Transformations of a 3D image reconstruction algorithm for data transfer and storage optimisation

When implementing a 3D image reconstruction algorithm on a DSP architecture , we find ourselves confronted with a large memory transfer overhead, reduc ing the possible speedup attainable on recent multi-media oriented architec tures. This paper describes how the critical part of the algorithm is re-sp ecified and aggressively transformed at the algorithm code level, to improv e the data access locality of the multi-dimensional image signal, while pre serving the input/output behaviour. Experiments show that a close to optima l reuse of the data in the foreground memory and registers is obtained, rem oving the data transfer and storage bottleneck and enabling real-time proto typing of the algorithm on a DSP architecture.