Skip to content

Releases: MikaelSlevinsky/FastTransforms

Remove Windows mklinks

05 Nov 23:15
20834eb

Choose a tag to compare

Remove Windows mklinks (#39)

* can i just integrate install?

* add include and bin flags

* conflicting includes?

* include openblas first?

* find openblas.dll

* integrate necessary?

Define ALIGN_SIZE backup

22 Oct 20:51

Choose a tag to compare

define ALIGN_SIZE backup

in case not even using sse2, which would be the case on a 32-bit arch

Add compiler flags for Mojave

11 Sep 02:38

Choose a tag to compare

/usr/local is no longer part of macOS' default search path.

Modernize makefile

03 Sep 04:06

Choose a tag to compare

v0.2.6

add a point

Add gcc 8 & 9

30 Aug 20:40

Choose a tag to compare

v0.2.5

test up to gcc-9

Expanded GCC versions in GitHub releases

29 Aug 03:31

Choose a tag to compare

Trial support of Travis GitHub releases uploading (continued)

24 Aug 20:49

Choose a tag to compare

v0.2.3

${TRAVIS_BUILD_DIR} appears to be the right place

Trial support of Travis GitHub releases uploading

24 Aug 20:22

Choose a tag to compare

v0.2.2

one deploy, two providers

New example for nonlocal diffusion on the sphere

22 Aug 18:30

Choose a tag to compare

Add ft_mpfr_trsv and ft_mpfr_trsm, tidy up some mpfr routines, make mpfr optional through the use of -DFT_USE_MPFR, simplify test scripts.

Fast 1D orthogonal polynomial transforms

19 Aug 17:27
9ee8818

Choose a tag to compare

This release adds fast solutions to classical orthogonal polynomial connection problems via divide and conquer for triangular banded generalized eigenproblems accelerated by the FMM. Dense (direct) connection coefficients are supported in multi-precision via mpfr. See the documentation starting here for a full list of new features and also the pull request #37.

Nothing has changed for exported multivariate transforms, but divide and conquer for symmetric definite generalized eigenproblems can also be used to accelerate the multivariate cases, see https://arxiv.org/abs/1711.07866, and most of the technology is in place. Subtleties regarding the rootfinding technique (passing tests on all platforms) will delay these to a subsequent release.

@dlfivefifty @ajt60gaibb @ahbarnett @kburns @gradywright