Light Transport Simulation Using Decomposed Distribution Functions

The distribution function is decomposed into multiple discrete probability distributions, each representing different scattering scenarios. This segmentation allows the complex measured material data to be processed in manageable discrete parts, reducing instruction divergence by grouping similar operations together while maintaining measurement precision through faithful representation of the original distribution.

The patent transforms the continuous measured material data into discrete probability distributions with specific parameters (weights and discrete values). This parameter transformation enables efficient SIMD/SIMT processing by converting complex continuous data into structured discrete form that reduces memory access patterns and instruction divergence, while the decomposition preserves the essential characteristics of the original measured data.

Instead of using a single complex continuous model, the patent applies multiple discrete probability distributions that collectively cover the full range of scattering behaviors. This partial action approach breaks down the complex simulation into simpler discrete components that are easier to process, while the cumulative effect of multiple distributions achieves the precision of more complex continuous models.

The patent creates discrete copies of the measured material data in the form of probability distribution tables that can be efficiently accessed and processed. These discrete representations copy the essential scattering characteristics of the original measured data into a format suitable for SIMD/SIMT architectures, maintaining visual quality while reducing computational complexity.

The patent replaces floating-point arithmetic operations with integer-based discrete probability sampling. By substituting the continuous mathematical mechanism with discrete lookup tables and integer comparisons, the system eliminates expensive floating-point computations while maintaining sampling accuracy. This substitution is particularly effective on SIMD/SIMT architectures where integer operations execute more efficiently and with less divergence.

The probability distribution functions are pre-computed and stored as discrete tables before the actual light transport simulation. This preliminary action converts complex continuous distributions into ready-to-use discrete forms, eliminating the need for real-time floating-point calculations during rendering. The pre-processing step maintains measurement precision by accurately representing the original distributions, while the runtime execution achieves high productivity through simple discrete sampling.