|
Symmetric multiprocessing enables users to break a program into separate tasks and run multiple tasks simultaneously, vastly improving the throughput of large computational applications such as solutions to linear systems or fast Fourier transforms. IMSL CNL enables C programmers to leverage the power of SMP in their programs, allowing them to achieve significant performance improvements as problem size and the number of processors increase. Benchmark programs demonstrated a 3X-performance increase when solving a 2000 x 2000 linear system just by increasing the number of processors during execution from one to four.
In addition to bringing SMP capabilities to users, IMSL CNL 2.5 delivers performance enhancements that improve computational efficiency in the areas of memory management and array processing. The software helps C development programmers reduce execution times and improve performance by efficiently utilizing the cache memory through the use of dynamic workspace and array manipulation.
IMSL CNL 2.5 also includes functionality improvements in the following areas: Statistics Probability distribution functions and inverses Multivariate linear regression Random number generation Survival analysis Fourier Transforms Real and complex one-dimensional fast Fourier transforms Complex two- dimensional fast Fourier transforms Linear Systems Sparse linear solvers General linear solvers Positive definite linear solvers Interpolation and Approximation Radial basis approximation and fit evaluation Optimization Nonlinear programming Nonlinear least squares with and without bounds Linearly constrained minimization Unconstrained multivariate minimization Miscellaneous Error processing Version information License manager