[ADOL-C] ADOL-C on a GPU

[Andrea Walther]

Hi,

> The problem is not in exploiting the sparsity (at least for now),
> but just in computing the darivatives of the same function at many different points (more than 500).

For this case, there are at least two possibilities

* Either one can distribute the function evaluation on  several cores
  using the ADOL-C version as it is now. For this purpose, one might
  use OpenMP. The OpenMP example contained in the ADOL-C distribution
  does exactly this.

* Alternatively, one might use GPUs. I have serious doubts that this
  will work with the current ADOL-C version. The starting point that
  comes closest to that is the tapeless variant of ADOL-C

  I will discuss this issue with a student who currently extends this
  tapeless variant to higher order derivatives whether we can also
  integrate a version for GPUs here.

So as short term solution one might use OpenMP, as long term solution
we aim at GPUs in these cases.

Best regards,

Andrea

-- 
Prof. Dr. Andrea Walther
Lehrstuhl fuer Mathematik und ihre Anwendungen
Institut fuer Mathematik
Universitaet Paderborn
Warburger Str. 100
33098 Paderborn

Email: andrea.walther at uni-paderborn.de
Phone: ++49 5251 602721
Fax:   ++49 5251 603728

**********