[Feature Request] CUDA hardware acceleration integration

[z4rathustr4]

Hi there, this is my first post and I don't know neither if it's been proposed by anyone nor if it's possible with the current codebase, but here it goes:

• Is it possible to adapt the core codebase to enable CUDA hardware acceleration?
  I can imagine it would be a terrible nightmare to update all class modules and can't imagine the complexity of modifying the core for this purpose, but it has good potential when running very large simulations (e. g. Dungeon Slice simulations, long APLs, multiple stat scaling, also the new talent trees, which, for example Mage class tree has at a first glance a cyclomatic complexity of 15, but going into @seriallos Talent Possibilities excel, the raw count of iterations ascend to 39,483,984... not to mention DK's class nodes 😣)

Some resources:

• Nvidia Developer Homepage
• Nvidia Dev Forums - Accelerated Computing - GPU Accelerated Libraries Category
• Talent Possibility Excel (credit to @seriallos obviously)

I hope it inspires the devs to add a new layer of possibilities to this wonderful project!

[scamille]

Please have a look at some of the discussion in #2119 . I don't think the assessment there has changed much - simc isn't really well suited for the parallelization that GPUs offer.

[z4rathustr4]

Thanks for the swift answer, I did not see that issue, I'll look at the implementation example linked there and look if I can do something better with it, just for fun 😄 .

Happy weekend.