This paper explores the performance of different nonlinear information encoding strategies in diffractive optical processors using linear materials. The study compares the statistical inference performance of phase encoding, which is simpler to implement, against data repetition-based nonlinear encoding strategies. Data repetition within a diffractive volume, such as through an optical cavity or cascaded input data, results in the loss of the universal linear transformation capability of a diffractive optical processor, making it unable to perform fully connected or convolutional layers commonly used in digital neural networks. However, data repetition-based diffractive blocks can still achieve enhanced accuracy for specific inference tasks. Phase encoding, without data repetition, provides a simpler nonlinear encoding strategy with comparable statistical inference accuracy to data repetition-based diffractive processors. The results highlight the push-pull relationship between linear material-based diffractive optical systems and nonlinear encoding strategies in visual information processing. The study also discusses the advantages and limitations of data repetition-based approaches, including noise resilience, and suggests future directions for integrating optical nonlinear materials and developing hybrid systems with both optical and digital components.This paper explores the performance of different nonlinear information encoding strategies in diffractive optical processors using linear materials. The study compares the statistical inference performance of phase encoding, which is simpler to implement, against data repetition-based nonlinear encoding strategies. Data repetition within a diffractive volume, such as through an optical cavity or cascaded input data, results in the loss of the universal linear transformation capability of a diffractive optical processor, making it unable to perform fully connected or convolutional layers commonly used in digital neural networks. However, data repetition-based diffractive blocks can still achieve enhanced accuracy for specific inference tasks. Phase encoding, without data repetition, provides a simpler nonlinear encoding strategy with comparable statistical inference accuracy to data repetition-based diffractive processors. The results highlight the push-pull relationship between linear material-based diffractive optical systems and nonlinear encoding strategies in visual information processing. The study also discusses the advantages and limitations of data repetition-based approaches, including noise resilience, and suggests future directions for integrating optical nonlinear materials and developing hybrid systems with both optical and digital components.