@predsci 's near-real-time data-driven simulation of the solar corona and heliosphere used to predict the 2026 total eclipse is nominated for an HPCWire Reader's Choice Award for "Top HPC/AI-Enabled Scientific Achievement"
Check out the simulation here:
eclipse.predsci.com
and feel free to vote here!
hpcwire.com/2026-hpcwire-rea…
Check out our open source Magnetohydrodynamic Algorithm outside a Sphere (MAS) code!
github.com/predsci/mas
It uses Fortran standard parallelism (with a sprinkle of OpenACC) to run on many GPUs in parallel with NVIDIA's HPC SDK compiler!
Check out its use to model the 2024 total solar eclipse: predsci.com/eclipse2024@ctnzr#NVIDIAGTC
At @NVIDIAGTC#GTC25 - if you are here, feel free to come by the poster session Monday night where I will be presenting results of using @fortranlang#Fortran standard parallelism on the GH200 Grace-Hopper Superchip! @predsci
Timing results for a small run of the HipFT code (github.com/predsci/hipft) on various CPU and GPUs. The code uses standard #Fortran@fortranlang "do
concurrent" to run in parallel on CPUs and offload to GPUs. For NVIDIA GPUs @NVIDIAHPCDev , unified memory is used, while for Intel GPUs, the use of @OpenMP_ARB Target directives are needed for data movement only.
A neat result here is the use of the new $250 Intel Arc B580 GPU @intelhpc@IntelDevTools@IntelGraphics with its FP64 cores, performing right near where it should based on its memory bandwidth (HipFT is memory bandwidth bound).
Our new workstation: NVIDIA A100 80GB PCIe in a Desktop Ryzen 7 9700X PC, custom liquid-cooled. Shout-out to bizon-tech.com for making this work! @NVIDIAHPCDev