Cuda Big Int?

Does anyone know of a BigInt port to cuda? I have seen some attempts, but nothing that was finished. Wanted to make sure nothing was out there before starting a port myself.


here’s some student work done for a project. Certainly not a feature complete library, but possibly rather useful for crypto and prime numbers:

and CUMP (which is floating point multiprocision arithmetics, but it needs an underlying Integer core to work)

I have this lib 99% ported. Most all functions are ported. I have used then in host code to ensure I have not broke anything during the port. I have not tested in GPU yet. If anyone wants to join in on the efforts, I will be glad to send the current code. I had plans to submit the final code to the org author in hope that he will put it on his page so that others can use it (hoping google will find it).


If you could publish your preliminary work on some site like github, I’d be grateful.
Maybe I can even contribute a few things here and there.


Msg me an email address and I will zip the project and send to you. I don’t want to publish anything just yet because it would not be good for someone to try and use it and not get the correct results due to a porting error. Some things I have noticed that are going to hold it back in CUDA is that an IF/ELSE in 1 thread will delay the execution in another until they are re-aligned with each other. At least from what I understand about how CUDA executes.