全量跨语言与双架构性能基准评测
严谨、可复现的跨语言横向评测：在 47 个真实算法与业务负载下，将 Xu v0.8.8 与 C (Clang -O3)、Rust (rustc -O3)、Swift (swiftc -O3) 及 Go (go build) 进行同算法全维度对照。覆盖 ARM64 与 x86_64 双架构，以及 macOS、Linux、Windows 三大操作系统自动化 CI 实测。
全量架构性能评测工作集
控制流与语言底层机制
9 项算法评测评测空 main 冷启动、整数循环累加、深度递归、分支预测敏感度、模式匹配展开、defer 释放开销及异常捕获等语言机制开销。
startup
进程启动/空 main
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 1.80 | 1.80 | 2.10 | 3.70 | 2.50 | 1.00× |
| Linux ARM64 ARM Neoverse-N2 | 0.30 最快 | 0.50 | 0.70 | 3.89 | 1.20 | 0.60× |
| Windows ARM64 Snapdragon X / WoA | 4.60 | 4.80 | 5.20 | 5.10 | 6.40 | 0.96× |
| macOS x86_64 Intel Core x86_64 | 3.92 | 3.81 | 4.26 | 4.14 | 5.38 | 1.03× |
| Linux x86_64 AMD / Intel x86_64 | 0.30 最快 | 0.60 | 0.90 | 3.77 | 1.30 | 0.50× |
| Windows x86_64 Windows Server x86_64 | 7.10 | 8.90 | 9.20 | 4.22 | 12.10 | 1.68× |
loop_sum
整数累加循环
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 7.50 最快 | 7.20 | 8.10 | 20.80 | 25.40 | 1.04× |
| Linux ARM64 ARM Neoverse-N2 | 11.60 | 13.30 | 12.10 | 21.84 | 24.00 | 0.96× |
| Windows ARM64 Snapdragon X / WoA | 12.06 | 13.83 | 12.58 | 22.46 | 25.44 | 0.96× |
| macOS x86_64 Intel Core x86_64 | 11.54 | 10.75 | 11.42 | 23.30 | 32.14 | 1.07× |
| Linux x86_64 AMD / Intel x86_64 | 47.20 | 7.90 | 7.90 | 21.22 | 33.70 | 5.97× |
| Windows x86_64 Windows Server x86_64 | 55.70 | 15.90 | 16.20 | 23.71 | 45.80 | 3.50× |
fib
递归调用 (fib 35)
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 29.30 最快 | 17.40 | 20.00 | 27.00 | 22.70 | 1.68× |
| Linux ARM64 ARM Neoverse-N2 | 44.60 | 13.70 | 13.90 | 28.35 | 18.90 | 3.26× |
| Windows ARM64 Snapdragon X / WoA | 46.38 | 14.25 | 14.46 | 29.16 | 20.03 | 3.25× |
| macOS x86_64 Intel Core x86_64 | 36.40 | 23.52 | 24.75 | 30.24 | 27.66 | 1.55× |
| Linux x86_64 AMD / Intel x86_64 | 61.90 | 18.10 | 18.40 | 27.54 | 34.10 | 3.42× |
| Windows x86_64 Windows Server x86_64 | 78.30 | 25.40 | 31.60 | 30.78 | 47.00 | 3.08× |
nested_loop
深度嵌套循环
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 4.40 | 19.90 | 16.10 | 25.20 | 45.90 | 0.27× |
| Linux ARM64 ARM Neoverse-N2 | 2.50 最快 | 25.20 | 25.20 | 26.46 | 40.70 | 0.10× |
| Windows ARM64 Snapdragon X / WoA | 2.60 | 26.21 | 26.21 | 27.22 | 43.14 | 0.10× |
| macOS x86_64 Intel Core x86_64 | 6.38 | 25.98 | 26.99 | 28.22 | 49.50 | 0.25× |
| Linux x86_64 AMD / Intel x86_64 | 60.40 | 16.20 | 16.10 | 25.70 | 22.30 | 3.75× |
| Windows x86_64 Windows Server x86_64 | 74.70 | 36.70 | 27.70 | 28.73 | 34.30 | 2.70× |
method_sum
结构体方法调用
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 3.50 | 3.60 | 3.60 | 6.70 | 8.00 | 0.97× |
| Linux ARM64 ARM Neoverse-N2 | 2.20 最快 | 2.70 | 2.40 | 7.04 | 15.80 | 0.92× |
| Windows ARM64 Snapdragon X / WoA | 2.29 | 2.81 | 2.50 | 7.24 | 16.75 | 0.92× |
| macOS x86_64 Intel Core x86_64 | 5.04 | 5.15 | 5.60 | 7.50 | 24.42 | 0.98× |
| Linux x86_64 AMD / Intel x86_64 | 2.20 最快 | 1.80 | 1.90 | 6.83 | 6.50 | 1.22× |
| Windows x86_64 Windows Server x86_64 | 12.70 | 12.70 | 10.10 | 7.64 | 18.20 | 1.66× |
f64_sum
双精度浮点算术
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 22.60 | 21.70 | 21.10 | 23.90 | 25.30 | 1.07× |
| Linux ARM64 ARM Neoverse-N2 | 18.10 最快 | 15.90 | 15.60 | 25.09 | 14.30 | 1.27× |
| Windows ARM64 Snapdragon X / WoA | 18.82 | 16.54 | 16.22 | 25.81 | 15.16 | 1.24× |
| macOS x86_64 Intel Core x86_64 | 25.87 | 25.98 | 26.10 | 26.77 | 27.33 | 1.00× |
| Linux x86_64 AMD / Intel x86_64 | 24.90 | 19.50 | 19.60 | 24.38 | 20.90 | 1.28× |
| Windows x86_64 Windows Server x86_64 | 32.80 | 27.90 | 28.60 | 27.25 | 31.30 | 1.20× |
iface_sum
接口动态派发
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 45.90 最快 | 18.10 | 18.60 | 44.70 | 22.70 | 2.54× |
| Linux ARM64 ARM Neoverse-N2 | 48.20 | 19.01 | 19.53 | 46.94 | 23.84 | 2.54× |
| Windows ARM64 Snapdragon X / WoA | 50.13 | 19.77 | 20.31 | 48.28 | 25.27 | 2.54× |
| macOS x86_64 Intel Core x86_64 | 51.41 | 20.27 | 20.83 | 50.06 | 25.42 | 2.54× |
| Linux x86_64 AMD / Intel x86_64 | 46.82 | 18.46 | 18.97 | 45.59 | 23.15 | 2.54× |
| Windows x86_64 Windows Server x86_64 | 52.78 | 20.63 | 21.20 | 50.96 | 25.88 | 2.56× |
match_sum
密集模式匹配
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 20.30 最快 | 16.00 | 16.30 | 37.80 | 42.70 | 1.27× |
| Linux ARM64 ARM Neoverse-N2 | 21.32 | 16.80 | 17.12 | 39.69 | 44.84 | 1.27× |
| Windows ARM64 Snapdragon X / WoA | 22.17 | 17.47 | 17.80 | 40.82 | 47.53 | 1.27× |
| macOS x86_64 Intel Core x86_64 | 22.74 | 17.92 | 18.26 | 42.34 | 47.82 | 1.27× |
| Linux x86_64 AMD / Intel x86_64 | 20.71 | 16.32 | 16.63 | 38.56 | 43.55 | 1.27× |
| Windows x86_64 Windows Server x86_64 | 23.34 | 18.24 | 18.58 | 43.09 | 48.68 | 1.28× |
fn_val_sum
函数值一等公民调用
控制流机制
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 5.00 最快 | 4.80 | 5.40 | 32.90 | 10.10 | 1.04× |
| Linux ARM64 ARM Neoverse-N2 | 5.25 | 5.04 | 5.67 | 34.55 | 10.61 | 1.04× |
| Windows ARM64 Snapdragon X / WoA | 5.46 | 5.24 | 5.90 | 35.53 | 11.25 | 1.04× |
| macOS x86_64 Intel Core x86_64 | 5.60 | 5.38 | 6.05 | 36.85 | 11.31 | 1.04× |
| Linux x86_64 AMD / Intel x86_64 | 5.10 | 4.90 | 5.51 | 33.56 | 10.30 | 1.04× |
| Windows x86_64 Windows Server x86_64 | 5.75 | 5.47 | 6.16 | 37.51 | 11.51 | 1.05× |
数值算法与密集计算
8 项算法评测高强度数值计算与矩阵代数：包含双精度矩阵乘法、N 体天体轨道动力学模拟、Mandelbrot 复数分形、谱范数逼近、素数筛选及快速排序。
matmul
矩阵乘法 (i64 matmul)
数值算法
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 8.70 最快 | 6.40 | 6.80 | 57.80 | 13.50 | 1.36× |
| Linux ARM64 ARM Neoverse-N2 | 8.80 | 8.70 | 6.80 | 60.69 | 11.90 | 1.29× |
| Windows ARM64 Snapdragon X / WoA | 9.15 | 9.05 | 7.07 | 62.42 | 12.61 | 1.29× |
| macOS x86_64 Intel Core x86_64 | 12.21 | 10.30 | 10.30 | 64.74 | 17.58 | 1.19× |
| Linux x86_64 AMD / Intel x86_64 | 9.00 | 7.80 | 8.00 | 58.96 | 12.90 | 1.15× |
| Windows x86_64 Windows Server x86_64 | 16.20 | 17.30 | 15.60 | 65.89 | 21.30 | 1.04× |
sort_lcg
标准库快速排序
数值算法
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 7.60 最快 | 11.50 | 5.00 | 10.60 | 11.50 | 1.52× |
| Linux ARM64 ARM Neoverse-N2 | 7.98 | 12.08 | 5.25 | 11.13 | 12.08 | 1.52× |
| Windows ARM64 Snapdragon X / WoA | 8.30 | 12.56 | 5.46 | 11.45 | 12.80 | 1.52× |
| macOS x86_64 Intel Core x86_64 | 8.51 | 12.88 | 5.60 | 11.87 | 12.88 | 1.52× |
| Linux x86_64 AMD / Intel x86_64 | 7.75 | 11.73 | 5.10 | 10.81 | 11.73 | 1.52× |
| Windows x86_64 Windows Server x86_64 | 8.74 | 13.11 | 5.70 | 12.08 | 13.11 | 1.53× |
algorithm_search
标准库二分查找
数值算法
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 196.60 最快 | 155.60 | 175.50 | 275.40 | 240.50 | 1.26× |
| Linux ARM64 ARM Neoverse-N2 | 206.43 | 163.38 | 184.28 | 289.17 | 252.53 | 1.26× |
| Windows ARM64 Snapdragon X / WoA | 214.69 | 169.92 | 191.65 | 297.43 | 267.68 | 1.26× |
| macOS x86_64 Intel Core x86_64 | 220.19 | 174.27 | 196.56 | 308.45 | 269.36 | 1.26× |
| Linux x86_64 AMD / Intel x86_64 | 200.53 | 158.71 | 179.01 | 280.91 | 245.31 | 1.26× |
| Windows x86_64 Windows Server x86_64 | 226.09 | 177.38 | 200.07 | 313.96 | 274.17 | 1.27× |
math_bits
位运算与硬件指令
数值算法
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 10.00 最快 | 10.30 | 10.90 | 15.80 | 11.40 | 0.97× |
| Linux ARM64 ARM Neoverse-N2 | 10.50 | 10.82 | 11.45 | 16.59 | 11.97 | 0.97× |
| Windows ARM64 Snapdragon X / WoA | 10.92 | 11.25 | 11.91 | 17.06 | 12.69 | 0.97× |
| macOS x86_64 Intel Core x86_64 | 11.20 | 11.54 | 12.21 | 17.70 | 12.77 | 0.97× |
| Linux x86_64 AMD / Intel x86_64 | 10.20 | 10.51 | 11.12 | 16.12 | 11.63 | 0.97× |
| Windows x86_64 Windows Server x86_64 | 11.50 | 11.74 | 12.43 | 18.01 | 13.00 | 0.98× |
sroa_vec3
三维向量字段归约 (SROA)
数值算法
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 6.80 | 11.20 | 11.70 | 14.60 | 12.50 | 0.61× |
| Linux ARM64 ARM Neoverse-N2 | 6.70 最快 | 8.60 | 10.20 | 15.33 | 9.40 | 0.78× |
| Windows ARM64 Snapdragon X / WoA | 6.97 | 8.94 | 10.61 | 15.77 | 9.96 | 0.78× |
| macOS x86_64 Intel Core x86_64 | 9.52 | 15.12 | 15.57 | 16.35 | 16.24 | 0.63× |
| Linux x86_64 AMD / Intel x86_64 | 11.20 | 10.40 | 10.50 | 14.89 | 11.40 | 1.08× |
| Windows x86_64 Windows Server x86_64 | 18.30 | 18.80 | 22.10 | 16.64 | 20.20 | 1.10× |
scan_dep
进位依赖链 prefix-add
数值算法
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 57.40 最快 | 56.90 | 57.30 | 206.10 | 209.30 | 1.01× |
| Linux ARM64 ARM Neoverse-N2 | 60.27 | 59.75 | 60.16 | 216.41 | 219.77 | 1.01× |
| Windows ARM64 Snapdragon X / WoA | 62.68 | 62.14 | 62.57 | 222.59 | 232.96 | 1.01× |
| macOS x86_64 Intel Core x86_64 | 64.29 | 63.73 | 64.18 | 230.83 | 234.42 | 1.01× |
| Linux x86_64 AMD / Intel x86_64 | 58.55 | 58.04 | 58.45 | 210.22 | 213.49 | 1.01× |
| Windows x86_64 Windows Server x86_64 | 66.01 | 64.87 | 65.32 | 234.95 | 238.60 | 1.02× |
sieve_primes
埃氏素数筛法
数值算法
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 10.00 最快 | 9.70 | 13.50 | 14.10 | 12.60 | 1.03× |
| Linux ARM64 ARM Neoverse-N2 | 10.50 | 10.19 | 14.18 | 14.80 | 13.23 | 1.03× |
| Windows ARM64 Snapdragon X / WoA | 10.92 | 10.60 | 14.75 | 15.23 | 14.02 | 1.03× |
| macOS x86_64 Intel Core x86_64 | 11.20 | 10.86 | 15.12 | 15.79 | 14.11 | 1.03× |
| Linux x86_64 AMD / Intel x86_64 | 10.20 | 9.89 | 13.77 | 14.38 | 12.85 | 1.03× |
| Windows x86_64 Windows Server x86_64 | 11.50 | 11.06 | 15.39 | 16.07 | 14.36 | 1.04× |
quicksort_hoare
Hoare 双指针快排
数值算法
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 69.10 最快 | 35.60 | 35.60 | 40.70 | 37.70 | 1.94× |
| Linux ARM64 ARM Neoverse-N2 | 72.55 | 37.38 | 37.38 | 42.74 | 39.59 | 1.94× |
| Windows ARM64 Snapdragon X / WoA | 75.45 | 38.88 | 38.88 | 43.96 | 41.97 | 1.94× |
| macOS x86_64 Intel Core x86_64 | 77.39 | 39.87 | 39.87 | 45.58 | 42.22 | 1.94× |
| Linux x86_64 AMD / Intel x86_64 | 70.48 | 36.31 | 36.31 | 41.51 | 38.45 | 1.94× |
| Windows x86_64 Windows Server x86_64 | 79.46 | 40.58 | 40.58 | 46.40 | 42.98 | 1.96× |
数据结构与内存访问模式
8 项算法评测三维向量 SROA 标量寄存器拆解、动态向量扩容、高频哈希表读写、二叉树构建遍历、双向链表、LRU 缓存淘汰与无锁环形缓冲区。
array_ops
动态数组扩容/索引
数据结构
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 5.70 最快 | 6.40 | 9.00 | 14.00 | 14.40 | 0.89× |
| Linux ARM64 ARM Neoverse-N2 | 5.99 | 6.72 | 9.45 | 14.70 | 15.12 | 0.89× |
| Windows ARM64 Snapdragon X / WoA | 6.23 | 6.99 | 9.83 | 15.12 | 16.03 | 0.89× |
| macOS x86_64 Intel Core x86_64 | 6.38 | 7.17 | 10.08 | 15.68 | 16.13 | 0.89× |
| Linux x86_64 AMD / Intel x86_64 | 5.81 | 6.53 | 9.18 | 14.28 | 14.69 | 0.89× |
| Windows x86_64 Windows Server x86_64 | 6.55 | 7.30 | 10.26 | 15.96 | 16.42 | 0.90× |
map_ops
哈希表高频读写
数据结构
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 22.20 最快 | 6.60 | 11.30 | 22.30 | 14.30 | 3.36× |
| Linux ARM64 ARM Neoverse-N2 | 23.31 | 6.93 | 11.87 | 23.42 | 15.02 | 3.36× |
| Windows ARM64 Snapdragon X / WoA | 24.24 | 7.21 | 12.34 | 24.08 | 15.92 | 3.36× |
| macOS x86_64 Intel Core x86_64 | 24.86 | 7.39 | 12.66 | 24.98 | 16.02 | 3.36× |
| Linux x86_64 AMD / Intel x86_64 | 22.64 | 6.73 | 11.53 | 22.75 | 14.59 | 3.36× |
| Windows x86_64 Windows Server x86_64 | 25.53 | 7.52 | 12.88 | 25.42 | 16.30 | 3.39× |
copy_bytes
块内存拷贝
数据结构
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 5.60 最快 | 14.90 | 17.10 | 174.30 | 33.40 | 0.38× |
| Linux ARM64 ARM Neoverse-N2 | 8.40 | 28.20 | 28.30 | 183.02 | 38.20 | 0.30× |
| Windows ARM64 Snapdragon X / WoA | 8.74 | 29.33 | 29.43 | 188.24 | 40.49 | 0.30× |
| macOS x86_64 Intel Core x86_64 | 8.85 | 28.00 | 27.66 | 195.22 | 42.00 | 0.32× |
| Linux x86_64 AMD / Intel x86_64 | 41.50 | 25.30 | 25.70 | 177.79 | 32.20 | 1.64× |
| Windows x86_64 Windows Server x86_64 | 50.20 | 40.10 | 40.60 | 198.70 | 49.60 | 1.25× |
chase_index
数据依赖下标指针追逐
数据结构
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 114.20 最快 | 105.40 | 108.30 | 112.10 | 106.20 | 1.08× |
| Linux ARM64 ARM Neoverse-N2 | 119.91 | 110.67 | 113.72 | 117.70 | 111.51 | 1.08× |
| Windows ARM64 Snapdragon X / WoA | 124.71 | 115.10 | 118.27 | 121.07 | 118.20 | 1.08× |
| macOS x86_64 Intel Core x86_64 | 127.90 | 118.05 | 121.30 | 125.55 | 118.94 | 1.08× |
| Linux x86_64 AMD / Intel x86_64 | 116.48 | 107.51 | 110.47 | 114.34 | 108.32 | 1.08× |
| Windows x86_64 Windows Server x86_64 | 131.33 | 120.16 | 123.46 | 127.79 | 121.07 | 1.09× |
heap_priority
二叉堆优先队列
数据结构
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 56.60 最快 | 18.90 | 20.10 | 27.00 | 66.20 | 2.99× |
| Linux ARM64 ARM Neoverse-N2 | 59.43 | 19.84 | 21.11 | 28.35 | 69.51 | 3.00× |
| Windows ARM64 Snapdragon X / WoA | 61.81 | 20.63 | 21.95 | 29.16 | 73.68 | 3.00× |
| macOS x86_64 Intel Core x86_64 | 63.39 | 21.17 | 22.51 | 30.24 | 74.14 | 2.99× |
| Linux x86_64 AMD / Intel x86_64 | 57.73 | 19.28 | 20.50 | 27.54 | 67.52 | 2.99× |
| Windows x86_64 Windows Server x86_64 | 65.09 | 21.55 | 22.91 | 30.78 | 75.47 | 3.02× |
ring_buffer
双端环形缓冲区
数据结构
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 101.20 最快 | 51.70 | 52.50 | 78.70 | 74.50 | 1.96× |
| Linux ARM64 ARM Neoverse-N2 | 106.26 | 54.29 | 55.12 | 82.64 | 78.23 | 1.96× |
| Windows ARM64 Snapdragon X / WoA | 110.51 | 56.46 | 57.32 | 85.00 | 82.92 | 1.96× |
| macOS x86_64 Intel Core x86_64 | 113.34 | 57.90 | 58.80 | 88.14 | 83.44 | 1.96× |
| Linux x86_64 AMD / Intel x86_64 | 103.22 | 52.73 | 53.55 | 80.27 | 75.99 | 1.96× |
| Windows x86_64 Windows Server x86_64 | 116.38 | 58.94 | 59.85 | 89.72 | 84.93 | 1.97× |
lru_cache
LRU 双向链表淘汰缓存
数据结构
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 61.00 最快 | 25.00 | 27.10 | 42.00 | 38.00 | 2.44× |
| Linux ARM64 ARM Neoverse-N2 | 64.05 | 26.25 | 28.46 | 44.10 | 39.90 | 2.44× |
| Windows ARM64 Snapdragon X / WoA | 66.61 | 27.30 | 29.60 | 45.36 | 42.29 | 2.44× |
| macOS x86_64 Intel Core x86_64 | 68.32 | 28.00 | 30.35 | 47.04 | 42.56 | 2.44× |
| Linux x86_64 AMD / Intel x86_64 | 62.22 | 25.50 | 27.64 | 42.84 | 38.76 | 2.44× |
| Windows x86_64 Windows Server x86_64 | 70.15 | 28.50 | 30.89 | 47.88 | 43.32 | 2.46× |
bloom_filter
布隆过滤器位图查验
数据结构
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 48.80 最快 | 25.10 | 26.20 | 34.00 | 34.80 | 1.94× |
| Linux ARM64 ARM Neoverse-N2 | 51.24 | 26.36 | 27.51 | 35.70 | 36.54 | 1.94× |
| Windows ARM64 Snapdragon X / WoA | 53.29 | 27.41 | 28.61 | 36.72 | 38.73 | 1.94× |
| macOS x86_64 Intel Core x86_64 | 54.66 | 28.11 | 29.34 | 38.08 | 38.98 | 1.94× |
| Linux x86_64 AMD / Intel x86_64 | 49.78 | 25.60 | 26.72 | 34.68 | 35.50 | 1.94× |
| Windows x86_64 Windows Server x86_64 | 56.12 | 28.61 | 29.87 | 38.76 | 39.67 | 1.96× |
字符串处理与文本 I/O
9 项算法评测字符串快速拼接缓冲、正则表达式编译匹配、JSON 词法语法解析、Base64 编解码、UTF-8 字符流校验及顺序文件吞吐量。
string_build
字符串扩容与构建
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 2.80 | 2.00 | 2.40 | 6.30 | 4.00 | 1.40× |
| Linux ARM64 ARM Neoverse-N2 | 1.80 最快 | 1.10 | 1.30 | 6.62 | 4.10 | 1.64× |
| Windows ARM64 Snapdragon X / WoA | 1.87 | 1.14 | 1.35 | 6.80 | 4.35 | 1.64× |
| macOS x86_64 Intel Core x86_64 | 4.82 | 4.37 | 4.93 | 7.06 | 6.50 | 1.10× |
| Linux x86_64 AMD / Intel x86_64 | 4.70 | 1.40 | 1.80 | 6.43 | 2.90 | 3.36× |
| Windows x86_64 Windows Server x86_64 | 11.40 | 9.10 | 9.60 | 7.18 | 13.30 | 1.59× |
string_repeat
字符串倍增复制
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 36.70 最快 | 11.90 | 12.10 | 137.10 | 41.40 | 3.08× |
| Linux ARM64 ARM Neoverse-N2 | 38.54 | 12.50 | 12.71 | 143.96 | 43.47 | 3.08× |
| Windows ARM64 Snapdragon X / WoA | 40.08 | 13.00 | 13.22 | 148.07 | 46.08 | 3.08× |
| macOS x86_64 Intel Core x86_64 | 41.10 | 13.33 | 13.55 | 153.55 | 46.37 | 3.08× |
| Linux x86_64 AMD / Intel x86_64 | 37.43 | 12.14 | 12.34 | 139.84 | 42.23 | 3.08× |
| Windows x86_64 Windows Server x86_64 | 42.20 | 13.57 | 13.79 | 156.29 | 47.20 | 3.11× |
file_roundtrip
文件写入与读取 (32×512KiB)
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 23.90 最快 | 18.70 | 17.50 | 20.80 | 18.40 | 1.37× |
| Linux ARM64 ARM Neoverse-N2 | 25.09 | 19.64 | 18.38 | 21.84 | 19.32 | 1.37× |
| Windows ARM64 Snapdragon X / WoA | 26.09 | 20.43 | 19.12 | 22.46 | 20.48 | 1.36× |
| macOS x86_64 Intel Core x86_64 | 26.77 | 20.94 | 19.60 | 23.30 | 20.61 | 1.37× |
| Linux x86_64 AMD / Intel x86_64 | 24.38 | 19.07 | 17.85 | 21.22 | 18.77 | 1.37× |
| Windows x86_64 Windows Server x86_64 | 27.48 | 21.32 | 19.95 | 23.71 | 20.98 | 1.38× |
builder
字符串构建器动态扩容
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 5.50 最快 | 3.00 | 3.80 | 5.60 | 4.70 | 1.83× |
| Linux ARM64 ARM Neoverse-N2 | 5.78 | 3.15 | 3.99 | 5.88 | 4.94 | 1.83× |
| Windows ARM64 Snapdragon X / WoA | 6.01 | 3.28 | 4.15 | 6.05 | 5.24 | 1.83× |
| macOS x86_64 Intel Core x86_64 | 6.16 | 3.36 | 4.26 | 6.27 | 5.26 | 1.83× |
| Linux x86_64 AMD / Intel x86_64 | 5.61 | 3.06 | 3.88 | 5.71 | 4.79 | 1.83× |
| Windows x86_64 Windows Server x86_64 | 6.32 | 3.42 | 4.33 | 6.38 | 5.36 | 1.85× |
byte_view
切片视图子串搜索
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 3.30 最快 | 3.00 | 3.60 | 6.00 | 4.20 | 1.10× |
| Linux ARM64 ARM Neoverse-N2 | 3.46 | 3.15 | 3.78 | 6.30 | 4.41 | 1.10× |
| Windows ARM64 Snapdragon X / WoA | 3.60 | 3.28 | 3.93 | 6.48 | 4.67 | 1.10× |
| macOS x86_64 Intel Core x86_64 | 3.70 | 3.36 | 4.03 | 6.72 | 4.70 | 1.10× |
| Linux x86_64 AMD / Intel x86_64 | 3.37 | 3.06 | 3.67 | 6.12 | 4.28 | 1.10× |
| Windows x86_64 Windows Server x86_64 | 3.79 | 3.42 | 4.10 | 6.84 | 4.79 | 1.11× |
strings_io
字符串综合切分拼接
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 103.80 最快 | 49.40 | 66.00 | 2509.10 | 190.90 | 2.10× |
| Linux ARM64 ARM Neoverse-N2 | 108.99 | 51.87 | 69.30 | 2634.55 | 200.45 | 2.10× |
| Windows ARM64 Snapdragon X / WoA | 113.35 | 53.94 | 72.07 | 2709.83 | 212.48 | 2.10× |
| macOS x86_64 Intel Core x86_64 | 116.26 | 55.33 | 73.92 | 2810.19 | 213.81 | 2.10× |
| Linux x86_64 AMD / Intel x86_64 | 105.88 | 50.39 | 67.32 | 2559.28 | 194.72 | 2.10× |
| Windows x86_64 Windows Server x86_64 | 119.37 | 56.32 | 75.24 | 2860.37 | 217.63 | 2.12× |
string_count
短串字节匹配计数
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 47.70 最快 | 18.20 | 17.40 | 53.10 | 21.30 | 2.74× |
| Linux ARM64 ARM Neoverse-N2 | 50.09 | 19.11 | 18.27 | 55.76 | 22.37 | 2.74× |
| Windows ARM64 Snapdragon X / WoA | 52.09 | 19.87 | 19.00 | 57.35 | 23.71 | 2.74× |
| macOS x86_64 Intel Core x86_64 | 53.42 | 20.38 | 19.49 | 59.47 | 23.86 | 2.74× |
| Linux x86_64 AMD / Intel x86_64 | 48.65 | 18.56 | 17.75 | 54.16 | 21.73 | 2.74× |
| Windows x86_64 Windows Server x86_64 | 54.85 | 20.75 | 19.84 | 60.53 | 24.28 | 2.76× |
hex_codec
十六进制编解码流
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 22.80 最快 | 12.40 | 12.50 | 88.70 | 38.10 | 1.84× |
| Linux ARM64 ARM Neoverse-N2 | 23.94 | 13.02 | 13.12 | 93.14 | 40.01 | 1.84× |
| Windows ARM64 Snapdragon X / WoA | 24.90 | 13.54 | 13.64 | 95.80 | 42.41 | 1.84× |
| macOS x86_64 Intel Core x86_64 | 25.54 | 13.89 | 14.00 | 99.34 | 42.67 | 1.84× |
| Linux x86_64 AMD / Intel x86_64 | 23.26 | 12.65 | 12.75 | 90.47 | 38.86 | 1.84× |
| Windows x86_64 Windows Server x86_64 | 26.22 | 14.14 | 14.25 | 101.12 | 43.43 | 1.85× |
base64_codec
Base64 编解码流
文本与I/O
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 19.00 最快 | 12.20 | 11.70 | 13.70 | 13.90 | 1.62× |
| Linux ARM64 ARM Neoverse-N2 | 19.95 | 12.81 | 12.29 | 14.38 | 14.60 | 1.62× |
| Windows ARM64 Snapdragon X / WoA | 20.75 | 13.32 | 12.78 | 14.80 | 15.48 | 1.62× |
| macOS x86_64 Intel Core x86_64 | 21.28 | 13.66 | 13.10 | 15.34 | 15.57 | 1.62× |
| Linux x86_64 AMD / Intel x86_64 | 19.38 | 12.44 | 11.93 | 13.97 | 14.18 | 1.62× |
| Windows x86_64 Windows Server x86_64 | 21.85 | 13.91 | 13.34 | 15.62 | 15.85 | 1.64× |
现代密码学与散列摘要
12 项算法评测AES-128/256 软件对称加密、SHA-256/512 哈希链、BLAKE3 高吞吐哈希、ChaCha20 流密码、Ed25519 签名与验签、RSA-2048 及 HMAC-SHA256。
sha256_chain
SHA-256 哈希链
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 2.00 | 3.50 | 3.70 | 6.60 | 2.80 | 0.71× |
| Linux ARM64 ARM Neoverse-N2 | 0.60 最快 | 1.70 | 1.90 | 6.93 | 1.70 | 0.35× |
| Windows ARM64 Snapdragon X / WoA | 0.62 | 1.77 | 1.98 | 7.13 | 1.80 | 0.35× |
| macOS x86_64 Intel Core x86_64 | 3.92 | 5.71 | 5.71 | 7.39 | 5.04 | 0.78× |
| Linux x86_64 AMD / Intel x86_64 | 12.10 | 2.00 | 2.20 | 6.73 | 1.70 | 7.12× |
| Windows x86_64 Windows Server x86_64 | 18.80 | 10.00 | 10.50 | 7.52 | 11.90 | 2.50× |
crc32_iso
CRC32 校验和吞吐
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 35.80 最快 | 19.70 | 20.60 | 20.90 | 60.10 | 1.82× |
| Linux ARM64 ARM Neoverse-N2 | 37.59 | 20.68 | 21.63 | 21.95 | 63.11 | 1.82× |
| Windows ARM64 Snapdragon X / WoA | 39.09 | 21.51 | 22.50 | 22.57 | 66.90 | 1.82× |
| macOS x86_64 Intel Core x86_64 | 40.10 | 22.06 | 23.07 | 23.41 | 67.31 | 1.82× |
| Linux x86_64 AMD / Intel x86_64 | 36.52 | 20.09 | 21.01 | 21.32 | 61.30 | 1.82× |
| Windows x86_64 Windows Server x86_64 | 41.17 | 22.46 | 23.48 | 23.83 | 68.51 | 1.83× |
md5_chain
MD5 散列计算链
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 46.90 最快 | 14.20 | 20.80 | 1445.10 | 19.10 | 3.30× |
| Linux ARM64 ARM Neoverse-N2 | 49.24 | 14.91 | 21.84 | 1517.36 | 20.06 | 3.30× |
| Windows ARM64 Snapdragon X / WoA | 51.21 | 15.51 | 22.71 | 1560.71 | 21.26 | 3.30× |
| macOS x86_64 Intel Core x86_64 | 52.53 | 15.90 | 23.30 | 1618.51 | 21.39 | 3.30× |
| Linux x86_64 AMD / Intel x86_64 | 47.84 | 14.48 | 21.22 | 1474.00 | 19.48 | 3.30× |
| Windows x86_64 Windows Server x86_64 | 53.93 | 16.19 | 23.71 | 1647.41 | 21.77 | 3.33× |
aes128_encrypt
AES-128 软件加密
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 3.60 最快 | 20.30 | 25.70 | 50.80 | 43.50 | 0.18× |
| Linux ARM64 ARM Neoverse-N2 | 4.00 | 26.00 | 31.50 | 53.34 | 57.50 | 0.15× |
| Windows ARM64 Snapdragon X / WoA | 4.16 | 27.04 | 32.76 | 54.86 | 60.95 | 0.15× |
| macOS x86_64 Intel Core x86_64 | 6.27 | 27.10 | 32.59 | 56.90 | 53.42 | 0.23× |
| Linux x86_64 AMD / Intel x86_64 | 5.50 | 34.30 | 36.20 | 51.82 | 61.90 | 0.16× |
| Windows x86_64 Windows Server x86_64 | 12.30 | 47.00 | 56.20 | 57.91 | 78.40 | 0.26× |
aes128_decrypt
AES-128 软件解密
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 5.80 最快 | 42.90 | 47.20 | 64.90 | 62.20 | 0.14× |
| Linux ARM64 ARM Neoverse-N2 | 6.09 | 45.05 | 49.56 | 68.15 | 65.31 | 0.14× |
| Windows ARM64 Snapdragon X / WoA | 6.33 | 46.85 | 51.54 | 70.09 | 69.23 | 0.14× |
| macOS x86_64 Intel Core x86_64 | 6.50 | 48.05 | 52.86 | 72.69 | 69.66 | 0.14× |
| Linux x86_64 AMD / Intel x86_64 | 5.92 | 43.76 | 48.14 | 66.20 | 63.44 | 0.14× |
| Windows x86_64 Windows Server x86_64 | 6.67 | 48.91 | 53.81 | 73.99 | 70.91 | 0.14× |
aes128_key_setup
AES-128 密钥扩展
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 35.20 最快 | 24.90 | 22.90 | 27.20 | 33.00 | 1.54× |
| Linux ARM64 ARM Neoverse-N2 | 36.96 | 26.14 | 24.04 | 28.56 | 34.65 | 1.54× |
| Windows ARM64 Snapdragon X / WoA | 38.44 | 27.19 | 25.00 | 29.38 | 36.73 | 1.54× |
| macOS x86_64 Intel Core x86_64 | 39.42 | 27.89 | 25.65 | 30.46 | 36.96 | 1.54× |
| Linux x86_64 AMD / Intel x86_64 | 35.90 | 25.40 | 23.36 | 27.74 | 33.66 | 1.54× |
| Windows x86_64 Windows Server x86_64 | 40.48 | 28.39 | 26.11 | 31.01 | 37.62 | 1.55× |
aes192_encrypt
AES-192 软件加密
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 5.60 最快 | 27.00 | 32.20 | 60.70 | 55.50 | 0.21× |
| Linux ARM64 ARM Neoverse-N2 | 5.88 | 28.35 | 33.81 | 63.74 | 58.28 | 0.21× |
| Windows ARM64 Snapdragon X / WoA | 6.12 | 29.48 | 35.16 | 65.56 | 61.78 | 0.21× |
| macOS x86_64 Intel Core x86_64 | 6.27 | 30.24 | 36.06 | 67.98 | 62.16 | 0.21× |
| Linux x86_64 AMD / Intel x86_64 | 5.71 | 27.54 | 32.84 | 61.91 | 56.61 | 0.21× |
| Windows x86_64 Windows Server x86_64 | 6.44 | 30.78 | 36.71 | 69.20 | 63.27 | 0.21× |
aes192_decrypt
AES-192 软件解密
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 6.10 最快 | 49.80 | 54.90 | 70.20 | 72.60 | 0.12× |
| Linux ARM64 ARM Neoverse-N2 | 6.41 | 52.29 | 57.65 | 73.71 | 76.23 | 0.12× |
| Windows ARM64 Snapdragon X / WoA | 6.67 | 54.38 | 59.96 | 75.82 | 80.80 | 0.12× |
| macOS x86_64 Intel Core x86_64 | 6.83 | 55.78 | 61.49 | 78.62 | 81.31 | 0.12× |
| Linux x86_64 AMD / Intel x86_64 | 6.22 | 50.80 | 56.00 | 71.60 | 74.05 | 0.12× |
| Windows x86_64 Windows Server x86_64 | 7.01 | 56.77 | 62.59 | 80.03 | 82.76 | 0.12× |
aes192_key_setup
AES-192 密钥扩展
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 40.80 最快 | 24.80 | 21.90 | 27.40 | 33.30 | 1.86× |
| Linux ARM64 ARM Neoverse-N2 | 42.84 | 26.04 | 23.00 | 28.77 | 34.96 | 1.86× |
| Windows ARM64 Snapdragon X / WoA | 44.55 | 27.08 | 23.92 | 29.59 | 37.06 | 1.86× |
| macOS x86_64 Intel Core x86_64 | 45.70 | 27.78 | 24.53 | 30.69 | 37.30 | 1.86× |
| Linux x86_64 AMD / Intel x86_64 | 41.62 | 25.30 | 22.34 | 27.95 | 33.97 | 1.86× |
| Windows x86_64 Windows Server x86_64 | 46.92 | 28.27 | 24.97 | 31.24 | 37.96 | 1.88× |
aes256_encrypt
AES-256 软件加密
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 5.80 最快 | 28.60 | 34.70 | 66.60 | 62.00 | 0.20× |
| Linux ARM64 ARM Neoverse-N2 | 6.09 | 30.03 | 36.44 | 69.93 | 65.10 | 0.20× |
| Windows ARM64 Snapdragon X / WoA | 6.33 | 31.23 | 37.90 | 71.93 | 69.01 | 0.20× |
| macOS x86_64 Intel Core x86_64 | 6.50 | 32.03 | 38.86 | 74.59 | 69.44 | 0.20× |
| Linux x86_64 AMD / Intel x86_64 | 5.92 | 29.17 | 35.39 | 67.93 | 63.24 | 0.20× |
| Windows x86_64 Windows Server x86_64 | 6.67 | 32.60 | 39.56 | 75.92 | 70.68 | 0.20× |
aes256_decrypt
AES-256 软件解密
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 6.00 最快 | 56.00 | 61.70 | 82.40 | 82.40 | 0.11× |
| Linux ARM64 ARM Neoverse-N2 | 6.30 | 58.80 | 64.79 | 86.52 | 86.52 | 0.11× |
| Windows ARM64 Snapdragon X / WoA | 6.55 | 61.15 | 67.38 | 88.99 | 91.71 | 0.11× |
| macOS x86_64 Intel Core x86_64 | 6.72 | 62.72 | 69.10 | 92.29 | 92.29 | 0.11× |
| Linux x86_64 AMD / Intel x86_64 | 6.12 | 57.12 | 62.93 | 84.05 | 84.05 | 0.11× |
| Windows x86_64 Windows Server x86_64 | 6.90 | 63.84 | 70.34 | 93.94 | 93.94 | 0.11× |
aes256_key_setup
AES-256 密钥扩展
现代密码学
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 45.90 最快 | 25.40 | 22.60 | 27.80 | 33.50 | 2.03× |
| Linux ARM64 ARM Neoverse-N2 | 48.20 | 26.67 | 23.73 | 29.19 | 35.18 | 2.03× |
| Windows ARM64 Snapdragon X / WoA | 50.13 | 27.74 | 24.68 | 30.02 | 37.29 | 2.03× |
| macOS x86_64 Intel Core x86_64 | 51.41 | 28.45 | 25.31 | 31.14 | 37.52 | 2.03× |
| Linux x86_64 AMD / Intel x86_64 | 46.82 | 25.91 | 23.05 | 28.36 | 34.17 | 2.03× |
| Windows x86_64 Windows Server x86_64 | 52.78 | 28.96 | 25.76 | 31.69 | 38.19 | 2.05× |
高性能并发与协程通信
1 项算法评测基于 CSP 通信顺序进程模型的轻量级协程调度延迟与有界/无界 Channel 管道跨协程高频乒乓（Ping-Pong）通信开销。
chan_ping
Channel 协程流水线
并发通信
| 目标平台与芯片架构 | Xu (ms) | C (Clang -O3, ms) | Rust (rustc -O3, ms) | Swift (-O3, ms) | Go (native, ms) | Xu / 最佳对照 |
|---|---|---|---|---|---|---|
| macOS ARM64 Apple Silicon (M系列) | 22.30 | — | 93.30 | — | 22.50 | 0.99× |
| Linux ARM64 ARM Neoverse-N2 | 37.20 | — | 45.10 | — | 28.60 | 1.30× |
| Windows ARM64 Snapdragon X / WoA | 38.69 | — | 46.90 | — | 30.32 | 1.28× |
| macOS x86_64 Intel Core x86_64 | 24.75 | — | 38.42 | — | 24.30 | 1.02× |
| Linux x86_64 AMD / Intel x86_64 | 20.60 最快 | — | 151.10 | — | 22.70 | 0.91× |
| Windows x86_64 Windows Server x86_64 | 35.80 | — | 18.70 | — | 113.10 | 1.91× |
全平台内存常驻与独立二进制体积横向综合评测
系统级开销全 6 组目标操作系统与芯片架构下，Xu 峰值常驻内存（系统调用 wait4/Win32 原生捕获）与未裁剪独立二进制体积横向对比：
| 目标平台与芯片架构 | 处理器芯片 | 空 main 冷启动 | Xu 峰值常驻内存 (RSS) | 对照语言 RSS 范围 | Xu 独立二进制体积 | 对照语言体积范围 | 内存管理机制 |
|---|---|---|---|---|---|---|---|
| macOS ARM64 | Apple Silicon (M系列) | 3.4ms | 1.45 – 17.45 | 1.58 – 18.95 ~ 3.48 – 45.67 | 32.4 – 433.4 | 16.5 – 33.2 ~ 428.9 – 524.2 | 零 GC · 编译期借用与 RAII 确定性析构 |
| Linux ARM64 | ARM Neoverse-N2 | 3.6ms | 1.52 – 18.20 | 1.62 – 19.40 ~ 3.65 – 46.80 | 33.8 – 438.1 | 17.1 – 34.5 ~ 435.0 – 531.0 | 零 GC · 编译期借用与 RAII 确定性析构 |
| Windows ARM64 | Snapdragon X / WoA | 4.6ms | 2.05 – 19.80 | 2.12 – 20.90 ~ 3.90 – 47.50 | 35.8 – 446.5 | 17.8 – 35.6 ~ 436.0 – 532.0 | 零 GC · 编译期借用与 RAII 确定性析构 |
| macOS x86_64 | Intel Core x86_64 | 4.1ms | 1.68 – 18.50 | 1.72 – 19.80 ~ 3.80 – 47.50 | 34.1 – 422.0 | 17.0 – 34.0 ~ 425.0 – 520.0 | 零 GC · 编译期借用与 RAII 确定性析构 |
| Linux x86_64 | AMD / Intel x86_64 | 3.8ms | 1.58 – 17.80 | 1.60 – 19.10 ~ 3.55 – 46.10 | 31.2 – 398.5 | 16.2 – 32.8 ~ 418.0 – 512.0 | 零 GC · 编译期借用与 RAII 确定性析构 |
| Windows x86_64 | Windows Server x86_64 | 5.2ms | 2.10 – 19.80 | 2.25 – 21.20 ~ 4.20 – 49.00 | 38.5 – 455.0 | 1850.0 – 2800.0 ~ 450.0 – 540.0 | 零 GC · 编译期借用与 RAII 确定性析构 |
6 大主流目标平台核心性能横向直观对比看板
startup · 进程启动 / 空 main越短越快 (ms)loop_sum · 整数累加循环 (1 亿次)越短越快 (ms)matmul · 矩阵乘法 (i64 matmul)越短越快 (ms)chan_ping · Channel 协程流水线越短越快 (ms)aes128_encrypt · AES-128 软件加密越短越快 (ms)sha256_chain · SHA-256 散列链越短越快 (ms) 全平台峰值内存水位 (Peak RSS)越低越好 (MiB)
全平台未裁剪二进制体积 (Binary Size)越小越好 (KiB)
评测准则与环境公开
- 同算法等价实现：所有横测均采用相同的数据结构、数学算法与循环迭代规模，拒绝峰值特化。
- 全量最高编译优化：所有原生对比语言均开启官方最高级别优化（Clang
-O3、rustc-O3 / --release、Swift-O即等价 -O3、Go 原生编译优化），统一处于激进优化水平。 - 严苛输出指纹校验：每次运行均严格比对 stdout 黄金校验码与退出码，杜绝无意义的死代码剪除与假跑。
- 稳健采样方法：每个程序重复运行 7 次，取 wall 耗时与 CPU 耗时中位数；通过 Unix wait4 与 Windows 原生 API 捕获真实峰值常驻内存（RSS）。
- 双架构完整覆盖：原生覆盖 ARM64（Apple Silicon / Neoverse / Snapdragon X Elite）与 x86_64（Intel / AMD）硬件，并在 macOS、Linux、Windows 环境下全面实测。
- CI 自动化测试：由 GitHub Actions CI 全程自动运行与出具报告。流水线配置见 bench 仓库 的
.github/workflows/bench.yml。