Culises: A Library for Accelerated CFD on Hybrid GPU-CPU Systems
Culises: A Library for Accelerated CFD on Hybrid GPU-CPU Systems
Culises: A Library for Accelerated CFD on Hybrid GPU-CPU Systems
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<str<strong>on</strong>g>Library</str<strong>on</strong>g> <str<strong>on</strong>g>Culises</str<strong>on</strong>g>Schematic overviewOpenFOAM® (1.7.1/2.0.1/2.1.0)MPI-parallelized <strong>CPU</strong> implementati<strong>on</strong>based <strong>on</strong> domain decompositi<strong>on</strong><str<strong>on</strong>g>Culises</str<strong>on</strong>g>:Solves linear system(s)<strong>on</strong> multiple <strong>GPU</strong>sOpenFOAM:<strong>CPU</strong>0<strong>CPU</strong>1<strong>CPU</strong>2linear system Ax=bprocessor partiti<strong>on</strong>ingsoluti<strong>on</strong> xInterface:cudaMemcpy(….cudaMemcpyHostToDevice)cudaMemcpy(….cudaMemcpyDeviceToHost)<str<strong>on</strong>g>Culises</str<strong>on</strong>g>:PCGPBiCGAMGPCG<strong>GPU</strong>0<strong>GPU</strong>1<strong>GPU</strong>2MPI-parallel assembly ofsystem matrices remains <strong>on</strong> <strong>CPU</strong>sSlide 7<str<strong>on</strong>g>Culises</str<strong>on</strong>g> - A <str<strong>on</strong>g>Library</str<strong>on</strong>g> <str<strong>on</strong>g>for</str<strong>on</strong>g> <str<strong>on</strong>g>Accelerated</str<strong>on</strong>g> <str<strong>on</strong>g>CFD</str<strong>on</strong>g> <strong>on</strong> <strong>Hybrid</strong> <strong>GPU</strong>-<strong>CPU</strong> <strong>Systems</strong>B. Landmann