-
I was looking through `compute_local_error` to see how the distributed version was implemented.
I have a question or suggestion for the following lines
https://github.com/dealii/dealii/blob/41e71ffb…
-
The relevant state-of-the-art projects and publications should be explored in order to get inspiration and come up with good ideas. Below is the list of things to start exploration with. The outcomes …
-
Planned improvements for LinearAlgebra module.
## Features
- [ ] Distributed support
- [ ] GPU support via cuBLAS / clBLAS
- [ ] Feature completeness of dense local linear algebra (https://githu…
-
JAX already has excellent parallelism support via `jax.Array`, the compiler could choose sharding so that maximum parallelize for elementwise operations and matrix multiplication. But the linear algeb…
-
Hello,
I am not sure about the underlying multi GPU concept of amgX.
The application I have in mind has the following features:
- distributed memory CFD domain decomposition using scotch
- M…
-
For finite element codes (ultimately based on T8code.jl) we would like to try incorporating t8code.
Naturally the results would be point/vertex based and not element based in most cases.
The solu…
-
# Recent radical innovation
There has been some radical innovation in graph processing the last few years
GraphBLAS efficiently represents and operates on graphs as sparse matrices. It provides …
-
This is a meta-issue tracking the feature completeness dense local linear algebra. Implementations can be wrapped (BLAS/LAPACK) or native for this checklist. We are using [`numpy.linalg`](https://nump…
-
Hello,
I stumbled upon this package and was intrigued by its implementation of the abstract matrix interface.
What are the implications of this? Does this mean a blob backed by an arbitrary out of c…
-
## 🐛 Bug
Using `torch.det()` inside `nn.DataParallel` in a multi gpu environment (tested 4) is suffering from a race condition. If the execution is done sequentially, the error doesn't trigger.
…