The paper presents an approach for solving unsymmetric sparse systems of linear equations using the PARDISO solver, which employs supernode pivoting, supernode partitioning, and asynchronous computation to achieve high performance on shared memory parallel computers. Supernode pivoting allows dynamic column and row interchanges during factorization, while prepermutation of rows places large entries on the diagonal. The BLAS-3 level efficiency is retained, and an enhanced left-right looking scheduling scheme results in good speedup on SMP machines without increasing the operation count. The algorithm integrates static pivoting and prepermutation to improve scalability and robustness. The PARDISO solver uses nested dissection for ordering and supernode technology to treat adjacent rows and columns with the same structure as one supernode. If a suitable pivot cannot be found within a supernode block, a static pivoting strategy is used. The numerical behavior of the approach is illustrated with experiments showing that it can solve nearly all unsymmetric matrices. The parallel LU algorithm uses a two-level scheduling approach to reduce synchronization events and improve scalability. Experimental results show that PARDISO performs well compared to other solvers like WSMP, with better scalability on SMPs. The paper concludes that PARDISO's approach provides stability and efficiency for unsymmetric sparse systems, with potential for further improvements in reordering schemes, dynamic pivoting, and postprocessing steps.The paper presents an approach for solving unsymmetric sparse systems of linear equations using the PARDISO solver, which employs supernode pivoting, supernode partitioning, and asynchronous computation to achieve high performance on shared memory parallel computers. Supernode pivoting allows dynamic column and row interchanges during factorization, while prepermutation of rows places large entries on the diagonal. The BLAS-3 level efficiency is retained, and an enhanced left-right looking scheduling scheme results in good speedup on SMP machines without increasing the operation count. The algorithm integrates static pivoting and prepermutation to improve scalability and robustness. The PARDISO solver uses nested dissection for ordering and supernode technology to treat adjacent rows and columns with the same structure as one supernode. If a suitable pivot cannot be found within a supernode block, a static pivoting strategy is used. The numerical behavior of the approach is illustrated with experiments showing that it can solve nearly all unsymmetric matrices. The parallel LU algorithm uses a two-level scheduling approach to reduce synchronization events and improve scalability. Experimental results show that PARDISO performs well compared to other solvers like WSMP, with better scalability on SMPs. The paper concludes that PARDISO's approach provides stability and efficiency for unsymmetric sparse systems, with potential for further improvements in reordering schemes, dynamic pivoting, and postprocessing steps.