Unisys Corporation

Gary J Lucas

John Steven Kuslak

David C Johnson

Mitchell A Bauman

Donald W Mackenthun

Kenneth L Engelbrecht

Glen R Kregness

Lawrence R Fontaine

Nathan A Eckel

Ronald G Arnold

James L Federici

Thomas D Hartnett

Nguyen Thai Tran

Peter Levinshteyn

Michael David Pelarski

Ferris T Price Deceased

Kenichi Tsuchiva

The current invention provides a mechanism for detecting and recovering from glitches on data strobes. In one embodiment, data is captured from an interlace by a receiver using at least one data strobe that is provided by the transmitter along with the data. A write address counter that is clocked by the data strobe is used to count the active edges of the data strobe. A read address counter that is periodically synchronized with the write address counter, but that is clocked by an internal clock of the receiver, is used to count units of data being received from the interface. Periodically, the contents of the read and write counters are compared. If the contents are not the same, a glitch has occurred on the data strobe. The glitch is recoverable if it occurs on, or after, a last strobe edge of a data transfer.

System and method for detecting glitches on a high-speed interface

A branch prediction method and system are provided that accurately predict a branch condition early in an instruction pipeline of a data processing system. By accurately predicting the branch condition, the correct target instruction can be fetched early, thereby avoiding many of the inefficiencies associated with branch mispredictions. To accurately predict if a branch condition is satisfied, one or more pre-calculated status bits are stored along with a digital value that is read by the conditional branch instruction to determine if the branch condition is satisfied. By including such a status bit, the condition of the conditional branch instruction may be immediately determined, without waiting for the instruction to be processed by an arithmetic unit or the like in a subsequent pipeline stage.

Condition indicator for use by a conditional branch instruction

A synchronous pipeline design is provided that includes a first predetermined number of fetch logic sections, or 'stages', and a second predetermined number of execution stages. Instructions are retrieved from memory and undergo instruction pre-decode and decode operations during the fetch stages of the pipeline. Thereafter, decoded instruction signals are passed to the execution stages of the pipeline, where the signals are dispatched to other execution logic sections to control operand address generation, operand retrieval, any arithmetic processing, and the storing of any generated results. Instructions advance within the various pipeline fetch stages in a manner that may be independent from the way instructions advance within the execution stages. Thus, in certain instances, instruction execution may stall such that the execution stages of the pipeline are not receiving additional instructions to process. This may occur, for example, because an operand required for instruction execution is unavailable. It may also occur for certain instructions that require additional processing cycles. Even though instructions are not entering the execution stages, instructions may continue to enter the fetch stages of the pipeline until all fetch stages are processing a respective instruction. As a result, when normal instruction execution resumes within the execution stages of the pipeline, all fetch stages of the pipeline have been filled, and pre-decode and decode operations have been completed for those instructions awaiting the entry into the execution stages of the pipeline.

Pipeline controller for providing independent execution between the preliminary and advanced stages of a synchronous pipeline

A method and apparatus for reducing processor response time to selected transfer instructions in an multi-instruction processor. The response time is shortened by using a fast path to generate addresses for selected transfer instructions. In this fast path a base address, retained in a register from a previous instruction, is summed with an offset from the current instruction to obtain an absolute address for memory accessing. Before the fast path is entered determinations are made whether the instruction is a particular transfer instruction of a particular class and subclass, and whether the base address is different than the base address for the previous instruction. Even through the fast path is entered the usual absolute address generator path is also entered where the instruction is subjected to both high and low limit tests. If the high and low limit test determine a different base is to be used, the absolute address from the main address generator is used, instead of the absolute address from the LXJ fast path, and the system is restored to the conditions that would have prevailed if the fast path had not been entered.

Method of and apparatus for saving time performing certain transfer

A method and apparatus for reducing processor response time to selected transfer instructions in an instruction processor using a plurality of memory banks including four banks in a basic mode and one memory bank in an extended mode. This invention provides fast transfer hardware to improve the response time by a speed up transfer for normal extended mode transfer instructions only. The bank descriptor of the instruction is used to determine appropriate transfer instructions which are then tested for characteristics indicating whether a fast transfer is possible. The fast transfer process requires fewer checks than the previous apparatus which accelerates the response to selected transfer instructions by one cycle.

Method of and apparatus for speeding up the execution of normal extended

Method and apparatus for detecting and correcting memory storage data errors in a system utilizing parity error detection. An error detected in the memory storage device results in a parity error being reported, thereby causing the corresponding address location to be deactivated. Once deactivated, no further reading or writing is performed at that address location for a predetermined time period. The parity error reporting and address deactivation is accomplished without an access time penalty and requires a reduced number of I/O pins.

Second level cache having instruction cache parity error control

A system and method for selectively injecting parity errors into instructions after the instructions are fetched from a storage device and are resident within the instruction processor in a data processing system. The parity errors are selectively injected according to programmable indicators, each programmable indicator being associated with one or more instructions stored in the storage device. The error-injection system also includes programmable operating modes whereby error injection will occur after every fetch of an associated instruction, or alternatively, after alternate fetches of an associated instruction. The system allows for comprehensive testing of error detection and recovery logic in an instruction processor, and further allows for comprehensive testing of the logic associated with performing a data re-fetch from the storage device.

Parity-error injection system for an instruction processor

Method and apparatus for changing the sequential execution of instructions in a pipelined instruction processor by using a microcode controlled redirect controller. The execution of a redirect instruction by the pipelined instruction processor provides a number of microcode bits including a target address to the redirect controller, a predetermined combination of the microcode bits then causes the redirect controller to redirect the execution sequence of the instructions from the next sequential instruction to a target instruction.

Instruction flow control for an instruction processor

A method of and apparatus for efficiently halting the operation of the instruction processor when a cache miss is detected. Generally, this is accomplished by preventing unwanted address incrementation of an instruction address pipeline and by providing a null instruction to an instruction pipeline when a cache miss is detected. Accordingly, the present invention may eliminate a recovery period after a cache miss, thereby enhance the performance of the data processing system. Further, the present invention may eliminate recovery hardware required to support the recovery process.

High performance instruction data path

Disclosed is a system for improved instruction fetch prediction. When a jump instruction is encountered, the preceding instruction is considered in predicting the next instruction to fetch. If the preceding instruction is a skip instruction, the result of evaluating a condition specified by the skip instruction is used in predicting the next instruction to fetch. Prediction designators for skip/jump sequences of instructions are maintained in a jump prediction RAM.

Apparatus and method for processing a jump instruction preceded by a

A method and apparatus for granting exclusive access to a selected portion of addressable memory to a requesting processor in a large scale multiprocessor system. An instruction processor having a store-through operand cache executes an instruction requiring exclusive access to an address in a shared memory. If the address upon which the lock is requested is not in the local cache, the instruction processor simultaneously sends a lock and read request to the coupled storage controller. Otherwise, a no-operand-read and lock request is sent to the storage controller. If, while processing the lock request, no conflict is detected by the storage controller, the address is marked as locked and a lock granted signal is issued to the requesting processor. Concurrent with the processing the lock request the storage controller processes the read request. The lock granted signal and requested data are returned to the requesting processor asynchronously. The requesting processor can continue processing the lock instruction when the lock granted and required data have been returned from the storage controller. When two or more processors contend for a lock on a the same portion of addressable memory, one processor is granted the lock while the other contending processor(s) are forced to wait. Lock contention is arbitrated by a round robin priority scheme.

Growing community of inventors

Pine Springs, MN, United States of America

Gary J Lucas