improve rust bench to exclude drop time

README.md

@@ -46,7 +46,7 @@ fmt.Println(tk.Decode([]uint32{2829, 4419, 14523, 2058, 1996, 13971, 3899}, true
 
 ## Benchmarks
 
-Decoding overhead (due to CGO) is between negligible and 9% depending on the benchmark.
+Decoding overhead (due to CGO and extra allocations) is between 2% to 9% depending on the benchmark.
 
 ```bash
 go test . -bench=. -benchmem -benchtime=10s

benches/decode_benchmark.rs

@@ -26,8 +26,11 @@ fn bench_decode_n_tokens(c: &mut Criterion) {
             let tokens: Vec<u32> = (0..iters).map(|_| rng.gen_range(0..max_token_id) as u32).collect();
 
             let start = Instant::now();
-            decode(&tokenizer, black_box(&tokens), black_box(true));
-            start.elapsed()
+            let result = decode(&tokenizer, black_box(&tokens), black_box(true));
+            let duration = start.elapsed();
+            // we do this to exclude drop time for the result
+            _ = result;
+            duration
         })}
     );
 }

test/benchmark/rust_results.txt

@@ -1,14 +1,15 @@
 run with `cargo bench`
 
-ecode_n_times          time:   [3.9812 µs 3.9874 µs 3.9939 µs]
-                        change: [-0.4103% -0.1338% +0.1275%] (p = 0.33 > 0.05)
-                        No change in performance detected.
-Found 7 outliers among 100 measurements (7.00%)
-  7 (7.00%) high mild
+decode_n_times          time:   [3.9349 µs 3.9588 µs 3.9846 µs]
+                        change: [-1.6315% -1.2605% -0.8858%] (p = 0.00 < 0.05)
+                        Change within noise threshold.
+Found 10 outliers among 100 measurements (10.00%)
+  6 (6.00%) high mild
+  4 (4.00%) high severe
 
-decode_n_tokens         time:   [651.72 ns 661.73 ns 675.78 ns]
-                        change: [+0.3504% +2.0016% +3.5507%] (p = 0.01 < 0.05)
+decode_n_tokens         time:   [628.40 ns 630.04 ns 632.03 ns]
+                        change: [-3.6926% -2.1495% -0.5598%] (p = 0.01 < 0.05)
                         Change within noise threshold.
-Found 7 outliers among 100 measurements (7.00%)
+Found 4 outliers among 100 measurements (4.00%)
   2 (2.00%) high mild
-  5 (5.00%) high severe
+  2 (2.00%) high severe