摘要:
In a data processing system having a plurality of tightly coupled data processing units connected by an asynchronous system bus, apparatus and an associated method are described for maintaining the coherency of data groups stored in instruction cache units and execution cache units. The apparatus includes a monitor unit as part of the bus interface unit, and a bus interface unit coupling each associated data processing unit to the system bus. The monitor unit receives signals, applied to the system bus, identifying data groups transferred between the memory unit and the data processing units, including those data groups originating from the bus interface unit of which the monitor unit is a component. The bus interface unit includes directories duplicating the contents of the instruction cache unit directory and the execution cache unit directory. The monitor unit, in response to signals applied to the system bus, identifies operations that can compromise the integrity of signals stored in the associated data processing unit. The monitor unit accesses the appropriate duplicate directory to determine if the address of a compromised data group validly exists in the duplicate directory. When an address is stored in the duplicate directory along with a valid signal, the valid signal is removed from both the cache directory and the duplicate directory.
摘要:
The pipelined central processing system (CSS) units of a multiprocessor system are tightly coupled to connect in common to a system bus for sharing main memory and input/output controllers/devices. The CSS includes several circuit boards for the different VLSI circuit chip pipelined stages and associated control circuits in addition to the bus interface unit (BIU) circuits. Each board includes one or more unusual event (UEV) detector circuits for signaling when the behavior of a stage is abnormal. The UEV fault signals from each board are collected by the BIU board. When a UEV fault is detected, the BIU board circuits prevent any further communications with the system bus and broadcasts the UEV fault signal to the other boards causing the different pipelined stages to emulate the completion of the instructions within the pipeline thereby flushing it. It is thereafter placed in a nonpipelined mode. Control circuits execute a UEV handler microcode routine which reads the contents of syndrome registers included on each board containing the UEV indicator states in addition to error signals into register file working locations. The UEV signal is then cleared enabling the BIU to resume communications with the sytem bus.
摘要:
A symmetric multiprocessing system with a unified environment and distributed system functions provides unified address space for all functional units in the system while distributing the execution of various system functions over the functional units of the system whereby each functional unit assumes responsibility for its own aspects of these operations. In addition, the system provides improved system bus operation for transfer of data from memory.
摘要:
A multiprocessor system includes a number of system processors which tightly couple to a system bus to share a main or system memory and a number of on-board memory processors which also are tightly coupled to the system bus. Each processor has a high performance microprocessor which tightly couples to an on-board or local memory through the microprocessor's local bus. System memory is accessible using a memory lock protocol while the local memory is accessible through a bus lock protocol. Each on-board memory processor includes a lock mechanism which enables the processing of memory lock commands directed to its local memory received via the system bus from any other processor and for issuing memory lock commands to system memory.
摘要:
A data processing system includes the functionality of a commercial instruction processor, a scientific instruction processor and a basic instruction processor integrated into a single semiconductor logic element.
摘要:
If the firmware calls for an operand rounding operation, apparatus in the Scientific Instruction Processor (SIP) tests the bit to the right of the low order bit of the normalized operand to determine if a rounding cycle is required. If the operand requires a normalization cycle or a mantissa overflow correction cycle, the rounding operation is performed in those cycles.
摘要:
A data processing system includes a plurality of system units all connected in common to a system bus. The system units include a central processor (CPU), a memory system and a high speed buffer or cache system. The cache system is word oriented and comprises a directory, a data buffer and associated control logic. The CPU requests data words by sending a main memory address of the requested data word to the cache system. If the cache does not have the information, apparatus in the cache requests the information from main memory, and in addition, the apparatus requests additional information from consecutively higher addresses. If main memory is busy, the cache has apparatus to request fewer words.
摘要:
Timing calculator means in a computer system are used to adaptively generate an appropriate access signal, to one of a plurality of memory types, based on first and second timing control values, wherein the first timing control value represents information specific to and limited to the start of a memory operation and wherein the second timing control value represents information representing other timing events. That is, the state machine of the present invention requires a distinct starting control value, separate from other timing values, for calculation of appropriate memory access parameters.
摘要:
A segment descriptor unit (SDU) includes a divided random access memory (RAM), a content addressable memory (CAM) and decoder circuits interconnected for performing dynamic and static address translation operations within a minimum of chip area and power. The CAM is arranged to store a number of entries which include segment number and validity information associated with a corresponding number of segment descriptors. The RAM contains locations allocated for storing segment descriptor words (SDW's) and working data. Each SDW is logically divided into two fields, a static translation word (STW) field containing all of the bits required for performing a static address translation operation and an access control word (ACW) field containing all of the bits required for verifying compliance with system security. The bits of each STW and ACW are stored in alternate bit positions of the SDW locations. Each pair of RAM bit locations couple to a common read/write amplifier and multiplexer circuit. Through the use of microinstruction commands coded to specify different address translation functions, the STW and ACW fields selected by the CAM are read out from RAM during different intervals for carrying out the steps of those operations.
摘要:
A translator is organized to include at least a pair of content addressable memories (CAMs), each for storing a different portion of the total number of bits of each of the words to be translated. The outputs from each CAM are logically combined within a multiple input random access memory (RAM). Both CAMs are interrogated simultaneously and deliver the results of comparing the word portions of an input word and the CAM contents to the RAM in substantially less time then required for a single CAM memory. The results are logically combined with in the RAM which, in response to a match condition, delivers the results of the translation as an output.