Hal Computer Systems, Inc.

Fujitsu Corporation

John R Szeto

Michael C Shebanow

Gene W Shen

Niteen Aravind Patkar

Michael A Simone

Hideki Osone

Deforest W Tovey

Takumi Maruyama

John Gmuender

A high-performance processor is disclosed with structure and methods for: (1) aggressively scheduling long latency instructions including load/store instructions while maintaining precise state; (2) maintaining and restoring state at any instruction boundary; (3) tracking instruction status; (4) checkpointing instructions; (5) creating, maintaining, and using a time-out checkpoint; (6) tracking floating-point exceptions; (7) creating, maintaining, and using a watchpoint for plural, simultaneous, unresolved-branch evaluation; and (9) increasing processor throughput while maintaining precise state. In one embodiment of the invention, a method of restoring machine state in a processor at any instruction boundary is disclosed. For any instruction which may modify control registers, the processor is either synchronized prior to execution or an instruction checkpoint is stored to preserve state; and for any instruction that creates a program counter discontinuity an instruction checkpoint is stored. Instruction execution status is monitored to detect a fault condition and if a fault occurs, the instruction identifier is saved according to predetermined rules and used as an endpoint to backup the program counter. If the faulting instruction is a checkpointed instruction, then the state information is restored from the checkpoint prior to reexecuting the faulting instruction, but if one of the instructions for which state was stored is sequentially between the faulting instruction and the last issued instruction, then (i) backing-up said processor to the closest checkpoint after the faulting instruction, (ii) back stepping the processor to restore state, and (iii) decrementing the program counter to the faulting instruction.

Structure and method for instruction boundary machine state restoration

Apparatus and method provide for tracking and maintaining precise state by assigning a unique identification tag to each instruction at the time of issue, associating the tag with a storage location in a first active instruction data structure, updating the data stored in the storage location in response to instruction activity status changes for each instruction, and maintaining a plurality of pointers to the storage locations that move in response to the instruction activity status. Status information includes an activity data item, such as an activity bit, that is set at the time the instruction is issued and cleared when execution completes without error. Pointers are established that point to the last issued instruction, the last committed instruction pointer, and reclaimed instruction pointer. These three pointers are moved forward toward the later issued (newer) instructions along the data structure based on comparisons of the active-bit for each location associated with one instruction in the data structure and predetermined rules. Exceptions or error conditions for any instruction prevent changing the active-bit so that movement of the pointers is controlled and prevented under these error conditions.

Processor structure and method for tracking instruction status to

An instruction selector receives M instructions per clock cycle and stores N instructions in an instruction queue memory. An instruction queue generates a precedence matrix indicative of the age of the N instructions. A dependency checker determines the available registers for executing the instructions ready for execution. An oldest-instruction selector selects the M oldest instructions responsive to the precedence matrix and the eligible queue entry signals. The instruction queue provides the M selected instructions to execution units for execution. Upon completing the instructions, the execution units provide register availability signals to the dependency checker to release the registers used for the instructions.

Method and apparatus for selecting the oldest queued instructions

An out of program control order execution data processor that comprises an issue unit, execution means, a floating point exception unit a precise state unit, a floating point status register, and writing means. The issue unit issues instructions in program control order for execution. The issued instructions include floating point instructions and non-floating point instructions. The execution means executes the issued instructions such that at least the floating point instructions may be executed out of program control order by the execution means. The floating point exception unit includes a data storage structure including storage elements. Each issued instruction corresponds to one of the storage elements. Each storage element has a floating point instruction identifying field and a floating point trap type field. The floating point exception unit also includes first logic to write, for each issued instruction, data in the floating point instruction identifying field of the corresponding storage element which indicates whether or not the corresponding issued instruction is a floating point instruction. It further includes second logic to write, for each issued floating point instruction which causes during execution one or more floating point execution exceptions that will result in a corresponding one of a plurality of predefined types of floating point execution traps, data in the floating point trap type field of the corresponding storage element which identifies the one of the predefined types of floating point execution traps that will result. The precise state means retires each issued instruction which does not cause an execution exception during execution and for which all issued instructions preceding it in program control order have been retired. When a first one of the predefined execution exceptions is caused by an issued instruction, the execution means continues execution of issued instructions and the precise state means engages in execution trap sequencing by continuing to retire issued instructions until it encounters an issued instruction that cannot be retired. The issued instruction that cannot be retired being one of (a) the issued instruction that caused the first execution exception, and (b) an issued instruction that was issued earlier than the issued instruction that caused the first execution exception but which caused a second execution exception occurring later than the first execution exception. The floating point status register has a floating point trap type field. The writing means writes data to the floating point trap type field of the floating point status register which identifies the type of floating point execution trap identified by the data in the floating point trap type field of the storage element corresponding to the instruction that cannot be retired when the data in the floating point identifying field of the storage element corresponding to the instruction that cannot be retired indicates that the instruction that cannot be retired is a floating point instruction.

Processor structure and method for tracking floating-point exceptions

Time-out checkpoints are formed based on a predetermined time-out condition or interval since the last checkpoint was formed rather than forming a checkpoint to store current processor state based merely on decoded instruction attributes. Such time-cut conditions may include the number of instructions issued or the number of clock cycles elapsed, for example. Time-out checkpointing limits the maximum number of instructions within a checkpoint boundary and bounds the time period for recovery from an exception condition. The processor can restore time-out based checkpointed state faster than an instruction decode based checkpoint technique in the event of an exception so long as the instruction window size is greater than the maximum number of instructions within a checkpoint boundary, and such method eliminates processor state restoration dependency on instruction window size. Time-out checkpoints may be implemented with conventional. checkpoints, or in a novel logical and physical register rename map checkpointing technique. Timeout checkpoint formation may be used with conventional processor backup techniques as well as with a novel backtracking technique including processor backup and backstepping.

Processor structure and method for checkpointing instructions to

Time-out checkpoints are formed based on a predetermined time-out condition or interval since the last checkpoint was formed rather than forming a checkpoint to store current processor state based merely on decoded instruction attributes. Such time-out conditions may include the number of instructions issued or the number of clock cycles elapsed, for example. Time-out checkpointing limits the maximum number of instructions within a checkpoint boundary and bounds the time period for recovery from an exception condition. The processor can restore time-out based checkpointed state faster than an instruction decode based checkpoint technique in the event of an exception so long as the instruction window size is greater than the maximum number of instructions within a checkpoint boundary, and such method eliminates processor state restoration dependency on instruction window size. Time-out checkpoints may be implemented with conventional checkpoints, or in a novel logical and physical register rename map checkpointing technique. Timeout checkpoint formation may be used with conventional processor backup techniques as well as with a novel backtracking technique including processor backup and backstepping.

Processor structure and method for aggressively scheduling long latency

A high-performance processor is disclosed with structure and methods for: (1) aggressively scheduling long latency instructions including load/store instructions while maintaining precise state; (2) maintaining and restoring precise state at any instruction boundary; (3) tracking instruction status to maintain precise state; (4) checkpointing instructions to maintain precise state; (5) creating, maintaining, and using a time-out checkpoint; (6) tracking floating-point exceptions; (7) creating, maintaining, and using a watchpoint for plural, simultaneous, unresolved-branch evaluation; and (9) increasing processor throughput while maintaining precise state.

Processor structure and method for maintaining and restoring precise

Time-out checkpoints are formed based on a predetermined time-out condition or interval since the last checkpoint was formed rather than forming a checkpoint to store current processor state based merely on decoded Instruction attributes. Such time-out conditions may include the number of instructions issued or the number of clock cycles elapsed, for example. Time-out checkpointing limits the maximum number of instructions within a checkpoint boundary and bounds the time period for recovery from an exception condition. The processor can restore time-out based checkpointed state faster than an instruction decode based checkpoint technique in the event of an exception so long as the instruction window size is greater than the maximum number of instructions within a checkpoint boundary, end such method eliminates processor state restoration dependency on instruction window size. Time-out checkpoints may be implemented with conventional checkpoints, or in a novel logical and physical register rename map checkpointing technique. Timeout checkpoint formation may be used with conventional processor backup techniques as well as with a novel backtracking technique including processor backup and backstepping.

Growing community of inventors

Oakland, CA, United States of America

John R Szeto