International Business Machines Corporation

Troy N Hicks

Steven Wayne White

John Stephen Muhich

Larry Edward Thatcher

Richard E Fry

Hung Qui Le

Myhong Nguyenphu

A method and system in a data processing system for accessing information using an instruction specifying a memory address is disclosed. The method and system comprises issuing the instruction to an execution unit and storing an address derived from the specified address. The method and system also includes accessing a cache to obtain the information, using the derived address and determining, in response to a signal indicating that there has been a cache miss, if there is a location available to store the specified address in a queue. According to the system and method disclosed herein, the present invention allows for dynamic pipeline expansion of a processor without splitting this function between components depending upon the reason expansion was required, thereby increasing overall system performance.

Dynamic expansion of execution pipeline stages

The present invention is directed to a method and apparatus for dispatching instructions in an information handling system. A pre-execution queue stores instructions, and at least one execution cluster is operably coupled to the pre-execution queue. An execution cluster comprises an early execution unit for executing a first instruction dispatched from the pre-execution queue to generate and forward a first result and a late execution unit for executing a second instruction dispatched from the pre-execution queue to generate and forward a second result after the first execution unit forwards the first result. The invention further includes circuitry operably associated with the pre-execution queue, and a method for prioritizing the order in which the instructions in the pre-execution queue are dispatched to the execution cluster.

Instruction dispatch unit and method for dynamically classifying and

A system that allows the continuous accessing of data on a floating point processor unit (FPU), by providing two data ports and corresponding buses between the data cache and the FPU. Further, synchronization between the fixed point unit (FXU), which provides the addresses, and the FPU is provided so that data can be loaded in the event of a data cache miss. This synchronization allows data to be transferred from the DCU to the FPU independent of an error condition (cache miss) on one of the buses. If a cache miss occurs that affects a first one of the buses, then the instruction corresponding to this data is held. Subsequent floating point data is received by the FPU on the second bus not subject to the miss. Synchronization signals include, load ready (LD1.sub.-- RDY) indicating to the FPU that data is on the bus and ready to be moved to the FPU and load not ready (LD1.sub.-- NRDY) that means a floating point load has executed successfully in the FXU, but due to a cache miss, the data is not yet available to the FPU. The FXU sends an address, along an address bus corresponding to the data bus not affected by the miss, for data in the cache. Thus, even in the case of a data cache miss, the FPU can continue operations by loading data, that has been addressed to the data port which does not correspond to the cache miss.

Synchronization for out of order floating point data loads

A data processing system is provided that includes a plurality of execution units each including independent circuits for storing and executing instructions. A circuit is also included for providing instructions from a sequence of instructions to the execution units where each instruction is provided to only one of the execution units. The system includes a circuit for detecting when an instruction in a first execution unit must complete execution prior to execution of an instruction in a second execution unit to produce correct results. A circuit is further included, responsive to the circuit for detecting, for delaying executing the instruction in the second execution unit until the instruction in the first execution unit has completed execution.

Data processing system with multiple execution units capable of

A data processing system having an instruction execution circuit that executes a first type of instruction. Also included is a fetch circuit that fetches instructions from a memory and fetches data from the memory in response to a second type of instruction. An instruction decoder is included that decodes fetched instructions and dispatches instructions of the first type to an instruction queueing circuit. The instruction decoder further dispatches instructions of the second type to the fetching circuit. The instruction queueing circuit includes the capability to store decoded instructions of the first type while tagging these instructions when data required for the execution of these instructions has not been fetched. The instruction queueing circuit further clears these tags of these instructions of the first type when data that is required for the execution has been fetched. The instruction queueing circuit serially provides the untagged instructions of the first type to the instruction execution circuit.

Growing community of inventors

Austin, TX, United States of America

Troy N Hicks