Abstract
Simulating specific data with known distributions is essential in various research fields. To generate values of a random variable, Monte Carlo simulations are widely used, relying on the Cumulative Distribution Function (CDF) of the process. Modern FPGAs can perform this task by combining a uniform random number generator with memory that stores CDF values. However, histogram regions with low occupancy tend to suffer from reduced performance, since their contribution to the integral is negligible. To address this issue, this work proposes the use of Read-Only Memory (ROM) blocks with varying resolutions, adapted to the occupancy of different regions. As demonstrated, this approach improves data representation without increasing circuit complexity or consuming excessive FPGA resources. The proposed architecture uses less than 3% of the FPGA logic and supports operation at high clock frequencies.