission of result data, which are in each case effected through the system bus and the shift steps of the garland-shaped shift register chain, overlap one another.

Testing and diagnostic device for digital computers

For error testing and diagnostics in EDP systems, particular storage elements (e.g., 6) are connected to form an addressable matrix which is coupled to a maintenance and service processor (5) or an external tester through the system bus (9). During normal operation, the logic subsystems (10), of which processors (1) and processing units consist, are connected by the storage elements. Through the system bus, the maintenance and service processor or the tester transfers addresses to the matrix and test data to the addressed storage elements from where they are fed to the logic subsystems which in turn respond to such test data, transferring the (partial) result data thus received to the storage elements. In the next step, the maintenance and service processor or the external tester causes the result data for error analysis and diagnostics to be fetched through the system bus from the storage elements reconnected in the form of a matrix.

An integrated circuit chip includes a tristate driver which assumes an active logical state in response to a data signal at its data input and assumes a high impedance state in response to a control signal at its control input. The integrated circuit chip also includes a control signal generating network which is connected to the tristate driver's control input for producing the control signal. The control signal generating network may be tested by connecting the control signal generating network to the data input and overriding the control input to prevent the tristate driver from assuming the high impedance state. Thus, for testing purposes, the proper response of the control signal generating circuit may be ascertained by monitoring the active state of the tristate driver.

Testing an integrated circuit containing a tristate driver and a control

LSI circuitry conforming to LSSD rules and techniques usually requires at least a small portion of circuitry used only for check and test purposes. The disclosed circuitry meets the LSSD design rules and techniques and considerably reduces the test circuit overhead. The disclosure modifies the known shift register latch (SRL) strategy by logically removing the master latches from the slave latches and by providing the slave latches with multiple shift inputs, e.g., two shift inputs (FIG. 2). The LSSD shifting philosophy remains unchanged to the extent that at the time of shifting, the virtual (not available slave latch) becomes real (physical) by assigning the only physical slave latch to the respective master latch. The present disclosure provides for multiple master latches to be dynamically assigned to one slave latch during shifting. This is in contrast to the known SRL chain approach requiring one slave latch for each master latch. Level Sensitive Scan Design Rules and Techniques are extensively disclosed in the testing art. See for example (1) U.S. Pat. No. 3,783,254 entitled 'Level Sensitive Logic System' filed Oct. 16, 1972, granted Jan. 1, 1974 to E. B. Eichelberger, of common assignee herewith, or; (2) 'A Logic Design Structure for LSI Testability' by E. B. Eichelberger and T. W. Williams, 14th Design Automation Conference Proceedings, pages 462-468, June 20, 21 and 22, 1977, New Orleans, Louisiana, IEEE Catalog Number 77, CH1216-1C.

Shift register latch circuit means contained in LSI circuitry conforming

LSI circuitry conforming to LSSD rules and techniques usually requires at least a small portion of circuitry used only for check and test purposes. The disclosed circuitry meets the LSSD design rules and techniques and considerably reduces the test circuit overhead. The disclosure modifies the known shift register latch (SRL) strategy by replacing the SRL's by master latches in such a manner that the information contained in them is shifted in cascades, using the 'division by two' principle for the master latches on the chip. The shift chain having only master latches is selected in response to shift clock signals. By consecutively shifting the respective cascade element, detailed information is obtained for all the master latches on the chip (without the information of the master latches temporarily used as slave latches during shifting) being lost in the cascade element. Level Sensitive Scan Design Rules and Techniques are extensively disclosed in the testing art. See for example: (1) U.S. Pat. No. 3,783,254 entitled 'Level Sensitive Logic System' filed Oct. 16, 1972, granted Jan. 1, 1974 to E. B. Eichelberger, of common assignee herewith; or (2) 'A Logic Design Structure for LSI Testability' by E. B. Eichelberger and T. W. Williams, 14th Design Ahtomation Conference Proceedings, pp. 462-468, June 20, 21 and 22, 1977, New Orleans, Louisiana, IEEE Catalog Number 77, CH1216-1C.

Shift register latch circuit means for check and test purposes and

In a microprogrammed processor consisting of several circuitized chips, which are to be synchronously operated, each chip is provided with its own local clock generator or T-ring for deriving therefrom timing signals required during the subphases of micro instruction execution. A master clock connected to all of the T-rings by lines of equal length forces the individual T-rings to operate synchronously and keeps them operating in such a manner. In addition, reset circuitry is provided for forcing all of the T-rings to their first timing interval for initial synchronization thereof or, at an appropriate time, when a micro instruction requiring less than the maximum number of available T-ring timing signals is executed. The timing signals which are locally produced are subject to little delay on their way to the various local switching points. Thus, the entire system can be operated at a higher oscillator or master clock frequency to take advantage of the enhanced processing and transfer speeds of modern, highly integrated circuit chips.

Decentralized generation of synchronized clock control signals having

In a data processing or transmission system which includes at least two synchronized clocks, for example--T-rings A and B which generate timing pulses Tai and Tbi for microinstruction execution, synchronism is checked by logic circuitry which receives pulses from the clocks. At least one of the pulses is delayed by one or more pulse period durations ti. The logic circuit output signal is used as an input to an indicator latch which is periodically set by an independent check oscillator or clock. In a preferred embodiment, the delays are introduced by master-slave flip-flops, which receive predetermined combinations of the T-signals and set by the independent check clock. Several delay latches and associated AND gates may be used for different logical combinations of delayed and undelayed T-signals. This scheme can easily be expanded to accommodate more than two synchronously operating clocks. These circuits check not only the instantaneous synchronism of the clocks, but also the correct sequencing of clock pulses. The check is also feasible if the T-ring counters are operated with a variable number of clock pulses per microinstruction execution.

Clock check circuits using delayed signals

In a micro-controlled data handling system the number of lines and pins required to transfer control signals from the microprogram controls to be integrated circuit modules controlled by such signals is conserved by using two bussing paths for distributing the control signals to the modules. A first path is dedicated exclusively to pre-decoded control signal functions and a second path is shared for transferring both data and control signal functions. Each controlled module contains an additional decoding circuit for combinationally decoding control signal functions received through both paths.

Growing community of inventors

Gechingen, Germany

Arnold Blum