Digital Equipment Corporation

Michael A Callander

Douglas E Sanders

Linda Chao

Richard Lee Sites

Derrick R Meyer

Nicholas D Wade

Raymond L Strouble

G Michael Uhler

Rebecca L Stamm

W Hugh Durdan

R Iris Bahar

Multi-processor systems are often implemented using a common system bus as the communication mechanism between CPU, memory, and I/O adapters. It is also common to include features on each CPU module, such as cache memory, that enhance the performance of the execution of instructions in the CPU. Many architectures require that the hardware employ a mechanism by which the data in the individual CPU cache memories is kept consistent with data in main memory and with data in other cache memories. One such method involves each CPU monitoring transactions on the system bus, and taking appropriate action when a transaction appears on the bus which would render data in the CPU's cache incoherent. If the CPU uses queues to hold records of incoming transaction information until it can service them, the bus interface must guarantee that the queued items are processed by the cache in the correct order. If this is not done, certain types of shared data protocols fail to operate correctly. The present invention describes a method by which hardware can guarantee the serialization of transactions requiring service by the CPU cache. The serialization method described guarantees that shared memory protocols operate correctly.

Multi-processor computer system having shared memory, private cache

A CPU module has a processor, cache memory, cache controller, and system interface attached to a processor bus. The system interface is attached to a system bus shared by memory, I/O, and other CPU modules. The cache controller requests control of the processor bus from the processor, and grants control to the system interface. The system interface uses the processor bus to store fill data obtained from memory into the cache in response to a read miss. The system interface also monitors system bus traffic and forwards the addresses of cache blocks to be invalidated to the cache controller over an invalidate bus. The cache controller requests control of the processor bus during a read miss to perform invalidates and writebacks. The processor grants control to the cache controller before the read miss completes, enabling the cache controller to proceed, and then re-issues the read. A protocol between the cache controller and the system interface ensures that cache fills, invalidates, and writebacks are done in the correct order to maintain data coherency. As part of this protocol, the cache controller decides when the system interface may proceed with a fill, and grants the processor bus to the system interface accordingly.

Method and apparatus for controlling a processor bus used by multiple

This invention relates to a write-back cache which is protected with parity and error correction coding ('ECC'). The parity and ECC codes are generated by a memory interface when data is transferred by main memory to the central processing unit ('CPU') associated with the cache. Thus, all data originating in main memory will be parity and ECC encoded when it passes through the memory interface, and the data, and its related parity information and ECC codes will be stored in the cache. On the other hand, data which is taken from the cache and modified by the CPU during its processing operations is also transferred to the memory interface for ECC encoding. Thus, all data modified by the CPU is also protected, and the modified data, and its related parity information and ECC codes are also stored in the cache. The memory interface also contains ECC checking and correcting circuitry which can correct erroneous data, on the basis of its ECC code, if that data has been corrupted while stored in the cache, or during transmission on a bus. Therefore, if data in the cache is corrupted, it can be corrected when it is returned to main memory via the memory interface. Accordingly, the invention makes a write-back cache compatible with full ECC protection, even though, at times, the cache may contain the only correct, current copy of given data in the system.

Write-back cache with ECC protection

A write-back cache memory system is disclosed which includes a source of a sequence of memory addresses and a tag store coupled to the source of addresses and accessed by an index portion of said addresses, which produces information relating to the addresses. The write-back cache memory system also includes an error detector for detecting an error in the tag store information. Circuitry is included for reporting an error and saving the index which caused the error if an error is detected but no error has been previously detected. Comparing circuitry is included for comparing the index causing the current error to the previously saved address if an error is detected and an error has been previously detected; and if the address is not the same, then reporting a fatal error; otherwise, if the index is the same, then not reporting a fatal error.

Apparatus for suppressing an error report from an address for which an

A method and apparatus for monitoring transactions on a system bus for invalidate requests, including a queue for storing the invalidate requests which is divided into two parts. The first part of the queue is contained within a cache controller to ensure that an invalidate request is immediately available for processing when the cache controller is otherwise idle. The second part of the queue is contained within a system interface to ensure that the system interface can detect and respond to more system transactions before the first invalidate request has been processed and to enable the system interface to be immediately aware if the entire queue is full.

Two-level protocol for multi-component bus ownership, and implementation

A pipelined CPU executing instructions of variable length, and referencing memory using various data widths. Macroinstruction pipelining is employed (instead of microinstruction pipelining), with queueing between units of the CPU to allow flexibility in instruction execution times. A wide bandwidth is available for memory access; fetching 64-bit data blocks on each cycle. A hierarchical cache arrangement has an improved method of cache set selection, increasing the likelihood of a cache hit. A writeback cache is used (instead of writethrough) and writeback is allowed to proceed even though other accesses are suppressed due to queues being full. A branch prediction method employs a branch history table which records the taken vs. not-taken history of branch opcodes recently used, and uses an empirical algorithm to predict which way the next occurrence of this branch will go, based upon the history table. A floating point processor function is integrated on-chip, with enhanced speed due to a bypass technique; a trial mini-rounding is done on low-order bits of the result, and if correct, the last stage of the floating point processor can be bypassed, saving one cycle of latency. For CAL type instructions, a method for determining which registers need to be saved is executed in a minimum number of cycles, examining groups of register mask bits at one time. Internal processor registers are accessed with short (byte width) addresses instead of full physical addresses as used for memory and I/O references, but off-chip processor registers are memory-mapped and accessed by the same busses using the same controls as the memory and I/O. In a non-recoverable error detected by ECC circuits in the cache, an error transition mode is entered wherein the cache operates under limited access rules, allowing a maximum of access by the system for data blocks owned by the cache, but yet minimizing changes to the cache data so that diagnostics may be run. Separate queues are provided for the return data from memory and cache invalidates, yet the order or bus transactions is maintained by a pointer arrangement. The bus protocol used by the CPU to communicate with the system bus is of the pended type, with transactions on the bus identified by an ID field specifying the originator, and arbitration for bus grant goes one simultaneously with address/data transactions on the bus.

Growing community of inventors

Hudson, MA, United States of America

Michael A Callander