Sunday September 26, 2010
Sometimes speed does matter. For example, which matrix library is faster when run under an industry-standard benchmark using a given workload?
-Much of the work done at the low levels of simulations is performed by library routines, such as manipulating matrices and calculating the square root of 1 / x. Toolchains such as GNU's and Intel's come with their own libraries for performing this work (and a host of other tasks), and there are optimized libraries available such as Goto's BLAS library for manipulating matrices.

-LINPACK is one of the most popular HPC benchmarks. It can measure a wide variety of system performance metrics for shared memory, message-based, and hybrid computing systems. Like all benchmarks, it has strengths and weaknesses in what and how it measures, but by and large it is the most widely used measurement of HPC performance today.

-Use LINPACK to measure the performance of GNU's matrix library routines and compare them to the performance of either Intel's Math Kernel Library (MKL) or Goto's BLAS library with the same workload: in this case, the input file below.

Resources:
-LINPACK: Download/Main site
-Input File (Ignore the request to login)
Show solution
©1994-2014   |   Shodor   |   Privacy Policy   |   NSDL   |   XSEDE   |   Blue Waters   |   ACM SIGHPC   |   facebook  |   facebook   |   twitter   |   rss   |   youtube Not Logged In. Login