Skip to content

Comment on NextSilicon reveals new processor chip in challenge to Intel, AMDparent

Comments

It's a bit more complicated, you need to use their compiler (LVVM fork with clang+fortran). This in itself is not that special as most accelerators (ICC, nvcc, aoc) already require this.

Modifications are likely on the level of: Does this clang support my required c++ version? Actual work is only required when you want to bring something else, like Rust (AFAIK not supported).

However, to analyze the efficiency of the code and how it is interpreted by the card you need their special toolchain. Debugging also becomes less convenient.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.