Use performance chart from dtoa-benchmark

Author: dtolnay

Cargo.toml

@@ -6,7 +6,7 @@ categories = ["value-formatting", "no-std", "no-std::no-alloc"]
 description = "A double-to-string conversion algorithm based on Schubfach and yy"
 documentation = "https://docs.rs/zmij"
 edition = "2021"
-exclude = ["build.rs", "performance.png", "chart/**"]
+exclude = ["build.rs", "*.png", "chart/**"]
 keywords = ["float"]
 license = "MIT"
 repository = "https://github.com/dtolnay/zmij"

README.md

@@ -23,9 +23,17 @@ fn main() {
 }
 ```
 
-## Performance (lower is better)
+## Performance
 
-![performance](https://raw.githubusercontent.com/dtolnay/zmij/master/performance.png)
+The [dtoa-benchmark] compares this library and other Rust floating point
+formatting implementations across a range of precisions. The vertical axis in
+this chart shows nanoseconds taken by a single execution of
+`zmij::Buffer::new().format_finite(value)` so a lower result indicates a faster
+library.
+
+[dtoa-benchmark]: https://github.com/dtolnay/dtoa-benchmark
+
+![performance](https://raw.githubusercontent.com/dtolnay/zmij/master/dtoa-benchmark.png)
 
 <br>