Processor including replay queue to break livelocks

A method of breaking a stalled condition or livelock in a processor having a replay queue, comprising;

• detecting a livelock or stalled condition;
• temporarily storing one or more instructions in a replay queue, with the instructions in the replay queue not including an oldest instruction in the processor;
• detecting that the livelock or stalled condition has been broken or released;
• and
• unloading the instructions from the replay queue for replay or re-execution.

A method of breaking a stalled condition or livelock in a processor having a replay queue, comprising:

• detecting a livelock or stalled condition;
• temporarily storing one or more instructions in a replay queue by storing all of the instructions in the replay queue except an oldest instruction;
• detecting that the livelock or stalled condition has been broken or released;
• and unloading the instructions from the replay queue for replay or re-execution.

A method of breaking a stalled condition or livelock in a processor having a replay queue, comprising:

• detecting a livelock or stalled condition;
• temporarily storing one or more instructions in a replay queue;
• detecting that the livelock or stalled condition has been broken or released by detecting that a retirement pointer has moved;
• and
• unloading the instructions from the relay queue for replay or re-execution.

A method of breaking a stalled condition in a multi-threaded processor having a replay queue, comprising:

• detecting a stall in one of threads by detecting that a retirement pointer of a thread has not changed for a minimum period of time;
• temporarily storing one or more instructions of the stalled thread in a replay queue;
• detecting that the stall has been broken or released;
• and
• unloading the instructions from the replay queue for replay or re-execution.

A method of breaking a stalled condition in a multi-threaded processor having a replay queue, comprising:

• detecting a stall in one of threads;
• temporarily storing one or more instructions of the stalled thread in a replay queue by storing a plurality of the instruction of the stalled thread in the replay queue, the plurality of instructions in the replay queue not including an oldest instruction of the stalled thread;
• detecting that the stall has been broken or released;
• and
• unloading the instructions from the replay queue for replay or re-execution.

A method of breaking a stalled condition in a multi-threaded processor having a replay queue, comprising:

• detecting a stall in one of threads;
• temporarily storing one or more instructions of the stalled thread in a replay queue except an oldest instruction of the stalled thread;
• detecting that the stall has been broken or released;
• and
• unloading the instructions from the replay queue for replay or re-execution.

A method of breaking a stalled condition in a multi-threaded processor having a replay queue, comprising:

• detecting a stall in one of threads;
• temporarily storing one or more instructions of the stalled thread in a replay queue by detecting an oldest instruction of the stalled thread based on a comparison of sequence numbers;
• and
• temporarily storing the instructions of the stalled thread in the replay queue except the oldest instruction of the stalled thread;
• detecting that the stall has been broken or released;
• and
• unloading the instructions from the replay queue for replay or re-execution.

A method of breaking a stalled condition in a multi-threaded processor having a replay queue, comprising:

• detecting a stall in one of threads;
• temporarily storing one or more instructions of the stalled thread in a replay queue;
• detecting that the stall has been broken or released by detecting that a retirement pointer of the stalled thread has moved or changed;
• and
• unloading the instructions from the replay queue for replay or re-execution.

A method of breaking a stalled condition in a multi-threaded processor having a replay queue, comprising:

• detecting a stall in one of threads by detecting a first thread to stall, the first thread to stall being detected by having a retirement pointer that is the first retirement pointer to remain fixed for a minimum period of time;
• temporarily storing one or more instructions of the stalled thread in a replay queue;
• detecting that the stall has been broken or released;
• and
• unloading the instructions from the replay queue for replay or re-execution.

A method of breaking a stalled condition in a multi-threaded processor having a replay queue, the method comprising:

• detecting that a retirement pointer for a first thread has not moved for a minimum period of time;
• detecting an oldest instruction of the first thread;
• temporarily storing a plurality instructions of the first thread in a replay queue, while allowing the oldest instruction of the first thread and instructions of other threads to be replayed;
• detecting that the retirement pointer of the first thread has made forward progress;
• and
• unloading the instructions from the replay queue for replay or re-execution in response to the step of detecting that the retirement pointer of the first thread has made forward progress.

A processor comprising:

• an execution unit to execute instructions;
• a replay system coupled to the execution unit to replay instructions which have not executed properly, the replay system comprising:
• a checker to determine whether each instruction has executed properly;
• and
• a replay queue coupled to the checker to temporarily store one or more instructions for replay;
• and
• a circuit, coupled to the replay system, to detect a stalled condition of the processor, one or more instructions being loaded into the replay queue when the stalled condition is detected.

A multi-threaded processor comprising:

• an execution unit to execute instructions;
• a replay system coupled to the execution unit to replay instructions which have not executed properly, the replay system comprising:
• a checker to determine whether each instruction has executed properly;
• and
• a replay queue coupled to the checker to temporarily store one or more instructions for replay;
• and
• a circuit, coupled to the replay system, to detect that one of the threads has stalled, one or more instructions of the stalled thread being loaded into the replay queue when the stalled condition is detected to allow instructions of other threads access to the execution unit.

A method in a multi-threaded processor, comprising:

• detecting a stall or stalled condition in the processor by detecting that a retirement pointer of one or more of the threads has not changed for a minimum or predetermined period of time;
• temporarily storing one or more instructions of one or more of the threads in a replay queue;
• detecting that the stall or stalled condition has been broken or released;
• and
• unloading the instructions from the replay queue for replay or re-execution.

A method in a multi-threaded processor having a replay queue, the method comprising:

• temporarily storing one or more instructions of a first thread in a replay queue, while allowing the instructions of one or more remaining threads to execute;
• unloading the first thread from the replay queue to allow the instructions of the first thread to execute;
• and
• temporarily storing one or more instructions of a second thread in the replay queue, while allowing instructions of the one or more remaining threads to execute.

A processor comprising:

• an execution unit to execute instructions;
• a replay system coupled to the execution unit to replay instructions which have not executed properly, the replay system comprising:
• a checker to determine whether each instruction has executed properly;
• and
• a replay queue coupled to the checker to temporarily store one or more instructions for replay;
• and
• a circuit, coupled to the replay system, to temporarily store instructions of a first thread in the replay queue while allowing the instructions of one or more remaining threads to execute, and then to unload the first thread from the replay queue to allow the instructions of the first thread to execute.

A method in a multi-threaded processor having a replay queue, the method comprising:

• detecting a stall or stalled condition with a first thread;
• temporarily storing one or more instructions of the first thread in a replay queue, while allowing the instructions of one or more remaining threads to execute;
• performing the following if the stalled condition is not released within a predetermined period of time:
• unloading the first thread from the replay queue to allow the instructions of the first thread to execute;
• temporarily storing one or more instructions of a second thread in the replay queue, while allowing instructions of the one or more remaining threads to execute;
• unloading the second thread from the replay queue;
• and
• reloading one or more instructions of the first thread into the replay queue until the stalled condition with the first thread is released.