FPGA & data paths
Fixed-Point Rounding & Saturation
Compare rounding toward zero, floor and nearest-even before narrowing a signed value.
METHOD 1.0.0 · 2026-11-19
How the result is calculated
Scale the integer by 2^discarded_bits using exact BigInt arithmetic. Apply the selected rounding rule, then compare with the signed target range.
Reference example
−6 / 4 is −1.5. Nearest-even returns −2; truncation toward zero returns −1.
The example is a calculation or file-check case, not a hardware measurement.
Method references
- ZipCPU Rounding Numbers without Adding a Bias
- Project F Fixed Point Numbers in Verilog
- AMD UG1399 Fixed-Point Identifier Summary, 2025.1
Keep the conditions with the result
The exported record includes the inputs, calculation time, method version and limitations. Modified inputs clear the old result. Files are read locally; downloading does not send them to FPGA.io.
Use the actual device data and operating conditions. A positive budget or arithmetic check does not establish timing closure, compatibility, manufacturing acceptance or hardware performance.
Report a calculation issue with the tool name, inputs and expected result →