On Mon, Apr 14, 2025 at 7:47 PM Mark Adams <mfad...@lbl.gov> wrote:
> On Mon, Apr 14, 2025 at 5:39 PM Angus, Justin Ray via petsc-dev < > petsc-dev@mcs.anl.gov> wrote: > >> Hello, >> >> >> >> A project I work on uses GMRES via PETSc. In particular, we have had good >> successes using the Additive Schwarz Method + ILU preconditioner setup >> using a CPU-based code. I found online where it is stated that “Parts of >> most preconditioners run directly on the GPU” ( >> https://urldefense.us/v3/__https://petsc.org/release/faq/__;!!G_uCfscf7eWS!fm7OZhr_b9AIu2cEVb2FddEGQTnpwos8zXtgJpvMdCoX2pY6yVCfLsM6iFOQGNEkMjTrRFzTiFnKWR9rBWF8$ >> >> <https://urldefense.us/v3/__https://petsc.org/release/faq/__;!!G_uCfscf7eWS!bw6qeKcY7MKSvlEgcogdKR7fpjZSOFvka6zfDprUZ_sJHdE-YZmRD6UTqWQW3_uGVBII4P-AG0zaGTLbI67_fQ$>). >> Is ASM + ILU also available for GPU platforms? >> > > Yes, but it will be hard to get speedup from CPUs. > It is easy to test though. Jacobi or pbjacobi if you have a vector > equation is probably faster. > I assume you mean ILU(0), which is what PETSc supports. Computing it would not be fast, but applying it is a MatMult() which is not bad. If you take lots of iterates, then it will speedup like the bandwidth ratio. You could even compute it on the CPU and apply it on the GPU, which would be good for an initial test. Thanks, Matt > Mark > > >> >> >> -Justin >> > -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://urldefense.us/v3/__https://www.cse.buffalo.edu/*knepley/__;fg!!G_uCfscf7eWS!fm7OZhr_b9AIu2cEVb2FddEGQTnpwos8zXtgJpvMdCoX2pY6yVCfLsM6iFOQGNEkMjTrRFzTiFnKWckmgRh1$ <https://urldefense.us/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!G_uCfscf7eWS!fm7OZhr_b9AIu2cEVb2FddEGQTnpwos8zXtgJpvMdCoX2pY6yVCfLsM6iFOQGNEkMjTrRFzTiFnKWVdtLfo4$ >
