On Wed, Nov 20, 2019 at 11:10 PM Julien Blanc
Some additional remarks: - you don't specify the hardware you ran the benchmark on. Additionaly, a test on an arm architecture, like a raspberry pi, would be nice.
I listed my hardware on the page, and I re-collated the results to group them by input file to make comparisons easier. I don't have a Raspberry PI, so if someone wants to run the bench program and report the results that would be lovely.
- could you provide a link with the source code for the benchmarks, as well as the test files ?
Yep, the bench program and input files are here: https://github.com/vinniefalco/json/tree/develop/bench
A 200MB file would be a good start to see how the library deals with large files.
I don't have a 200MB JSON file, and I have doubts that it would provide much more information than what we have already. I'd also prefer not to put a 200MB file into the repository. The goal of this library is to be in the same ballpark as rapidjson, it doesn't try to be the top performer (in fact, rapidjson achieves its performance by making usability tradeoffs which are unsuited for vocabulary types). However if someone wants to run the bench program on a 200MB file or any other custom inputs and report the results, that would be great! Thanks