Simone Silvestri

Inconsistent behavior with multiple GPU

6

Hello! I am recently trying to exploit the single-process multi GPU capabilities of CUDA.jl 3.8.0 to run on multiple GPU. Unfortunately, for some weird reason, my kernel fails with `ERROR:...

Remove `Δt` from solvers

1

cc @glwagner

`ConstantCartesianCoriolis` test dominates GPU testing wall time

3

the `inertial_oscillations_work_with_rotation_in_different_axis` (in `test_dynamics.jl`) takes about 1.5 hours to complete on a GPU, despite testing only slanted coriolis force. It would be probably better to skip it or design a...

Add Multi GPU testing

3

with the development of multi GPU we have to expand our test capabilities to include this functionality. We should introduce a new Agent that can handle single-node and multi-node multi...

[WIP] Extend MultiRegion to NonhydrostaticModel

10

Now that finally `MultiRegion` is merged we can implement the single node multi GPU paradigm also in the Nonhydrostatic model cc @tomchor The work can be divided in three tasks...

This PR introduces the idea expressed in issue #2509 as well as introducing two unique buffers for _prognostic_ and _diagnostic_ fields which allow the parallelization of memory transfer

Tests for `OffsetArrays-Field` broadcasts

1

with @glwagner we introduced broadcasts across fields and `OffsetArrays` that include halos, so that ``` field1 = CenterField(grid) field2 = CenterField(grid) set!(field1, (x, y, z) -> x) data = field1.data...

BC and Immersed Reductions for Sliced Fields

4

This PR is the successor of #2723 The overall objective is to improve the support of `Sliced` and `Windowed` fields such that we can represent the Free surface as a...

Dynamic parallelism

1

I am trying to set up a dynamic kernel wherein a KA kernel launches a CUDA kernel. The final objective would be to have dynamic parallelism using only kernel abstractions....