Performing collective operations using software setup and partial software execution at leaf nodes in a multi-tiered full-graph interconnect architecture

This invention describes a method for a main processor to efficiently manage complex, coordinated tasks (collective operations) across many other processors in a large, multi-level computer system. The main processor's software first organizes the other processors into a logical hierarchy, then distributes the task, and finally its specialized hardware gathers and combines all the results into a single output. The claims specify this system must have a very particular hardware structure of "supernodes" containing "processor books" which contain individual processors.

Why it matters: The explosion of large-scale distributed computing and AI/ML training since 2007 makes efficient collective operations critical. Modern cloud and accelerator architectures could benefit from hierarchical coordination and hardware-accelerated aggregation, despite the specific 'supernode' hardware described.

Status: ExpiredProtection ended and was never renewed
How hard to build: SpecializedNeeds custom high-performance computing hardware, software logic

3 AI reinvention directions for this patent

See all three

AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.

Reinvent this with AI

Where this one came from

Every morning we go through millions of abandoned patents and pick out 15 worth your time. This was one of today’s.

This one

+10