cufinufft

cufinufft copied to clipboard

Added the precomputed sigma=1.25 kernel options for #136

AaronGhost

Rapidly degrading adjoint test in 3D

4

Dear all, I was testing `cufinufft` via the Python wrapper on some 3d problems, and I stumbled upon some accuracy issues when performing the adjoint test. That is, given the...

grizzuti

Per @ahbarnett in #126: > Any Py-based timing tests, that match a C++-based test like Melody just did, would be amazing to add to the Py tests dir, and would...

janden

Move from manylinux2010 to manylinux2014

6

As discussed in #121, the new version of NumPy only releases binary wheels for manylinux2014, while our CI (and our releases) are for manylinux2010 for now. This is currently resolved...

janden

Type 2 3D transform gives different result from (non-cu) finufft for some inputs

8

This is a bit of a strange error, so I will need to unpack a little here: For some inputs, the type 2 3D NUFFT implemented in cufinufft is giving...

JBlaschke

Documentation

3

How far do we want to go with this? I don't know if it's worth it to have a full-fledged site lite for Finufft, but something more substantial (like a...

janden

Idea for SM in double-precision 3D

2

So if I've understood correctly, the reason we can't perform SM in 3D with doubles (at high precision) is that the 3D box won't fit. In other words, for 12...

janden

As it stands now, the way we are setting (and restoring) the gpu device id, we are always clobbering any local device contexts with the primary context. This is because...

JBlaschke

Julia Package or Julia Compatibility?

3

Is it possible to use cufinufft in Julia at the moment? If not, will a Julia package be available in the future?

betttax

Multi Device Support

2

Longer term, I think it would be good if we could naively use multi gpu in an embarrassingly parallel way. That is, cufinufft keeps the responsibility of managing device choice...

garrettwrong