fuhuizn

第一次,站长亲自招 Gopher 了>>>

我用50亿字节的内存写入操作测试了`go`和`rust`的性能,发现`go`的内存操作比`rust`快! 我并不是一次性分配50亿字节内存,而是每次调用函数写入100000(大约100K不到一点)字节,反复调用5万次。我测试了两种写入方式,一种是初始化预分配内存的,另一种是初始化不分配内存的。结果是预分配模式下,`go`大大领先`rust`,速度达到`rust`的1.78倍;不预先分配内存模式下,双方性能相当。 `go`源代码(`app.go`和`app_test.go`) ``` //app.go package main import ( "fmt" "time" ) const L = 100000 const N = 50000 //预分配内存模式 func preAppend() []byte { buf := make([]byte, 0, L) for i := 0; i < L; i++ { buf = append(buf, 's') } return buf } //不分配内存模式 func normalAppend() []byte { buf := make([]byte, 0) for i := 0; i < L; i++ { buf = append(buf, 's') } return buf } func main() { start := time.Now() for i := 0; i < N; i++ { preAppend() } fmt.Printf("preAppend Time:%v\n", time.Now().Sub(start).Seconds()) start = time.Now() for i := 0; i < N; i++ { normalAppend() } fmt.Printf("normalAppend Time:%v\n", time.Now().Sub(start).Seconds()) } //app_test.go package main import ( "testing" ) func TestNormalAppend(t *testing.T) { for i := 0; i < N; i++ { normalAppend() } println("TestNormalAppend") } func TestPreAppend(t *testing.T) { for i := 0; i < N; i++ { preAppend() } println("TestPreAppend") } ``` `rust`源代码: ``` //main.rs use std::time; const N: isize = 50000; const L: usize = 100000; fn main() { let start = time::SystemTime::now(); for _i in 0..N { pre_append(); } let end = time::SystemTime::now(); println!("pre_append Time:{:?}", end.duration_since(start).unwrap()); let start = time::SystemTime::now(); for _i in 0..N { normal_append(); } let end = time::SystemTime::now(); println!( "normal_append Time:{:?}", end.duration_since(start).unwrap() ); } //不分配内存模式 fn normal_append() -> String { let mut buf = String::new(); let mut i: usize = 0; while i < L { buf.push('s'); i += 1; } return buf; } //预分配内存模式 fn pre_append() -> String { let mut buf = String::with_capacity(L); let mut i: usize = 0; while i < L { buf.push('s'); i += 1; } return buf; } #[cfg(test)] mod tests { use crate::normal_append; use crate::pre_append; use std::fmt::Error; const N: isize = 50000; #[test] fn test_pre_append() -> Result<(), Error> { for _i in 0..N { pre_append(); } Ok(()) } #[test] fn test_normal_append() -> Result<(), Error> { for _i in 0..N { normal_append(); } Ok(()) } } ``` ### 下面是各种模式下的运行结果: #### 测试模式的比拼 `go`语言: ``` $ go test -v . === RUN TestNormalAppend TestNormalAppend --- PASS: TestNormalAppend (11.51s) === RUN TestPreAppend TestPreAppend --- PASS: TestPreAppend (6.63s) PASS ok app_test_bench 18.146s ``` 运行 `TestNormalAppend`的时间是11.51秒。 运行`TestPreAppend`的时间是6.63秒。 `rust`语言: ``` $ cargo t Compiling string_test v0.1.0 (/home/fuhz/src/rust_test/string_test) Finished test [unoptimized + debuginfo] target(s) in 0.37s Running unittests src/main.rs (target/debug/deps/string_test-cf0723c9e88581f5) running 2 tests test tests::test_normal_append has been running for over 60 seconds test tests::test_pre_append has been running for over 60 seconds test tests::test_pre_append ... ok test tests::test_normal_append ... ok test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 78.64s ``` `test_normal_append`对应`go`程序中的`TestNormalAppend`,`test_pre_append`对应`go`程序中的`TestPreAppend`,两者的运行时间都超过了60秒,无法显示精确时间。 #### Debug模式的比拼 ``` $ go build $ ./app_test_bench preAppend Time:5.864950476 normalAppend Time:12.123574635 cargo build Compiling string_test v0.1.0 (/home/fuhz/src/rust_test/string_test) Finished dev [unoptimized + debuginfo] target(s) in 0.27s $ target/debug/string_test pre_append Time:69.011713959s normal_append Time:69.313903563s ``` `go`程序中,`preAppend`用时5.865秒,`normalAppend`用时12.123秒;`rust`程序中,`pre_append`用时69.012秒,`normal_append`用时69.314秒。 #### Release模式的比拼 ``` $ go build -ldflags '-s -w' $ ./app_test_bench preAppend Time:6.072494913 normalAppend Time:10.464082479 $ cargo build --release Compiling string_test v0.1.0 (/home/fuhz/src/rust_test/string_test) Finished release [optimized] target(s) in 0.30s $ target/release/string_test pre_append Time:10.794002015s normal_append Time:10.808211814s ``` `go`程序中,`preAppend`用时6.072秒,`normalAppend`用时10.464秒;`rust`程序中,`pre_append`用时10.794秒,`normal_append`用时10.808秒。

有疑问加站长微信联系(非本文作者))

3224 次点击

上一篇:在Go中使用Energy创建跨平台GUI应用 - 事件通信(IPC) 之 Go定义事件JS调用

下一篇:代理IP是按照流量计费还是数量计费好?

测试 分配内存 代码 分配模式

• 编辑
• 预览