Argonne National Laboratory

Science Highlights

Date Postedsort ascending
Photo courtesy of David Gleich (Purdue)
Grey is better than black

Stefan Wild argues that a black-box approach to optimization can be replaced by using families of local surrogates to infer information.

October 13, 2016
Workshop to be held on FPGAs

A workshop on FPGAS for scientific simulation and data analytics will take place October 12—13, 2016, at the National Center for Supercomputing Applications in Urbana, Illinois.

October 5, 2016
Danger Ahead: Silent Data Corruption

Two MCS Division researchers have teamed up to create a novel adaptive impact-driven detector designed to protect a large set of high-performance computing applications from influential SDCs.

October 3, 2016
Elia Merzari, a nuclear engineer at Argonne, received the George Westinghouse silver medal for his leadership and contributions to the power field.
Argonne engineer receives prestigious medal

Argonne nuclear engineer Elia Merzari has been awarded the George Westinghouse Silver Medal from the American Society of Mechanical Engineers.

August 18, 2016
Locking and arbitration in multithreaded MPI implementations

A team of researchers at Argonne National Laboratory, the Chinese Academy of Sciences, Intel, and the Tokyo Institute of Technology now have devised an approach that uses arbitration to favor execution paths that contribute better to the overall progress.

August 18, 2016
NekCEM/Nek5000 named R&D 100 finalist

The project NekCEM/Nek5000: Scalable High-Order Simulation Codes has been named a finalist in the 2016 the R&D 100 Awards, which are given by R&D Magazine to its top 100 new technologies for the year.

August 11, 2016
Modeling a million-node Slim Fly network

Researchers at Rensselaer Polytechnic Institute and Argonne National Laboratory have now developed an efficient and detailed model of the Slim Fly network using massively parallel discrete-event simulation.

August 2, 2016
Thirteen student participants received travel grants to attend the PETSc 2016 User Meeting.
Meeting highlights exciting work being done by the PETSc user community

The PETSc 2016 User Meeting kicked off June 28 in Vienna, Austria, with a daylong tutorial, followed by two days of plenary talks and poster sessions. PETSc, the Portable, Extensible Toolkit for Scientific Computing developed at the U.S. Department of Energy’s Argonne National Laboratory and collaborating institutions, is one of the world's most widely used software libraries for high-performance computational science.

July 25, 2016
Cappello gives tutorial on resilience for HPC

Franck Cappello co-presented a tutorial July 1 on resilience for high-performance computing at the 46th annual IEEE/IFIP international conference on Dependable Systems and Networks (DSN 2016).

July 11, 2016
I/O benchmark time reported by a 4,800-process IOR file per process
experiment that issued nearly 10 million distinct I/O operations on
NERSC's Edison Cray XC30 system. The experiment was repeated 15 times for each configuration to confirm that Darshan's impact onapplication performance is negligible.
Evaluating the New Modularized Version of Darshan on the Cray XC30

Darshan is a software tool designed and developed by a team of researchers at Argonne’s Mathematics and Computer Science (MCS) Division and widely used to gather I/O workload statistics from high-performance computing applications. Darshan’s original design goals were threefold: scalability on leadership-class computing systems, transparency to end users, and accuracy in the I/O characterization of applications. In 2015, the Darshan team – comprising Phil Carns, Shane Snyder, and Rob Ross – decided to add a fourth goal: modularization.

June 23, 2016