Paths to a chip

From an algorithm to a chip.

Any computation can be built as hardware, and it is worth a chip when it is fixed, heavy and repeated often enough. You bring the algorithm; we bring the architecture, the RTL and the route to tape-out.

For
Software and product teams whose algorithm has outgrown its processors
You provide
The algorithm as code or a model, test data, and your throughput, power and cost targets
You receive
A chip specified, designed and taped out for your product
Engagement
Spec-in: you bring the requirements; we deliver production silicon

How it gets there

From code to gates, step by step.

  1. Profile

    Where the algorithm spends its time, memory and energy, and which part is worth hardware.

  2. Fix, or keep programmable

    What becomes fixed logic, and what stays programmable so that the chip survives a change to the algorithm.

  3. Numerics

    The word lengths and number formats the result really needs.

  4. Architecture

    Datapath, memory and control sized to the workload.

  5. RTL

    Written and verified with your software as the reference model; high-level synthesis where it suits the block.

  6. FPGA prototype

    The design proven on FPGA with your real data, where the product calls for it.

  7. ASIC

    Physical design, sign-off and tape-out.

Questions about turning an algorithm into a chip

Can any software be turned into a chip?

Any computation can. Whether it should depends on three things: the algorithm will stay as it is, hardware does the work far more efficiently than a processor, and you will run it often enough to carry the one-off cost. A feasibility study answers all three.

Do we need to know hardware design?

No. In a spec-in engagement you bring the requirements and the reference software, and we do the hardware.

What if the algorithm changes later?

The architecture decides that. The parts likely to change stay programmable or configurable; the parts that will not change are fixed, which is where the efficiency comes from.

We already have RTL generated from our code. What now?

Start with an RTL assessment: it shows what that RTL still needs before it can be built.

How is our algorithm protected?

Every project runs under a mutual NDA, with access limited to the named project team; ownership of the design and its data is set in each project's contract.

An algorithm that has outgrown its processors?

Tell us what it computes, how often, and what it costs you today.

  1. Fill in the project form
  2. We reply by email within a few working days
  3. Onboarding: NDA, design review and a staged plan

The form is the fastest way to reach us. A person reads every request.