An integral aspect of the move toward exascale computing and systems 50 times more power than today’s is the advancement of tools that will allow results to be seen while a simulation runs—a process called in-situ visualization and analysis. The key benefit of this process is that it allows input/output bottlenecks to be circumvented.
Many computing community tools allow for in-situ visualization and analysis, but a product called ALPINE Ascent is the first one meant for next-generation supercomputers, said Cyrus Harrison of Lawrence Livermore National Laboratory (LLNL) and the US Department of Energy’s Exascale Computing Project (ECP) during an interview at the SC18 supercomputing conference in Dallas recently. ALPINE Ascent is referred to as a flyweight in-situ visualization and analysis library for ECP applications.
Joined by colleagues Eric Brugger and Matt Larsen, also of LLNL and the ECP ALPINE project team, for the interview, Harrison explained that the Ascent software embedded with a simulation code has, in running on more than 16,000 GPUs on the Sierra supercomputer at LLNL, achieved a significant milestone toward preparation of all of the ECP codes and software technologies for execution on heterogeneous systems, which, like Sierra, have both CPUs and GPUs.
“This was the first large-scale demonstration of using of some of this technology very, very successfully on GPUs, so we’re very excited about that, and, hopefully, that is the path that we will be able to continue, using this software to exascale,” Harrison said.
Larsen, who contributes to the development of the Ascent infrastructure and is involved in several other ECP projects, said the aim of Ascent is to create a next-generation in-situ visualization structure that is built from the ground up to use the node-level parallelism that is available on systems like Sierra.
He noted that Ascent can leverage the capabilities made possible by another ECP project, VTK-m (Visualization Toolkit-multi-cores), which allows exploitation of the node-level parallelism of GPUs and computers with multi-core or manycore processors such that operations can be executed directly on the same architectures on which the simulations are running.
Brugger explained that the VTK-m project is distinguished from the pre-existing effort called VTK in that it is completely new and separate, with the goal of providing an abstraction that will run across different manycore architectures. He added that the infrastructure that he and his team members are generating is not only available in Ascent but also in other in-situ visualization tools such as VisIt and ParaView.
Harrison said seeing Ascent work so well out of the box on Sierra was gratifying for the team and represents the fruition of several years of preparation.
“We’re really excited about using these tools on pre-exascale and exascale systems,” Harrison said, adding that the work has been a lot of fun and more challenges remain.
Video Chat Notes
A major milestone: running on more than 16,000 GPUs on Sierra [0:00:57]
ECP’s ALPINE project for in-memory computing [0:02:17]
ALPINE Ascent software infrastructure to use node-level parallelism [0:03:43]
Leveraging the VTK-m project to exploit node-level parallelism [0:04:13]
Building an infrastructure applicable to other efforts [0:05:23]
Years of preparation have paid off [0:06:35]
ECP presents immense opportunity for rich collaborative interactions [0:07:12]