Get the latest tech news
Glue and Coprocessor Architectures
Special thanks to Justin Drake, Georgios Konstantopoulos, Andrej Karpathy, Michael Gao, Tarun Chitra and various Flashbots contributors for feedback and review. If you analyze any resource-intensive computation being done in the modern world in even a medium amount of detail, one feature that you will find again and again is that the computation can be broken up into two parts: These two forms of computation are best handled in different ways: the former, with an architecture that may have lower efficiency but needs to have very high generality, and the latter, with an architecture that may have lower generality, but needs to have very high efficiency.
2024 Sep 02 See all posts Glue and coprocessor architectures Special thanks to Justin Drake, Georgios Konstantopoulos, Andrej Karpathy, Michael Gao, Tarun Chitra and various Flashbots contributors for feedback and review. The high-level business logic code is written in Python, a highly general and flexible language which is also very slow, and we just accept the inefficiency because it only touches a small part of the total computational cost. This is most obvious in cryptography, where it's clear what kinds of specific expensive operations are most likely to be used: modular arithmetic, elliptic curve linear combinations(aka multi-scalar multiplications), Fast Fourier transforms, and so on.
Or read this on Hacker News
