Abstract:
In a heterogenous symmetric multi-processing system, processors from distinct families of processors are integrated on a single platform. The processors are coupled to an implementation specific communication mechanism through family specific bus interface converters. Shared memory and I/O subsystems may be coupled to the implementation specific communication mechanism as well. An operating system maintains separate ready queues for each family of processors. Each ready queue is responsible for scheduling execution of process threads on its associated family of processors. The operating system facilitates execution of both single mode binary code files and mixed mode binary code files. When a thread is created, the operating system determines the initial processor family to associate with the thread based on the binary code stream that the thread will begin executing. The thread is placed in the ready queue of that family. As the thread executes it may require services from another family of processors in order to natively execute the next set of instructions in the binary code file. When services are required, the operating system reschedules those instructions on a processor which executes those instructions natively. Means are provided to return the thread to a processor in the previous family of processors in order to support mixed mode instruction stream subroutine support
Abstract:
A performance regulator program (22) monitors and controls in real time, the performance level which an application program (21-i) achieves when it is executed on a digital computer (10). With this performance regulator program, any external units (11, 12) which are coupled to the computer are prevented from being overloaded by excessive performance in the application program. Also with this performance regulator program, several models of the application program (21-i) can be easily generated such that each model achieves a different performance level.
Abstract:
Disclosed is a method for use in a distributed computing enterprise to provide a client or user with access to a requested object located outside the address space of the user. The disclosed method includes: receiving a request for access to a requested object from a user having a first address space; if the requested object exists and is not active, recreating the requested object in the first address space of the user; and, if the requested object is active, performing one of the following steps. If the requested object is in the first address space, providing to the user a pointer that indicates the address of the requested object in the first address space. If the requested object is not currently in the first address space, effecting a transition of the requested object into an inactive state, and then recreating the requested object in the first address space.
Abstract:
The present invention operates in a file server (11) having a peripheral storage subsystem (16) coupled thereto by means of a fibre channel (14). The file server (11) includes an apparatus (12) disposed between the file server (11) and the storage subsystem (16) for adapting fibre channel transmissions to and from an industry standard data (29) bus of the file server (11).
Abstract:
A data processing array (11) is partitioned by electronic control signals into multiple sub-arrays (40-46 of Fig. 2A) which are established and operate independently of each other. In the preferred embodiment, an operator's console (30) is provided for manually selecting the data processing nodes that are in each sub-array (40-46 of fig. 2A)), and a control module (20) is coupled by control channels (31, 32, 33) between the console and the data processing nodes (12). These control channels carry the control signals directly to the data processing nodes (12) without utilizing the input/output channels (from message routing circuits 10) which are intercoupled to form the array (11). One portion of these control signals prevents each node in a sub-array from sending messages on the input/output channels to any node in another sub-array; and another portion of the control signals selects a node in each sub-array as a boot node from which a separate operating system and user programs are loaded without utilizing the input/output channels.
Abstract:
An electronic data transmission system has a low peak-to-average power ratio by including a transmitter circuit which receives an input signal and in response generates a distorted output signal. This distorted output signal is generated by amplifying the input signal with one particular gain when the input signal is at a maximum magnitude which gives the distorted output signal a corresponding maximum magnitude, and by amplifying the input signal with a larger gain when the input signal is in a predetermined range below the maximum magnitude. The distorted output signal travels over a communication channel to a receiver circuit, which regenerates the input signal by amplifying the distorted output signal with a gain that is the inverse of the gain by which the distorted signal is generated.
Abstract:
A bridge circuit (10) includes a microprocessor (20) having a first I/O port which couples to a SCSI bus (11) and a second I/O port which is coupled through transceivers (21) to an EISA bus (12). Also, the bridge circuit includes an EISA interface controller (22), having control lines (32) coupled to the EISA bus and the transceivers, which enable the microprocessor to request and use the EISA bus in time-shared fashion. In order to achieve a high speed of operation, the bridge circuit further includes a memory module (23), coupled via a private bus (24) to the second I/O port of the microprocessor (20), which sends instructions on the private bus directly to the microprocessor, without generating any signals on the EISA bus. In addition, in order to prevent deadlocks on the private bus, the bridge circuit includes a deadlock prevention circuit (25) which is coupled to the microprocessor (20) and the private bus (24) and the EISA interface controller (22). This deadlock prevention circuit detects the occurrence of a predetermined event (figs. 5, 6, 7) during a series of data transmissions between the microprocessor and the EISA bus. When such detection occurs, the deadlock prevention circuit directs the microprocessor to stop the series data transmission over the EISA bus and private bus before the end of the series; and it directs the microprocessor to not restart the series on the private bus until after the EISA bus is required. Meanwhile, instruction fetches on the private bus do occur.
Abstract:
An electronic control module (25 of Figs. 1 and 2) reduces ringing in the digital signals on a transmission line (26, 27, 28) which has multiple transmitters (22, 23), receivers (24), and parasitic inductors (L) and capacitors (C) coupled to the line. Preferably, a copy of this control module (25) is provided at each node on the line which has a transmitter and/or receiver. Each copy of the control module has three main parts - a sensing circuit (25a), a pulse generating circuit (25b), and a switching circuit (25c). The sensing circuit (25a) is coupled to the transmission line, and it generates a control signal (CTL) when the digital signal on the transmission line changes from a low voltage to a high voltage. The pulse generating circuit (25b) receives the control signal and responds by generating a single pulse (P1). The switching circuit (25c) receives the pulse and, in response, couples a high supply voltage (+V') to the transmission line while the pulse occurs.
Abstract:
A fault tolerant three port (P1, P2, P3) communications module (CM1 of fig. 1) has two control ports (P1 and P2) for receiving commands from two computers (A and B), and a communications port (P3) for transferring data over a communications channel (CC1) in response to the commands. Each control port includes a select line (e.g. SEL1A) which carries a select signal with true and false states, mode lines (e.g. Mode (0-3)A) which carry codes that represent the commands, and a write line (e.g. WRA) which carries a rspective pulse in sync with each of the codes. The select line, mode lines, and write line of each control port are coupled in the module to a respective inter-processor command decoder (20 of fig. 3) having a lead stage (22) and a trail stage (23). The lead stage detects (22) when a predetermined code (e.g. 1001) occurs on the mode lines and the select signal is false during a first one of the pulses; and, the trail stage (23) generates an output signal (IPC) that indicates the receipt of an inter-processor command for the module if, during a second pulse that immediately folows the first pulse, the compliment (0110) of the predetermined code occurs on the mode lines and the select line is true.
Abstract:
In a check processing array, an imaging/illumination arrangement for illuminating and imaging checks at one or several imaging-sites as they are rapidly, continuously transported past two or more imaging stations, each station having, as its illumination source, a hollow Lambertian integrating vessel, housing lamps which project a highly uniform, diffuse Lambertian illumination-beam to its respective imaging site.