Mechanism to schedule threads on OS-sequestered sequencers without operating system intervention

A method comprising:

• executing a user-generated thread primitive;
• generating, by a first core of a multi-core processor, the first core visible to an operating system (OS) and a second computation unit of the multi-core processor invisible to the OS, responsive to the user-generated thread primitive, a descriptor to describe a thread including a beginning address for execution of the thread and a stack descriptor to identify a memory storage area;
• storing the descriptor in a work queue, wherein the work queue includes a first queue associated with the first core and a second queue associated with the second computation unit;
• and
• scheduling execution of the thread on the second computation unit using a distributed user-level scheduler invoked on the first core and the second computation unit.

At least one non-transitory computer-readable medium comprising instructions that when executed enable a system to:

• execute a user-generated thread primitive;
• generate, by a first core of a multi-core processor, the first core visible to an operating system (OS) and a second computation unit of the multi-core processor invisible to the OS, responsive to the user-generated thread primitive, a descriptor to describe a thread including a beginning address for execution of the thread and a stack descriptor to identify a memory storage area;
• store the descriptor in a work queue, wherein the work queue includes a first queue associated with the first core and a second queue associated with the second computation unit;
• and
• schedule execution of the thread on the second computation unit using a distributed user-level scheduler invoked on the first core and the second computation unit.