Fluid Numerics

In this livestream, Joe will review some of the self-imposed design criteria for the Spectral Element Library in Fortran. Specifically, we'll focus on the desire to provide default CPU-only builds of SELF with the option to build for AMD or Nvidia GPUs.

Draw a domain in inkscape and run a fluid simulation from that. Seriously. We're doing it. Check out our live stream this afternoon (at 3pm ET) to learn how Garrett made this happen. The livestream gives you a good opportunity to ask questions and interact with us. But don't worry. If you can't make it, the video will be up afterwards for you to watch at your leisure. https://lnkd.in/eb4njAhu

00:52 GPU accelerated platforms overview 03:52 Estimated performance of GPU accelerated platform components 06:25 Starting the codelab and introduction to Oram 08:00 Overview of the Fortran application 08:55 Overview of the porting process 09:45...

Before the year is out, learn with us. Get access to our Nvidia V100, AMD MI210 and AMD MI300A systems and learn how to run Fortran applications on GPUs, independent of the hardware vendor. Our three day course comes with one week of compute access plus expert support from both Fluid Numerics and Pelagos Consulting & Education. https://lnkd.in/eDA_aUeg

In this livestream, Joe will share some of our latest work on finding and optimal implementation for vector divergence in 2-D and 3-D. Specifically, we'll demonstrate how a hand-written HIP kernel that takes advantage of shared memory and the...

00:00 2km Resolution at 500m Depth 01:06 2km Resolution at 750m Depth 02:11 2km Resolution at 1000m Depth 03:16 1km Resolution at 500m Depth 04:13 1km Resolution at 750m Depth 05:08 1km Resolution at 1000m Depth Topographic Dynamics of the Gulf Stream...

In this video, we draw a model domain in Inkscape and convert the SVG to a spectral element mesh that can be ingested in the Spectral Element Library in Fortran. In the midst of convergence studies and more serious work, it's good to have some fun....

In this livestream, Joe will provide an overview of GPU programming with HIP and HIPFort. We will then walk through porting a simple stencil code in Fortran to the GPU using HIPFort. We'll compare runtimes with the original CPU-only code and with...

Break the vendor lock. Work with us on our dev cluster to rapidly port and optimize your code to AMD GPUs, while maintaining portability with Nvidia GPUs. When you are ready to scale up, we will help you get up and running on Hot Aisle Inc. , where you will have access to MI300X GPUs hosted in a secure and reliable data center. https://lnkd.in/ebizZCh7

00:00 The Helmholtz Decomposition 01:19 Boundary Conditions for the Helmholtz potentials 09:22 Statement of the boundary value problems for the Helmholtz potentials 11:25 Relating the Helmholtz potentials to kinetic energy 16:20 Eliminating the...