Silicon and physical substrates

In-memory and memristive computing

Use physical memory state to perform selected computation near or within storage arrays.

emerging deviceexperimental platform

Working definition

In-memory computing reduces some data movement by performing operations where weights or state are stored; memristive and resistive devices may encode conductance and exploit array physics for accumulation or plasticity. Useful evaluation must include programming, readout, conversion, endurance, drift, variability, yield, peripheral circuitry, and correction rather than idealized array operations alone.

Mechanism

  • Program physical memory states.
  • Apply voltages or events that produce an aggregate response.
  • Digitize, calibrate, or update the resulting state.

Measurements

  • Programming energy and endurance
  • Read accuracy and drift
  • Array-plus-periphery task performance

Reproducibility controls

  • Version hardware, software, firmware, and analysis code.
  • Declare dataset, preprocessing, random seeds, and measurement boundary.
  • Report repeated runs, variation, exclusions, and failed trials.

Limits and failure modes

  • Device demonstrations may omit system overhead.
  • Analog crossbars do not solve arbitrary workloads.

Mathematical connection

Formal structure without substrate erasure

computational modelOptimization

Array operation and nonideality

Optimize a mapped array operation against measured array-plus-periphery constraints and error.

Inputs

  • Conductance matrix
  • Input vector
  • Nonideality model

Outputs

  • Measured transform
  • Residual error
  • Correction cost

Limit: The optimized array abstraction does not include unmeasured programming, endurance, conversion, routing, yield, or thermal behavior.

Technical and governance sources

  1. [1]NeuroBench: Advancing Neuromorphic Computing Through Collaborative, Fair and Representative Benchmarking · National Institute of Standards and Technology

    Establishes: A community framework separating algorithm and system tracks and defining task, correctness, efficiency, and reporting procedures intended to make neuromorphic results more comparable and reproducible.

    Boundary: A benchmark ranks submitted systems on declared tasks and metrics. It does not prove general intelligence, biological equivalence, safety, usefulness outside the benchmark, or superiority under unreported host and data costs.

Related concepts