Circle STARK GPU Acceleration: An Analysis of Performance and Implementation

The session will cover the overview of GPU acceleration for ZK-SNARK/STARK proof systems, focusing on the GPU implementation of Circle STARK. It will address challenges and solutions in implementation, delving into parallelizing computations in the FRI protocol, quotient operations, polynomial operations, and batching. We'll compare CPU and GPU performance, discussing available technologies (SIMD, AVX512, CUDA). Finally, we'll explore future directions for hardware acceleration (ie. FPGA).