GPU ray tracing with OptiX for a Monte Carlo particle transport software

This postdoctoral research subject is proposed in the context of the development by CEA List of the CIVA software platform, used for the simulation of non-destructive testing techniques, in particular X-radiography and computed tomography. This research is supported by a collaborative project between two CEA divisions (DRT and DES). It aims to optimize the performance of the Monte-Carlo method used in the CIVA software in order to improve the processing of complex industrial radiographic configurations. More precisely, More precisely, the following two tasks will be carried out:
- Developing a eTLE-fd type estimator on CPU in the CIVA environment, in order to accelerate the Monte-Carlo calculation by generating pseudo-particles transported directly to the detector. This estimator, already implemented in a DES code (TRIPOLI-4®), will have to be adapted to the NDT applications processed by CIVA.
- Port this estimator to GPU using Nvidia's OptiX library in order to optimize the Monte-Carlo calculation, including taking into account particle attenuation and energy deposition, to be coded in CUDA.