Vectorizing Unicode conversions on real RISC-V hardware
In this article we'll discuss how to achieve the speedup below for UTF-8 to UTF-16 conversion, using the RISC-V Vector extension.
I excluded the plain ASCII case from the graph above because it made the other results less readable, the speedup was: 8x for C908, and 11x for C920. More comprehensive measurements are at the end of the article.
Return to parent page
View source code
Introduction
The vast majority of text you'll come across will be encoded in the UTF-8 format, but some languages and ...
Read more at camel-cdr.github.io