There are two levels of precision that can be utilized in GROMACS - single and double.  Both are floating-point formats that allow data to occupy different amounts of memory.

    Single Precision (mixed precision)

    When installing GROMACS, the default behavior is to utilize single precision.  This level of precision is sufficient for most calculations. Note that this is actually mixed precision. The coordinate, velocity and force vectors are single precision, but where necessary, double precision variables are used. A typical example is the accumulation buffer for the virial. In many parts of the code we pay attention to summation order and/or reorder mathematical expressions to avoid rounding errors.

    Double Precision

    Some techniques, such as Normal Mode Analysis and more thorough energy minimization, require increased precision.  To install GROMACS in double precision, set GMX_DOUBLE=ON during configuration.

    Page last modified 14:22, 21 Jan 2013 by JLemkul?