Running Advanced Life-Cycle Models with CPU

1

Dear Rober,

Your VFI Toolkit is excellent!

I am especially interested in life-cycle models with housing. However, I just started at the beginning. I noticed that only Nvidia GPUs (with CUDA drivers) work well as GPUs in Matlab. I use Intel GPU, so I cannot use GPU here to solve problems faster. I have a few concerns regarding these:

  • To plot life-cycle profile, function jequaloneDist=zeros([…],‘gpuArray’) require GPU. How can I overcome the GPU requirement of this function with my CPU?

  • Would I encounter many problems, since I do not use GPU, till I reach model 35, which I plan to modify finally?

  • Can I even run a model similar to 35 with a CPU? How long would it take to solve a value function for such a model?

Could you please provide me with some advice? Thank you very much.

1 Like
2

Sorry but without GPU the toolkit is rather limited
http://discourse.vfitoolkit.com/t/what-can-i-run-without-a-gpu-with-just-cpu/

This is because the toolkit uses an algorithm (pure discretization) which is easy and robust, but is very very slow on CPU, decent runtimes on GPU. Hence for more advanced features it is just not worthwhile coding them except for GPU as they would be un-usably slow.

[There is also the ‘divide and conquer’ algorithm now in some parts of toolkit, but the way it is implemented is also specifically just for GPU]

1 Like
3

Thank you for your response. I am looking for solutions to upgrade my computer hardware right now. I hope to come back very soon.

1 Like