OpenCL Kernels and the Final Chip
May 13, 2022

Accelerator kernels should move to the best hardware without forcing programmers to configure every device, buffer, route, and security boundary. The Final Chip architecture places and manages OpenCL-style computation across an expandable cluster of reconfigurable silicon.

A new language can make allocation, ownership, communication, and reconfiguration part of one program while the physical fabric grows from a device to a data center and beyond.

Turn Kernels Into Managed Objects

Each kernel becomes a region with explicit inputs, outputs, state, and capabilities. The runtime can place that region near data, connect it to services, replicate it for load, and change its hardware allocation as demand shifts.

Local object-capability relationships protect sensitive resources without wrapping every interaction in an unrelated software stack.

Build the Research Partnership

Microsoft’s history with FPGA acceleration, cloud infrastructure, developer platforms, and HoloLens research makes it a natural partner for this architecture. The work connects chips, language design, distributed systems, and elastic service delivery.

Research labs, cloud providers, and semiconductor partners can host, fund, or co-develop the next implementation. Bring the kernels, infrastructure, and fabrication path that can carry the Final Chip into the world.

Build the Fabrication Path Behind OpenCL Kernels and the Final Chip

The Missing Maker Fab carries this work into the tools and processes required to fabricate fast active devices locally.

The Missing Maker Fab

Originally posted on LinkedIn

Brian Greenforest · (2022-05-13 01:38:06 UTC)

Open the original LinkedIn post · LinkedIn activity 6930692618993500160

LinkedIn status when archived: Visible to anyone on or off LinkedIn.

We have a good set of ideas, how OpenCL kernels can be automatically allocated and managed in a coding and configuration-less distributed and secure manner (in a new, not yet developed, programming language), with INFINITE chips cluster scalability. Our chips could be EVERYWHERE, if we succeed with design and tape out. We wanted to call it "Final Chip" (after "Final Fantasy"), but realized there were multiple Fantasies... Microsoft for Startups / Microsoft / Microsoft Leap / HoloLens Research Project / Microsoft Developers please give our startup a humble place at your research lab, or please fund us... There's no owner who will be happier. You knew what to do with FPGAs since Bing's regex parser, and what you're doing right now is amazing! We'll be a great fit for your company. We REALLY don't want Intel to get onto our IP by choice or an accident. PLEASE SAVE US ASAP. #scalability #startups #research #coding #microsoft #reactiveprogramming #distributedcomputing #hpc #fpga #chiplets #smartdust #claytronics #nanobots #curecancer