Single precision implementation of Phylogenetic Likelihood Kernel (newviewGTRGAMMA) in standard-RAxML

[Geert Roks]

Hello all,

In the paper of Simon Berger and Alexandros Stamatakis from 2009 (https://link.springer.com/chapter/10.1007/978-3-642-14403-5_29), they discussed the performance of a single-precission implementation of RAxML. This implementation should have been added since v7.2.3, however in the current version v8.2.x, I am unable to find such a single-precision implementation or any mention of it except for this paper. Also, when trying to look for version v7.2.3, I can't find it, since the github tags only go back to v7.7.1.

Has this SP implementation been deprecated? Or does it have a different name, other than `newviewGTRGAMMA`? Or is it in a different repo?
Could someone point me to the source code of the single-precision implementation of the Phylogenetic Likelihood Kernel (PLK)?

Kind regards,
Geert

[Alexandros Stamatakis]

Dear Geert,

We basically completely abandoned any further attempts with single
precision as it was too numerically unstable (especially all the single
parameter optimization routines had substantial convergence issues) to
be of any use.

I managed to dig out a v7.2.3. tarball and attach it here.

Alexis

On 08.04.24 15:13, Geert Roks wrote:
> Hello all,
>
> In the paper of Simon Berger and Alexandros Stamatakis from 2009
> (https://link.springer.com/chapter/10.1007/978-3-642-14403-5_29

> <https://link.springer.com/chapter/10.1007/978-3-642-14403-5_29>), they

> discussed the performance of a single-precission implementation of
> RAxML. This implementation should have been added since v7.2.3, however
> in the current version v8.2.x, I am unable to find such a
> single-precision implementation or any mention of it except for this
> paper. Also, when trying to look for version v7.2.3, I can't find it,
> since the github tags only go back to v7.7.1.
>
> Has this SP implementation been deprecated? Or does it have a different
> name, other than `newviewGTRGAMMA`? Or is it in a different repo?
> Could someone point me to the source code of the single-precision
> implementation of the Phylogenetic Likelihood Kernel (PLK)?
>
> Kind regards,
> Geert
>

> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/raxml/af637565-b5e2-411e-a2ff-26084729b610n%40googlegroups.com <https://groups.google.com/d/msgid/raxml/af637565-b5e2-411e-a2ff-26084729b610n%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Alexandros (Alexis) Stamatakis

ERA Chair, Institute of Computer Science, Foundation for Research and
Technology - Hellas
Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology

www.biocomp.gr (Crete lab)
www.exelixis-lab.org (Heidelberg lab)