[ADOL-C] Hessian Matrix with Cuda

[宗宝冯]

Hi, Authors of ADOL-C,

I am a developer of baidu apollo project. when I used ADOL-C in apollo, NLP
solve spend most time, So, we want to speed up by CUDA. This is
background,  I have two questions about ADPL-C with CUDA:

1. There is an example  using cuda kernel named kernel in file
https://gitlab.com/adol-c/adol-c/blob/cuda/ADOL-C/examples/additional_examples/cuda/traceless_cuda.cu,
would this kernel function can replaced by sparse_jac ? is this you point ?
2. If I want replace sparse_hess, do I need create a new kernel function
like get Jacobian matrix ? I have my function evaluation.


Thank you all.



-- 
fzb.me
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/adol-c/attachments/20200109/22a58f69/attachment.html>