Stochastic Computing

StochastiCore

v0.1Alpha· hardware

An open-source GPU architecture whose arithmetic datapath runs on stochastic computing — multiplication becomes a single AND gate, scaled addition a multiplexer.

§ 01

About StochastiCore

A complete, synthesizable GPU pipeline — command processing, stochastic vertex transform, binary rasterization, stochastic fragment shading, framebuffer/Z-buffer, and VGA output — implemented in open-source Verilog. The first open GPU whose arithmetic datapath is stochastic, with a reproducible Yosys synthesis flow and an activity-driven gate-level power model.

§ 02

Highlights

  • ▸First open GPU pipeline with a stochastic-computing arithmetic datapath
  • ▸Reproducible Yosys gate-level synthesis + activity-driven power model
  • ▸Deterministic low-discrepancy generator: 7.5× less generator area, 15.4× less switching energy
  • ▸Bit-accurate rendering at 36.7 dB PSNR; free anti-banding from white SC noise
  • ▸Manuscript under review at IEEE Transactions on VLSI Systems
§ 03

Publications from this lab

  • IEEE Trans. VLSI Systems · Under review · APR 30, 2026
    StochastiCore: A Complete GPU Architecture Based on Stochastic Computing

    Graphics and inference accelerators are built from binary multipliers and adders whose area and switching cost grow with operand width. Stochastic computing (SC) encodes a value as the probability of a bit being one in a random stream, reducing multiplication to a single AND gate and scaled addition to a multiplexer, with intrinsic tolerance to bit errors. We present StochastiCore, the first synthesizable GPU pipeline whose arithmetic datapath is stochastic — command processing, SC vertex transform, binary rasterization, SC fragment shading, and framebuffer/VGA output in open-source Verilog. Beyond integration we contribute a theoretical analysis of SC graphics, a reproducible Yosys gate-level power and area characterization, and a deterministic low-discrepancy bitstream generator that cuts generator area 7.5× and switching energy 15.4×. Bit-accurate rendering on the verified LFSR model reaches 36.7 dB PSNR; we report honestly that SC's advantage is area parity at higher functional density, graceful degradation, and free anti-banding rather than a per-operation energy win.