GPU accelerated computation of CREC

Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27240%2F15%3A86096376" target="_blank" >RIV/61989100:27240/15:86096376 - isvavai.cz</a>
Alternative codes found
RIV/61989100:27740/15:86096376
Result on the web
—
DOI - Digital Object Identifier
—

Result language
angličtina
Original language name
GPU accelerated computation of CREC
Original language description
One of the most important risks that a bank runs is credit risk: the risk of default of its customers. Credit risk can threaten the existence of the bank if there are concentrations in the bank portfolio and the different loans are closely correlated. Animportant question for a bank is how much buffer is needed to absorb credit related losses, typically with a very high level of confidence (e.g. 99.95%). Since the correlations in the portfolio cannot be assessed analytically only possible way is stochastic simulation via Monte Carlo. In this type of simulation, many different possible future economic scenarios are generated and the losses per scenario are determined. The problem with this approach is that it is very slow. In this project the problem was solved using parallel GPU-CUDA implementation. Optimized version of parallel implementation was able to solve problem over 50 times faster, current CPU parallel version took 3hours 20 minutes and new CUDA version took 4minutes.
Czech name
—
Czech description
—

Publication year
2015
Confidentiality
U - Předmět řešení projektu je utajovanou skutečností podle zvláštních právních předpisů nebo je skutečností, jejíž zveřejnění by mohlo ohrozit činnost zpravodajské služby. Údaje o projektu jsou upraveny tak, aby byly zveřejnitelné

Similar results(10)