LS-DYNA PERFORMANCE ON ULTRASPARCTM -III SERVERS AND WORKSTATIONS
Sun Microsystems recently announced a new line of high-performance Sun FireTM Midframe servers and Sun BladeTM 1000 workstations based on 750-MHz UltraSPARC-III microprocessors. These new computers offer exceptional performance for numerically intensive MCAE applications at reasonable prices. This paper presents single- and multi-processor performance and scalability results of LS-DYNA running on these systems. Results of both OpenMP and MPI binaries are compared to those of existing server platforms. Scalability and performance results are presented for the OpenMP binary that has been built with Sun's native OpenMP compiler extension. Comparisons with a previous version using libraries from KAI show significant improvements both in the absolute execution time and the scalability. The MPI executable tuned for the new Sun Fire server resulted in a 40-60% improvement in execution time over the entire range of processor counts.
https://www.dynalook.com/conferences/european-conf-2001/ls-dyna-performance-on-ultrasparctm-iii-servers.pdf/view
https://www.dynalook.com/@@site-logo/DYNAlook-Logo480x80.png
LS-DYNA PERFORMANCE ON ULTRASPARCTM -III SERVERS AND WORKSTATIONS
Sun Microsystems recently announced a new line of high-performance Sun FireTM Midframe servers and Sun BladeTM 1000 workstations based on 750-MHz UltraSPARC-III microprocessors. These new computers offer exceptional performance for numerically intensive MCAE applications at reasonable prices. This paper presents single- and multi-processor performance and scalability results of LS-DYNA running on these systems. Results of both OpenMP and MPI binaries are compared to those of existing server platforms. Scalability and performance results are presented for the OpenMP binary that has been built with Sun's native OpenMP compiler extension. Comparisons with a previous version using libraries from KAI show significant improvements both in the absolute execution time and the scalability. The MPI executable tuned for the new Sun Fire server resulted in a 40-60% improvement in execution time over the entire range of processor counts.
29.pdf
— 610.3 KB
