Implement AVX vectorized exponentiation

[xflouris]

Implement _mm256_exp_pd() for computing transition probability matrices (Issue #105 )

The intrinsic does not map to an instruction and is only available in the Intel Small Vector Library.

There exists a single-precision implementation, however it seems to differ a bit (in results) from the libc implementation, at least for some corner cases.

A good double-precision non-vectorized implementation seems to be in the old cephes library.

[stamatak]

I think the Intel MKL library also offers a fast implementation, also
there is this old paper with Nikos Alachiotis where we looked at SW
implementations of exp:

http://link.springer.com/chapter/10.1007%2F978-3-642-19475-7_34

alexis

On 05.09.2016 10:31, Tomas Flouri wrote:

Implement |_mm256_exp_pd()| for computing transition probability
matrices (Issue #105 #105 )

The intrinsic does not map to an instruction and is only available in
the Intel Small Vector Library
https://software.intel.com/en-us/node/583228.

There exists a single-precision implementation
http://software-lisc.fbk.eu/avx_mathfun/avx_mathfun.h, however it
seems to differ a bit (in results) from the libc implementation, at
least for some corner cases.

A good double-precision non-vectorized implementation seems to be in the
old cephes library.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#111, or mute the thread
https://github.com/notifications/unsubscribe-auth/AA1w-nJhcM7Pi8ByiekwsxpxsCVnpnbMks5qm9NrgaJpZM4J02cK.

Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
Adjunct Professor, Dept. of Ecology and Evolutionary Biology, University
of Arizona at Tucson

www.exelixis-lab.org