|
ИСТИНА |
Войти в систему Регистрация |
Интеллектуальная Система Тематического Исследования НАукометрических данных |
||
Parallel computing and challenges for thin film optics technologyтезисы доклада
- Авторы: Tikhonravov Alexander V., Trubetskov Michael K.
- Сборник: Proceedings of the Second International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering
- Тезисы
- Год издания: 2011
- Место издания: Civil-Comp Press Stirlingshire, UK
- Первая страница: 21
- DOI: 10.4203/ccp.95.21
- Аннотация: The most practical design must provide the highest production yield in a process of optical coating manufacturing. Instead of using costly production runs, computer simulations of real production processes are now considered as the most promising approach to choosing optimal designs for challenging applications [4]. The theory of probability predicts that reliable production yield estimations may require thousands of computational manufacturing experiments. Computational time of a single computational manufacturing experiment is dependent on a type of monitoring approach used for controlling thicknesses of coating layers. Experiments simulating production runs with optical monitoring are the most time consuming experiments. Depending on the number of layers of examined theoretical design, a single experiment with broadband optical monitoring may require up to several minutes on a personal computer. Thus performing thousands of computational experiments with multiple theoretical designs is possible only with the parallelization of computational manufacturing experiments. Parallelization of algorithms used for these experiments was performed both for SMP configurations on the basis of OpenMP and for GPU architecture on the basis of CUDA technology. For the latest Fermi-based GPUs the algorithm implementation is three to four times faster than the SMP configurations with 2 dual-core 3 GHz Intel Xeon processors.
- Добавил в систему: Тихонравов Александр Владимирович