VHDL look-up tables from a math function

What LutCraft does

LutCraft turns a floating-point reference model — a math expression like sqrt(x), log(x) or a sine — into a fixed-point VHDL look-up table, and reports the approximation error introduced by LUT sampling and output quantization. You choose the input regions and how many entries each gets, so you can spend resolution where the function changes fastest.

What it produces

• A synchronous VHDL ROM addressed by a LUT index.
• A combinational fixed-point input-to-index mapper.
• An error report (max / RMS error per region) with the sampled model, quantized values and local error.
• A machine-readable report.json and a manifest.json of the generated artifacts.
• The above packaged as a single downloadable bundle.

When it helps

Use LutCraft when you need a fixed-point approximation of a function on an FPGA and a hand-built table would drift away from its error budget as the design changes. Typical cases: sqrt / log / sin and other transcendental functions, ADC calibration curves, sigmoid and other activation functions, and waveform LUTs.

A free analysis shows the numerical tradeoff — error metrics, sampled values and optional plots — without writing any VHDL. Generating the synthesizable bundle is a paid step.

LUT generation from your own sampled data (CSV) isn't part of the web service yet — get in touch to discuss a job.

Sign in to start Talk to us