Hello. I currently have developed a Json parsing/serializing library called Jsonifier. It used to use simdjson's core algorithm for parsing but where I re-saturated the avx/neon registers upon initial collection of the index values instead of only operating on 64-bytes worth at a time (As can be seen here) and this seemed to accelerate the algorithm by roughly 30%. I have since however reduced the usage of avx/arm-simd instructions to only being used for the ptr-to-data-member hashmaps and string parsing/serializing, as well as string comparisons/char-searches. What I would like to request is pointers with respect to utilizing your library as efficiently as possible in my own benchmarks as I believe that may not currently be the case.
Here is the performance repo:
https://github.com/RealTimeChris/Json-Performance
Hope this reaches you well! Cheers!