Method and apparatus for implementing a parallel construct comprised of a single task

A system comprising:

• a translating unit to translate a code, the code having an instance of a parallel construct, the instance associated with a task, the translating unit to generate a code segment:
• the code segment when executed to cause each of a plurality of threads:
• to determine if the task has been undertaken by another thread by comparing a first private value of the thread with a second private value of the thread,
• if the task has not been undertaken by another thread, to undertake the task and to indicate with a shared value the undertaking of the task, and
• to indicate with the first private value the instance;
• a linker unit coupled to the translating unit, the linker unit to link the translated code with a library;
• and
• a set of processors coupled to the translating unit and the linker unit, the set of processors to host the plurality of threads.

A machine-accessible medium containing instructions that, when executed, cause a machine to:

• receive a first code segment, the first code segment having a set of instances of a parallel construct, each of the instances comprised of a task;
• and
• transforming the first code segment to a second code segment,
• the second code segment that when executed:
• to initialize a shared value for each of a team of threads, the shared value to indicate a most current instance encountered by one of the threads,
• to initialize a first private value and a second private value for each thread, the first private value to indicate one of the instances encountered by the first private value's corresponding thread,
• to determine if the task of the most current instance has been undertaken by comparing a first private value of the first thread and a second private value of the first thread,
• upon determining the task has not been undertaken, undertaking the task and indicating with the shared value the undertaking of the task,
• to update the shared value with the team of threads, and
• to update the private value of each of the team of threads with the private value's corresponding thread.

A machine-accessible medium containing instructions that, when executed, cause a machine to:

• receive a first code segment, the first code segment having a set of instances of a parallel construct, each of the set of instances being associated with a task;
• and
• for each instance, to generate a second code segment, the second code segment that when executed to:
• encounter one instance,
• determine if the task of the one instance has been undertaken by comparing a first private value of the first thread and a second private value of the first thread,
• upon determining the task has not been undertaken, undertaking the task and using a shared value to indicate the undertaking of the task,
• using the first private value of the first thread to indicate the one instance, and
• maintain the second private value of the first thread, the second private value being a boundary for accessing the shared value.