Optimizing the Use of Static Buffers for DMA on a CELL Chip
The CELL architecture has one Power Processor Element (PPE) core, and eight Synergistic Processor Element (SPE) cores that have a distinct instruction set architecture of their own. The PPE core accesses memory via a traditional caching mechanism, but each SPE core can only access memory via a small 256K software-controlled local store. The PPE cache and SPE local stores are connected to each other and main memory via a high bandwidth bus. Software is responsible for all data transfers to and from the SPE local stores. To hide the high latency of DMA transfers, data may be prefetched into SPE local stores using loop blocking transformations and static buffers. We find that the performance of an application can vary depending on the size of the buffers used, and whether a single-, double-, or triple-buffer scheme is used. Constrained by the limited space available for data buffers in the SPE local store, we want to choose the optimal buffering scheme for a given space budget. Also, we want to be able to determine the optimal buffer size for a given scheme, such that using a larger buffer size results in negligible performance improvement. We develop a model to automatically infer these parameters for static buffering, taking into account the DMA latency and transfer rates, and the amount of computation in the application loop being targeted. We test the accuracy of our prediction model using a research prototype compiler developed on top of the IBM XL compiler infrastructure.
Unable to display preview. Download preview PDF.
- 1.Chen, W., Iancu, C., Yelick, K.: Communication optimizations for fine-grained UPC applications. In: Parallel Architecture and Compilation Techniques (2005)Google Scholar
- 2.Eichenberger, A.E., et al.: Optimizing compiler for the CELL processor. In: Parallel Architecture and Compilation Techniques, September (2005)Google Scholar
- 3.Flachs, B., et al.: A streaming processing unit for a CELL processor. In: IEEE International Solid-State Circuits Conference (ISSCC), February 2005, IEEE Computer Society Press, Los Alamitos (2005)Google Scholar
- 4.Pham, D., et al.: The design and implementation of a first-generation CELL processor. In: IEEE International Solid-State Circuits Conference, February 2005, IEEE Computer Society Press, Los Alamitos (2005)Google Scholar
- 5.European Center for Parallelism of Barcelona (CEPBA). Paraver: Parallel program visualization and analysis tool reference manual (November 2000), http://www.cepba.upc.es/paraver
- 6.Hiranandani, S., Kennedy, K., Tseng, C-W.: Evaluation of compiler optimizations for Fortran D on MIMD distributed memory machines. In: International Conference on Supercomputing (1992)Google Scholar
- 7.Iancu, C., Husbands, P., Chen, W.: Message strip mining heuristics for high speed networks. In: 6th International Meeting of VECPAR (2004)Google Scholar
- 8.Iancu, C., Husbands, P., Hargrove, P.: HUNTing the overlap. In: Parallel Architecture and Compilation Techniques, September (2005)Google Scholar
- 9.Ishizaki, K., Komatsu, H., Nakatani, T.: A loop transformation algorithm for communication overlapping. International Journal of Parallel Programming (2000)Google Scholar
- 10.Kapasi, U.J., et al.: Stream scheduling. In: Proceedings of the 3rd Workshop on Media and Streaming Processors (2001)Google Scholar
- 11.Kistler, M., Perrone, M., Petrini, F.: CELL multiprocessor communication network: Built for speed. IEEE Micro 26(3) (2006)Google Scholar
- 12.Leu, J.S., Agrawal, D.P., Mauney, J.: Modeling of parallel software for efficient computation communication overlap. In: Proceedings of the 1987 Fall Joint Computer Conference on Exploring Technology: Today and Tomorrow (1987)Google Scholar
- 13.Palermo, D.J., et al.: Communication optimizations used in the PARADIGM compiler for distributed memory multicomputers. In: International Conference on Parallel Processing, August (1994)Google Scholar