Single vs. double precision
CFD solvers, such as ANSYS CFX and ANSYS FLUENT, have options for the user to select single- or double-precision versions of the software. Explicit dynamics solvers (a.k.a. hydrocodes) such as ANSYS LS-DYNA and ANSYS AUTODYN also have single- or double-precision executables.
The question may arise why ANSYS does not have the same option. ANSYS is always in double-precision, namely because equation solvers such as the direct sparse solver factorizes the matrix. Matrix factorization, along with ill-conditioned matrices that arise from Lagrange Multipliers or coupled-field elements, need double-precision to maintain sufficient accuracy.
(It is interesting to note that the PCG iterative solver in ANSYS can use single precision to save some memory via the PRECISION command. I do not recommend using this, however, since the reduced memory is not worth the possible loss of accuracy if matrices are ill-conditioned. Instead, consider using the MSAVE command to reduce memory requirements.)
Comments are closed.