Hi,
I have been developing FHE applications with TFHE-rs, mainly with the (high level) C APIs. Recently, I wish to deploy the application on a server equipped with GPU.
Is GPU supported in conjunction with C API? I have searched the generated tfhe.h
file but it seems not to contain the relevant interfaces (e.g. decompress_to_gpu).
I would really appreciate any help regarding this.
The function compressed_server_key_decompress_to_gpu
(example) exists but only if you have compiled the CAPI with the gpu
feature
If you are relying on our Makefile to build, you can use the make build_c_api_gpu
to build a GPU enabled C API
Thank you for your (extremely) prompt reply. After adding --features=high-level-c-api,gpu
and specifying the location of CUDA libraries in link-args, the generated tfhe.h header successfully contains the required interfaces. Thanks!