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

An apparatus comprising:

• a first thread execution unit that includes a first portion of a distributed operating-system-independent scheduler routine;
• a second thread execution unit that includes a second portion of a distributed operating-system-independent scheduler routine;
• wherein said first and second thread execution units are capable of concurrent multithreaded execution;
• wherein said first and second portions are further to query a work queue to contend for a work descriptor that describes a thread;
• and
• wherein said distributed scheduler routine is to schedule, without intervention of an operating system, the thread for execution on a selected one of the thread execution units, based on said contention.

A multi-sequencer multithreading system comprising:

• a memory system;
• a first sequencer of a first sequencer type;
• and
• a second sequencer of a second sequencer type;
• and
• a software library, stored in the memory system, the software library including one or more instructions to generate a work queue;
• the library further including a distributed scheduler to perform user-directed scheduling of work from the work queue for execution on the sequencers.

In a multi-threaded processor, a method of scheduling execution of software threads among a plurality of sequencers, comprising:

• responsive to a user-generated thread instruction, generating a thread descriptor that describes a thread;
• placing said thread descriptor in a queue;
• executing a plurality of scheduler routines, each of the routines being associated with a distinct sequencer, wherein said plurality of sequencers are not subject to an operating-system based thread scheduler;
• wherein each of said scheduler routines contends for the descriptor in the queue;
• and
• scheduling the thread for execution on one of the sequencers.

An article comprising a machine-accessible medium having a plurality of machine accessible instructions, wherein, when the instructions are executed by a processor, the instructions provide for:

• executing a user-generated thread primitive, wherein said executing further comprises invoking a library function that generates a thread control instruction;
• generating, responsive to the thread primitive, a work descriptor to describe a thread;
• causing the work descriptor to be stored in queue system;
• and
• scheduling execution of the thread on a sequencer without intervention of an operating system.