feat: add #[repr(u64)] to Repr to optimize clone cost (volo-rs#6)

Thanks to @wfly1998 @0xd34d10cc

Author: PureWhiteWu

Cargo.lock

@@ -1,6 +1,6 @@
 [package]
 name = "faststr"
-version = "0.2.12"
+version = "0.2.13"
 authors = ["Volo Team <[email protected]>"]
 edition = "2021"
 description = "Faststr is a string library that reduces the cost of clone."

Cargo.toml

@@ -61,118 +61,82 @@ $ cargo bench
 
 ### AARCH64
 
-#### M1Max
-
-```
-empty faststr           time:   [19.315 ns 19.345 ns 19.377 ns]
-
-empty string            time:   [2.2097 ns 2.2145 ns 2.2194 ns]
-
-static faststr          time:   [19.483 ns 19.598 ns 19.739 ns]
-
-inline faststr          time:   [20.447 ns 20.476 ns 20.507 ns]
-
-string hello world      time:   [17.215 ns 17.239 ns 17.263 ns]
-
-512B faststr            time:   [23.883 ns 23.922 ns 23.965 ns]
-
-512B string             time:   [50.733 ns 51.360 ns 52.041 ns]
-
-4096B faststr           time:   [23.893 ns 23.959 ns 24.033 ns]
-
-4096B string            time:   [78.323 ns 79.565 ns 80.830 ns]
-
-16384B faststr          time:   [23.829 ns 23.885 ns 23.952 ns]
-
-16384B string           time:   [395.83 ns 402.46 ns 408.51 ns]
-
-65536B faststr          time:   [23.934 ns 24.002 ns 24.071 ns]
-
-65536B string           time:   [1.3142 µs 1.3377 µs 1.3606 µs]
-
-524288B faststr         time:   [23.881 ns 23.926 ns 23.976 ns]
-
-524288B string          time:   [8.8109 µs 8.8577 µs 8.9024 µs]
-
-1048576B faststr        time:   [23.968 ns 24.032 ns 24.094 ns]
-
-1048576B string         time:   [18.424 µs 18.534 µs 18.646 µs]
-```
-
 #### M3Max
 
 ```
-empty faststr           time:   [16.178 ns 16.189 ns 16.201 ns]
+empty faststr           time:   [2.0188 ns 2.0271 ns 2.0356 ns]
 
 empty string            time:   [2.1306 ns 2.1333 ns 2.1365 ns]
 
-static faststr          time:   [16.310 ns 16.325 ns 16.341 ns]
+static faststr          time:   [2.0458 ns 2.0589 ns 2.0709 ns]
 
-inline faststr          time:   [17.024 ns 17.040 ns 17.056 ns]
+inline faststr          time:   [2.2270 ns 2.2332 ns 2.2399 ns]
 
 string hello world      time:   [12.553 ns 12.575 ns 12.597 ns]
 
-512B faststr            time:   [16.771 ns 16.790 ns 16.810 ns]
+512B faststr            time:   [3.8373 ns 3.8454 ns 3.8540 ns]
 
 512B string             time:   [36.895 ns 37.007 ns 37.121 ns]
 
-4096B faststr           time:   [16.785 ns 16.811 ns 16.840 ns]
+4096B faststr           time:   [3.8205 ns 3.8260 ns 3.8317 ns]
 
 4096B string            time:   [55.275 ns 55.355 ns 55.446 ns]
 
-16384B faststr          time:   [16.796 ns 16.819 ns 16.843 ns]
+16384B faststr          time:   [3.8191 ns 3.8246 ns 3.8306 ns]
 
 16384B string           time:   [338.18 ns 352.36 ns 365.02 ns]
 
-65536B faststr          time:   [16.968 ns 17.011 ns 17.062 ns]
+65536B faststr          time:   [3.8169 ns 3.8221 ns 3.8277 ns]
 
 65536B string           time:   [662.52 ns 663.75 ns 664.96 ns]
 
-524288B faststr         time:   [16.806 ns 16.834 ns 16.868 ns]
+524288B faststr         time:   [3.8140 ns 3.8178 ns 3.8219 ns]
 
 524288B string          time:   [6.2681 µs 6.2755 µs 6.2827 µs]
 
-1048576B faststr        time:   [16.823 ns 16.849 ns 16.880 ns]
+1048576B faststr        time:   [3.8235 ns 3.8290 ns 3.8348 ns]
 
 1048576B string         time:   [12.422 µs 12.438 µs 12.453 µs]
 ```
 
-### AMD EPYC 7Y83
+### amd64
+
+#### AMD EPYC 7Y83
 
 ```
-empty faststr           time:   [42.724 ns 42.728 ns 42.732 ns]
+empty faststr           time:   [4.3325 ns 4.3330 ns 4.3335 ns]
 
-empty string            time:   [4.6490 ns 4.6494 ns 4.6499 ns]
+empty string            time:   [4.6413 ns 4.6422 ns 4.6434 ns]
 
-static faststr          time:   [42.519 ns 42.525 ns 42.532 ns]
+static faststr          time:   [4.3328 ns 4.3333 ns 4.3339 ns]
 
-inline faststr          time:   [43.446 ns 43.450 ns 43.454 ns]
+inline faststr          time:   [4.6567 ns 4.6580 ns 4.6593 ns]
 
-string hello world      time:   [12.385 ns 12.385 ns 12.387 ns]
+string hello world      time:   [12.897 ns 12.929 ns 12.954 ns]
 
-512B faststr            time:   [42.232 ns 42.238 ns 42.244 ns]
+512B faststr            time:   [4.4218 ns 4.4253 ns 4.4291 ns]
 
-512B string             time:   [15.822 ns 15.846 ns 15.894 ns]
+512B string             time:   [16.087 ns 16.094 ns 16.105 ns]
 
-4096B faststr           time:   [41.741 ns 41.918 ns 42.069 ns]
+4096B faststr           time:   [4.4066 ns 4.4099 ns 4.4141 ns]
 
-4096B string            time:   [84.492 ns 84.668 ns 84.839 ns]
+4096B string            time:   [96.905 ns 97.401 ns 97.879 ns]
 
-16384B faststr          time:   [42.245 ns 42.250 ns 42.255 ns]
+16384B faststr          time:   [4.4150 ns 4.4277 ns 4.4414 ns]
 
-16384B string           time:   [225.36 ns 225.42 ns 225.47 ns]
+16384B string           time:   [229.25 ns 229.30 ns 229.34 ns]
 
-65536B faststr          time:   [41.987 ns 42.087 ns 42.166 ns]
+65536B faststr          time:   [4.4562 ns 4.4623 ns 4.4690 ns]
 
-65536B string           time:   [1.3212 µs 1.3215 µs 1.3219 µs]
+65536B string           time:   [1.3325 µs 1.3328 µs 1.3332 µs]
 
-524288B faststr         time:   [42.272 ns 42.277 ns 42.283 ns]
+524288B faststr         time:   [4.4167 ns 4.4240 ns 4.4326 ns]
 
-524288B string          time:   [14.373 µs 14.380 µs 14.388 µs]
+524288B string          time:   [18.268 µs 18.277 µs 18.287 µs]
 
-1048576B faststr        time:   [42.279 ns 42.287 ns 42.295 ns]
+1048576B faststr        time:   [4.4275 ns 4.4385 ns 4.4494 ns]
 
-1048576B string         time:   [27.995 µs 28.015 µs 28.038 µs]
+1048576B string         time:   [32.839 µs 33.777 µs 34.554 µs]
 ```
 
 ## Related Projects

README.md

@@ -9,16 +9,21 @@ fn criterion_benchmark(c: &mut Criterion) {
 
     let s = FastStr::from("Hello, world!");
     c.bench_function("static faststr", |b| b.iter(|| black_box(s.clone())));
+    #[allow(deprecated)]
     let s = FastStr::new_inline("Hello, world!");
     c.bench_function("inline faststr", |b| b.iter(|| black_box(s.clone())));
     let s = String::from("Hello, world!");
     c.bench_function("string hello world", |b| b.iter(|| black_box(s.clone())));
 
     for size in [512, 4 * 1024, 16 * 1024, 64 * 1024, 512 * 1024, 1024 * 1024] {
         let s = FastStr::from("a".repeat(size));
+        let _s1 = black_box(s.clone());
+        let _s2 = black_box(s.clone());
         c.bench_function(format!("{}B faststr", size).as_str(), |b| {
             b.iter(|| black_box(s.clone()))
         });
+        drop(_s1);
+        drop(_s2);
         let s = String::from("a".repeat(size));
         c.bench_function(format!("{}B string", size).as_str(), |b| {
             b.iter(|| black_box(s.clone()))

benches/faststr.rs

@@ -34,10 +34,17 @@ impl FastStr {
         Self(Repr::new(text))
     }
 
-    /// Create a new inline `FastStr` (up to 38 bytes long) from a string slice `s`.
+    /// Create a new inline `FastStr` (up to 30 bytes long) from a string slice `s`.
     ///
-    /// This constructor panics if the length of `s` is greater than 38.
+    /// This constructor panics if the length of `s` is greater than 30.
+    ///
+    /// Note: the inline length is not guaranteed.
     #[inline]
+    #[doc(hidden)]
+    #[deprecated(
+        since = "0.2.13",
+        note = "The inline threshold is not stable. Please use `FastStr::new()` instead."
+    )]
     pub const fn new_inline(s: &str) -> Self {
         if s.len() > INLINE_CAP {
             panic!("[FastStr] string is too long to inline");
@@ -496,9 +503,10 @@ impl From<Cow<'static, str>> for FastStr {
     }
 }
 
-const INLINE_CAP: usize = 38;
+const INLINE_CAP: usize = 30;
 
 #[derive(Clone)]
+#[repr(u64)]
 enum Repr {
     Empty,
     Bytes(Bytes),