3D graphics coding in a reconfigurable environment

• We introduce the concept of Reconfigurable Graphic Coding.
• We propose a set of Functional Units that combined can build 3D graphics decoders.
• We describe dataflow network for two decoders.
• We compare the results between the generated C and native C++ implementation.

The main objective of this paper is to introduce the concept of Reconfigurable Graphic Coding and its validation under the form of a Functional Units (FU) library. The heterogeneity of data for 3D graphics objects representation requires the adaptability of the compression schemas to various types of content. While such adaptation can be relatively easy to support in software implementations, the same is much more difficult to implement in hardware. Although compression schemas inherently share the same data processing chain, the components forming it may vary with respect to the number and type of components to encode, data range and correlation type. Based on the analysis of the state of the art on 3D graphics compression approaches, we propose a set of processing units. We show how this set can be configured/connected into a network, including hardware networks, to obtain reference decoders. Moreover, the network can be reconfigured at runtime, based on information that is provided with the encoded object. This modular concept of functional units, allows optimized management of computation (such as identification of parallelizable functions or functions that are suitable for acceleration) relative to the hardware architecture (CPU, GPU, FPGA, etc.). In the four decoders presented, at least half of the FUs are being reused at least once. The results were performed by generating and compiling C code from RVC-CAL code and comparing the results with the MPEG reference software implementation. The FUs described in this paper were standardized by MPEG as part of the ISO/IEC 23001-4.