Browsing by Author "Ojanguren, M."
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item Modelización global del proceso de colada continua(2005) Barco, J.; Palacios, J.; Ojeda, C.; Ojanguren, M.; Centros PRE-FUSION TECNALIA - (FORMER)In the continuous casting process the contribution of complicated physical phenomena has an important effect in the final quality of the product and productivity of the process. Moreover, the different physical phenomena do not act in isolation, they interact in a coupled way affecting the quality of the semiproduct. Therefore, a correct design and setting is a work that requires a good knowledge of the different phenomena that are present and their interactions. In this work, a simulation methodology that solves all the process as a whole is presented, where the variables of the process are the only boundary conditions. The interactions between the different phenomena are solved with the use of commercial applications (FLUENT and ABAQUS) and home developed models.Item Tuning LAPACK codes on hierarchical memory machines(1996-06) Ojanguren, M.; Alvarez, A.; Anza, J.; Longo, A.; Centros PRE-FUSION TECNALIA - (FORMER); PROMETALAs important as the advance of computer technology is the development of suitable software for the exploitation of computer capacity. In this sense LAPACK (Anderson et al., LAPACK User's Guide, release 1.0, SIAM, Philadelphia, 1992) appears as the most efficient library in the dense linear algebra field, obtaining good results on vector and parallel computers, and in general, on hierarchical memory machines. However, some deficiencies in the general matrix LU decomposition were detected in DGETRF subroutine. On hierarchical memory machines not only the cache and TLB faults have to be minimized, but also the page faults, which cause excessive I/O operations. The blocking strategy used by DGETRF (and LAPACK in general) makes good use of the cache memory, but does not seem to be enough to avoid unnecessary I/O operations. Therefore, DGETRF does not provide satisfactory run times for large dimension matrices. In this paper a new code using a double blocking strategy will be described, which attains better run times than DGETRF.