So I did re-run the benchmarks from @lemire upstream simdjson on the same box as our @rustlang port ran here some results (original vs. rust, tape parsing not DOM in 1.789 GB/s):
twitter: 1,789 s vs 1.740
canada: 0.957 vs 0.780
citm_catalog: 2.258 vs 1.980
1/2
-
-
So the original implementation is still faster but the gap is closing - it should also be noted that the benchmarks, while using the same input data are different too so some variation could come from that - as always take them with a grain of salt!
1 reply 0 retweets 2 likesShow this thread -
-
Thanks! And yes, I was really happy that it’s getting closer - there is still some tax in where we do things differently to allow turning it into a dome easier and some places were we pay a rust-tax but it’s getting fairly close :)
1 reply 0 retweets 1 like
Until we start cheating using unfair contextual knowledge 
10:07 AM - 7 Nov 2019
0 replies
0 retweets
1 like
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.