Benchmark with fast float

[sirinath]

How does the performance compare with;

• https://github.com/lemire/fast_double_parser
• https://github.com/fastfloat/fast_float

[jk-jeon]

I remember I did a benchmark against fast_double_parser by myself, and the result was that fast_double_parser is significantly faster than my implementation on reasonable (up to, say, 17 digits) length inputs. Of course, on longer inputs fp is much faster simply because fast_double_parser just gives up parsing such inputs and falls back to standard library implementations. It is my plan to dig into Dr. Lemire's implementation and reimplement it myself (with some modification) to replace my original implementation. (I will be happier if someone else can do that for me instead:) )

However, for me it is rather a bad idea to just reuse his implementation, because that means unnecessary duplication of otherwise mostly same cache tables.

[sirinath]

How can the cache tables be adopted to be shared?

[jk-jeon]

Because I expect most of the entries would be the same?
Even if it's not the case, I think there might be ways to workaround the issue.

[sirinath]

Looking forward to this implementation.