Reason for abandonment

[jon-chuang]

Did this project lose steam due to the rapid iteration of zk-SNARK schemes that convinced the development team that it is much wiser to iterate on CPU-based prototypes rather than park effort in a more complicated GPU implementation whose underlying scheme might become obsolete quickly?

Are there not some shared primitives to zk-SNARKs which one might reasonably believe would stand the test of time and hence may deserve to be accelerated on a GPU, so that future schemes could build atop of them?

[shamatar]

Hey! Partially it was due to the fact while rented GPU is faster in terms of time it happened to be as economically efficient (time * cost) as CPU prover. This fact along with a newer proof systems that need to be evaluated is a reason why GPU provers is not investigated on a scale. Nevertheless there is a "belle_cuda" repo with a complete Groth16 Cuda prover.

…

On Thu, Feb 6, 2020 at 4:34 PM jon-chuang ***@***.***> wrote: Did this project lose steam due to the rapid iteration of zk-SNARK schemes that convinced the development team that it is much wiser to iterate on CPU-based prototypes rather than park effort in a more complicated GPU implementation whose underlying scheme might become obsolete quickly? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABRK7535SVHS4W4ZUW2SJHTRBQGW7ANCNFSM4KQ47MBA> .

[jon-chuang]

Hi, I see. In theory then, GPUs, especially if you own a fleet, could still provide massive speedups and cost savings? Asking with a view towards future developments.