最近一位朋友问我“为什么同样的hello world 入门程序”为什么golang编译出来的二进制文件,比 C 大,而且大很多。我做了个测试,来分析这个问题。C 语言的hello world程序:
1
2
3
4
5
|
#include <stdio.h> int main() { printf ( "hello world!\n" ); return 0; } |
golang 语言的hello world程序:
1
2
3
4
5
6
7
|
package main import "fmt" func main() { fmt.Printf( "hello, world!\n" ) } |
编译,查看生成文件大小
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
|
root@cnxct: /home/cfc4n/go_vs_c # gcc -o c.out main.c root@cnxct: /home/cfc4n/go_vs_c # go build -o go_fmt.out main_fmt.go root@cnxct: /home/cfc4n/go_vs_c # ll total 1552 drwxr-xr-x 2 root root 4096 Sep 20 16:56 ./ drwxr-xr-x 8 cfc4n cfc4n 4096 Sep 20 16:54 ../ -rwxr-xr-x 1 root root 8600 Sep 20 16:56 c.out* -rwxr-xr-x 1 root root 1560062 Sep 20 16:56 go_fmt.out* -rw-r--r-- 1 root root 78 Sep 20 16:54 main.c -rw-r--r-- 1 root root 78 Sep 20 16:55 main_fmt.go root@cnxct: /home/cfc4n/go_vs_c # du -sh * 12K c.out 1.5M go_fmt.out 4.0K main.c 4.0K main_fmt.go |
正如这位朋友所说c.out是12K,而 go_fmt.out是1.5M,差距奇大无比….为什么呢?
这两个二进制可执行文件文件里,都包含了什么?
众所周知,linux 上的二进制可执行文件是 ELF Executable and Linkable Format 可执行和可链接格式
如上图,ELF 文件分为如下:
- ELF文件的组成:ELF header
- 程序头:描述段信息
- Section头:链接与重定位需要的数据
- 程序头与Section头需要的数据.text .data
在 Linux 上, 查看elf格式构成可以使用readelf
ELF Header:头的信息
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
|
root@cnxct: /home/cfc4n/go_vs_c # readelf -h c.out ELF Header: Magic: 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 Class: ELF64 Data: 2‘s complement, little endian Version: 1 (current) OS /ABI : UNIX - System V ABI Version: 0 Type: EXEC (Executable file ) Machine: Advanced Micro Devices X86-64 Version: 0x1 Entry point address: 0x400430 Start of program headers: 64 (bytes into file ) Start of section headers: 6616 (bytes into file ) Flags: 0x0 Size of this header: 64 (bytes) Size of program headers: 56 (bytes) Number of program headers: 9 Size of section headers: 64 (bytes) Number of section headers: 31 Section header string table index: 28 root@cnxct: /home/cfc4n/go_vs_c # readelf -h go_fmt.out ELF Header: Magic: 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 Class: ELF64 Data: 2‘s complement, little endian Version: 1 (current) OS /ABI : UNIX - System V ABI Version: 0 Type: EXEC (Executable file ) Machine: Advanced Micro Devices X86-64 Version: 0x1 Entry point address: 0x44e360 Start of program headers: 64 (bytes into file ) Start of section headers: 456 (bytes into file ) Flags: 0x0 Size of this header: 64 (bytes) Size of program headers: 56 (bytes) Number of program headers: 7 Size of section headers: 64 (bytes) Number of section headers: 23 Section header string table index: 3 |
ELF 头的长度都是一样的,不会带来总体体积的变化。区别是个别字节的值不一样,比如Entry point address 程序入口点的值不一样等。
接下来是 程序头:,也就是 section部分(在linker连接器的角度是section部分或者装载器角度的segment)
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
|
root@cnxct: /home/cfc4n/go_vs_c # readelf -d c.out Dynamic section at offset 0xe28 contains 24 entries: Tag Type Name /Value 0x0000000000000001 (NEEDED) Shared library: [libc.so.6] 0x000000000000000c (INIT) 0x4003c8 0x000000000000000d (FINI) 0x4005b4 0x0000000000000019 (INIT_ARRAY) 0x600e10 0x000000000000001b (INIT_ARRAYSZ) 8 (bytes) 0x000000000000001a (FINI_ARRAY) 0x600e18 0x000000000000001c (FINI_ARRAYSZ) 8 (bytes) 0x000000006ffffef5 (GNU_HASH) 0x400298 0x0000000000000005 (STRTAB) 0x400318 0x0000000000000006 (SYMTAB) 0x4002b8 0x000000000000000a (STRSZ) 61 (bytes) 0x000000000000000b (SYMENT) 24 (bytes) 0x0000000000000015 (DEBUG) 0x0 0x0000000000000003 (PLTGOT) 0x601000 0x0000000000000002 (PLTRELSZ) 48 (bytes) 0x0000000000000014 (PLTREL) RELA 0x0000000000000017 (JMPREL) 0x400398 0x0000000000000007 (RELA) 0x400380 0x0000000000000008 (RELASZ) 24 (bytes) 0x0000000000000009 (RELAENT) 24 (bytes) 0x000000006ffffffe (VERNEED) 0x400360 0x000000006fffffff (VERNEEDNUM) 1 0x000000006ffffff0 (VERSYM) 0x400356 0x0000000000000000 (NULL) 0x0 |
可以看到c.out里引用了一个动态链接库libc.so.6,再看下go_fmt.out的情况
1
2
3
|
root@cnxct: /home/cfc4n/go_vs_c # readelf -d go_fmt.out There is no dynamic section in this file . |
c.out的执行,依赖了libc.so.6, libc.so.6肯定需要ld.so的,看下依赖情况,
1
2
3
4
5
6
7
|
root@cnxct: /home/cfc4n/go_vs_c # ldd c.out linux-vdso.so.1 => (0x00007fff3a195000) libc.so.6 => /lib/x86_64-linux-gnu/libc .so.6 (0x00007f4ac4d06000) /lib64/ld-linux-x86-64 .so.2 (0x0000558ece3fe000) root@cnxct: /home/cfc4n/go_vs_c # ldd go_fmt.out not a dynamic executable |
也就是说,C的程序默认使用了libc.so动态链接库,go 的程序,默认进行了静态编译,不依赖任何动态链接库。所以体积变大了。 那么,只是这一个原因吗?
我在 golang 的官方文档里找到如下的解释:
Why is my trivial program such a large binary?
The linker in the gc tool chain creates statically-linked binaries by default. All Go binaries therefore include the Go run-time, along with the run-time type information necessary to support dynamic type checks, reflection, and even panic-time stack traces.A simple C “hello, world” program compiled and linked statically using gcc on Linux is around 750 kB, including an implementation of printf. An equivalent Go program using fmt.Printf is around 1.5 MB, but that includes more powerful run-time support and type information.
将c的程序也使用静态编译试试。。。
1
2
3
4
5
6
7
|
gcc -static -o c_static.out main.c root@cnxct: /home/cfc4n/go_vs_c # du -sh * 12K c.out 888K c_static.out 1.5M go_fmt.out 4.0K main.c 4.0K main_fmt.go |
可以看到,使用静态编译生成的二进制文件c_static.out为888K,仍然比 GO 写的小了一半,这到底是为什么呢?到底是哪里多了?
在ELF 可执行文件里,就需要以程序编译链接的角度来分析了,对于一个 ELF 文件的分析,上面部分分析过 ELF header部分,以及 dynamic section的情况了。再以看一下剩余的section信息。
ELF中的section主要提供给Linker使用, 而segment提供给Loader用,Linker需要关心.text, .rel.text, .data, .rodata等等,关键是Linker需要做relocation。而Loader只需要知道这个段的Read、Write、Execute的属性。
再去看go_fmt.out里都包含了什么,为了方便校对,写了一个程序来对比
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
|
package main import ( "debug/elf" "fmt" "os" ) func main() { if len(os.Args) != 3 { fmt.Println( "参数不对" ) os.Exit(0) } strFile1 := os.Args[1] strFile2 := os.Args[2] f1, e := elf.Open(strFile1) if e != nil { panic(e) } f2, e := elf.Open(strFile2) if e != nil { panic(e) } mapSection1 := make(map[string]string, 0) mapSection2 := make(map[string]string, 0) //[Nr] Name Type Address Offset Size EntSize Flags Link Info Align var size1 uint64 var size2 uint64 for _, s := range f1.Sections { mapSection1[s.Name] = fmt.Sprintf( "%s\t%s\t%s\t%010x\t%010x\t%d\t%x\t%s\t%x\t%x\t%x\t" , s.Name, strFile1, s.Type.String(), s.Addr, s.Offset, s.Size, s.Entsize, s.Flags.String(), s.Link, s.Info, s.Addralign) size1 += s.Size } for _, s := range f2.Sections { mapSection2[s.Name] = fmt.Sprintf( "%s\t%s\t%s\t%010x\t%010x\t%d\t%x\t%s\t%x\t%x\t%x\t" , s.Name, strFile2, s.Type.String(), s.Addr, s.Offset, s.Size, s.Entsize, s.Flags.String(), s.Link, s.Info, s.Addralign) size2 += s.Size } fmt.Println(fmt.Sprintf( "%s:%d\t%s:%d" , strFile1, size1, strFile2, size2)) fmt.Println( "Name\tFile\tType\tAddress\tOffset\tSize\tEntSize\tFlags\tLink\tInfo\tAlign" ) for k, v := range mapSection1 { fmt.Println(v) if v1, found := mapSection2[k]; found { fmt.Println(v1) delete (mapSection2, k) } } for _, v := range mapSection2 { fmt.Println(v) } } |
对比一下两个文件的section段信息
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
|
root@cnxct: /home/cfc4n/go_vs_c # ./diffelf c_static.out go_fmt.out c_static.out:910462 go_fmt.out:1674012 Name File Type Address Offset Size EntSize Flags Link Info Align .init_array c_static.out SHT_INIT_ARRAY 00006c8ed8 00000c8ed8 16 0 SHF_WRITE+SHF_ALLOC 00 8 .fini_array c_static.out SHT_FINI_ARRAY 00006c8ee8 00000c8ee8 16 0 SHF_WRITE+SHF_ALLOC 00 8 .data c_static.out SHT_PROGBITS 00006c9080 00000c9080 6864 0 SHF_WRITE+SHF_ALLOC 0 020 .data go_fmt.out SHT_PROGBITS 00004fa4e0 00000fa4e0 7440 0 SHF_WRITE+SHF_ALLOC 0 020 .strtab c_static.out SHT_STRTAB 0000000000 00000d6b40 26703 0 0x0 0 0 1 .strtab go_fmt.out SHT_STRTAB 0000000000 00001704e0 51486 0 0x0 0 0 1 __libc_subfreeres c_static.out SHT_PROGBITS 00004bd6a8 00000bd6a8 80 0 SHF_ALLOC 00 8 __libc_thread_subfreeres c_static.out SHT_PROGBITS 00004bd708 00000bd708 8 0 SHF_ALLOC 0 0 8 .got c_static.out SHT_PROGBITS 00006c8fe8 00000c8fe8 16 8 SHF_WRITE+SHF_ALLOC 0 08 .comment c_static.out SHT_PROGBITS 0000000000 00000cab50 52 1 SHF_MERGE+SHF_STRINGS 00 1 .fini c_static.out SHT_PROGBITS 00004a0470 00000a0470 9 0 SHF_ALLOC+SHF_EXECINSTR 0 04 .eh_frame c_static.out SHT_PROGBITS 00004bd710 00000bd710 44948 0 SHF_ALLOC 0 08 .bss c_static.out SHT_NOBITS 00006cab60 00000cab50 6264 0 SHF_WRITE+SHF_ALLOC 0 020 .bss go_fmt.out SHT_NOBITS 00004fc200 00000fc200 108808 0 SHF_WRITE+SHF_ALLOC 0 020 .gcc_except_table c_static.out SHT_PROGBITS 00004c86a4 00000c86a4 179 0 SHF_ALLOC 00 1 .tdata c_static.out SHT_PROGBITS 00006c8eb8 00000c8eb8 32 0 SHF_WRITE+SHF_ALLOC+SHF_TLS 00 8 .note.gnu.build- id c_static.out SHT_NOTE 00004001b0 00000001b0 36 0 SHF_ALLOC 00 4 __libc_freeres_fn c_static.out SHT_PROGBITS 000049de60 000009de60 9513 0 SHF_ALLOC+SHF_EXECINSTR 0 0 10 .plt c_static.out SHT_PROGBITS 00004002f0 00000002f0 160 0 SHF_ALLOC+SHF_EXECINSTR 0 010 .note.stapsdt c_static.out SHT_NOTE 0000000000 00000cab84 3864 0 0x0 0 0 4 .symtab c_static.out SHT_SYMTAB 0000000000 00000cbaa0 45216 18 0x0 20 2c7 8 .symtab go_fmt.out SHT_SYMTAB 0000000000 0000164000 50400 18 0x0 16 5f 8 .note.ABI-tag c_static.out SHT_NOTE 0000400190 0000000190 32 0 SHF_ALLOC 0 04 .rela.plt c_static.out SHT_RELA 00004001d8 00000001d8 240 18 SHF_ALLOC+SHF_INFO_LINK018 8 .rodata c_static.out SHT_PROGBITS 00004a0480 00000a0480 119332 0 SHF_ALLOC 0 0 20 .rodata go_fmt.out SHT_PROGBITS 000047e000 000007e000 212344 0 SHF_ALLOC 0 0 20 .data.rel.ro c_static.out SHT_PROGBITS 00006c8f00 00000c8f00 228 0 SHF_WRITE+SHF_ALLOC 00 20 .init c_static.out SHT_PROGBITS 00004002c8 00000002c8 26 0 SHF_ALLOC+SHF_EXECINSTR 0 04 .text c_static.out SHT_PROGBITS 0000400390 0000000390 645828 0 SHF_ALLOC+SHF_EXECINSTR 0 010 .text go_fmt.out SHT_PROGBITS 0000401000 0000001000 508779 0 SHF_ALLOC+SHF_EXECINSTR 0 010 __libc_atexit c_static.out SHT_PROGBITS 00004bd6f8 00000bd6f8 8 0 SHF_ALLOC 0 08 .stapsdt.base c_static.out SHT_PROGBITS 00004bd700 00000bd700 1 0 SHF_ALLOC 0 01 .jcr c_static.out SHT_PROGBITS 00006c8ef8 00000c8ef8 8 0 SHF_WRITE+SHF_ALLOC 0 08 c_static.out SHT_NULL 0000000000 0000000000 0 0 0x0 0 0 0 go_fmt.out SHT_NULL 0000000000 0000000000 0 0 0x0 0 0 0 __libc_thread_freeres_fn c_static.out SHT_PROGBITS 00004a0390 00000a0390 222 0 SHF_ALLOC+SHF_EXECINSTR 0 0 10 __libc_freeres_ptrs c_static.out SHT_NOBITS 00006cc3d8 00000cab50 48 0 SHF_WRITE+SHF_ALLOC 0 0 8 .shstrtab c_static.out SHT_STRTAB 0000000000 00000dd38f 361 0 0x0 0 0 1 .shstrtab go_fmt.out SHT_STRTAB 0000000000 00000b1d80 257 0 0x0 0 0 1 .tbss c_static.out SHT_NOBITS 00006c8ed8 00000c8ed8 48 0 SHF_WRITE+SHF_ALLOC+SHF_TLS 00 8 .got.plt c_static.out SHT_PROGBITS 00006c9000 00000c9000 104 8 SHF_WRITE+SHF_ALLOC 00 8 .itablink go_fmt.out SHT_PROGBITS 00004b29d8 00000b29d8 56 0 SHF_ALLOC 0 08 .gopclntab go_fmt.out SHT_PROGBITS 00004b2a20 00000b2a20 282414 0 SHF_ALLOC 0 020 .debug_abbrev go_fmt.out SHT_PROGBITS 000051c000 00000fd000 255 0 0x0 0 0 1 .debug_frame go_fmt.out SHT_PROGBITS 000052b2d5 000010c2d5 69564 0 0x0 0 0 1 .debug_aranges go_fmt.out SHT_PROGBITS 000054634c 000012734c 48 0 0x0 0 0 1 .debug_info go_fmt.out SHT_PROGBITS 00005463a6 00001273a6 248638 0 0x0 0 0 1 .note.go.buildid go_fmt.out SHT_NOTE 0000400fc8 0000000fc8 56 0 SHF_ALLOC 00 4 .debug_pubtypes go_fmt.out SHT_PROGBITS 000053e9e5 000011f9e5 31079 0 0x0 0 0 1 .debug_gdb_scripts go_fmt.out SHT_PROGBITS 000054637c 000012737c 42 0 0x0 0 01 .debug_line go_fmt.out SHT_PROGBITS 000051c0ff 00000fd0ff 61910 0 0x0 0 0 1 .typelink go_fmt.out SHT_PROGBITS 00004b1ea0 00000b1ea0 2872 0 SHF_ALLOC 0 020 .noptrdata go_fmt.out SHT_PROGBITS 00004f8000 00000f8000 9416 0 SHF_WRITE+SHF_ALLOC 00 20 .gosymtab go_fmt.out SHT_PROGBITS 00004b2a10 00000b2a10 0 0 SHF_ALLOC 0 01 .noptrbss go_fmt.out SHT_NOBITS 0000516b20 0000116b20 18080 0 SHF_WRITE+SHF_ALLOC 00 20 .debug_pubnames go_fmt.out SHT_PROGBITS 000053c291 000011d291 10068 0 0x0 0 0 1 |
发现go_fmt.out多了好多.debug_*开头的 section,这是用于 debug 的段信息。再次编译,去除这些信息,同时也把 C 静态编译的二进制也去除符号表和重定位信息。
1
2
3
4
5
6
7
8
9
|
root@cnxct: /home/cfc4n/go_vs_c # gcc -static -o c_static_gs.out -g -s main.c root@cnxct: /home/cfc4n/go_vs_c # go build -o go_fmt_sw.out -ldflags="-s -w" main_fmt.go root@cnxct: /home/cfc4n/go_vs_c # du -sh * 12K c.out 820K c_static_gs.out 888K c_static.out 1.5M go_fmt.out 1012K go_fmt_sw.out |
如上结果,go_fmt_sw.out为1012K,c_static_gs.out为820K,还大了近200KB。到底是哪里大的呢?
刚刚的两个elf 文件的section对比中,还有一个比较特殊的go_fmt.out中 有一个名字叫.gopclntab的段,类型是SHT_PROGBITS程序段,大小为 282414字节,也就是275K,在c_static.out里并没有这个段的,也没有.gosymtab这个段。二者不一样,section段名字有规范标准吗?
其实,对于linker链接器来说,会关心段(section)的名字,但对loader加载器来说,并不关心名字,只关心这个段(segment)的权限,是否可执行,所在的偏移地址,用于函数的执行。
那.gopclntab段包含了什么内容呢?我写了一个程序分析了这个段的内容,程序代码如下:
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
|
package main import ( "debug/elf" "debug/gosym" "fmt" "os" ) func main() { if len(os.Args) != 2 { fmt .Println( "参数不对" ) os.Exit(0) } strFile1 := os.Args[1] f1, err := elf.Open(strFile1) if err != nil { panic(err) } symtab, err := f1.Section( ".gosymtab" ).Data() if err != nil { f1.Close() panic( ".gosymtab 异常" ) } gopclntab, err := f1.Section( ".gopclntab" ).Data() if err != nil { f1.Close() panic( ".gopclntab 异常" ) } pcln := gosym.NewLineTable(gopclntab, f1.Section( ".text" ).Addr) var tab *gosym.Table tab, err = gosym.NewTable(symtab, pcln) if err != nil { f1.Close() panic(err) } for _, x := range tab.Funcs { fmt .Println( fmt .Sprintf( "addr:0x%x\t\tname:%s,\t" ,x.Entry,x.Name)) } } |
编译后执行
01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
|
root@cnxct: /home/cfc4n/go_vs_c # ./expelf go_fmt.out addr:0x401000 name: sync /atomic .StoreUint32, addr:0x401010 name: sync /atomic .StoreUint64, addr:0x401020 name: sync /atomic .StoreUintptr, addr:0x401030 name:runtime.memhash0, addr:0x401040 name:runtime.memhash8, ...... addr:0x427040 name:runtime.printnl, ...... addr:0x44dc80 name:runtime.memmove, addr:0x44e360 name:_rt0_amd64_linux, addr:0x44e380 name:main, addr:0x44e390 name:runtime. exit , addr:0x44ea70 name:runtime.epollwait, addr:0x44ea90 name:runtime.(*cpuProfile).(runtime.flushlog)-fm, addr:0x44eae0 name: type .. hash .runtime.uncommontype, ...... addr:0x452d60 name:math.init.1, addr:0x452e00 name:math.init, addr:0x452e70 name:math.hasSSE4, addr:0x452e90 name: type .. hash .[70]float64, addr:0x452f10 name: type .. eq .[70]float64, addr:0x452f50 name:errors.New, addr:0x452ff0 name:errors.(*errorString).Error, ...... addr:0x4534c0 name:unicode /utf8 .RuneCountInString, addr:0x453600 name:strconv.(*decimal).String, addr:0x453a00 name:strconv.digitZero, addr:0x453a30 name:strconv.trim, addr:0x453aa0 name:strconv.(*decimal).Assign, ...... addr:0x4599c0 name:strconv.init, addr:0x459ad0 name: type .. hash .strconv.decimal, addr:0x459ec0 name: type .. eq .[61]strconv.leftCheat, addr:0x459f80 name: sync .(*Mutex).Lock, ...... addr:0x45c6d0 name:syscall.Syscall6, addr:0x45c740 name: type .. hash .[133]string, addr:0x45c7c0 name: type .. eq .[133]string, addr:0x45c880 name: time .init, addr:0x45df30 name: type .. hash .os.PathError, addr:0x45dfc0 name: type .. eq .os.PathError, addr:0x45e0e0 name:reflect.makeMethodValue, addr:0x45eaf0 name:reflect.resolveReflectName, addr:0x45eb40 name:reflect.(*rtype).nameOff, ...... addr:0x4710d0 name:reflect.(*funcTypeFixed64).Comparable, addr:0x4710f0 name: type .. hash .reflect.funcTypeFixed128, addr:0x471170 name: type .. eq .reflect.funcTypeFixed128, addr:0x471230 name:reflect.(*funcTypeFixed128).uncommon, ...... addr:0x471c50 name:reflect.(*sliceType).Comparable, addr:0x471c70 name: type .. hash .struct { reflect.b bool; reflect.x interface {} }, addr:0x471cf0 name: type .. eq .struct { reflect.b bool; reflect.x interface {} }, addr:0x471d70 name: type .. hash .[27]string, addr:0x471df0 name: type .. eq .[27]string, addr:0x471ea0 name: fmt .(* fmt ).writePadding, addr:0x472020 name: fmt .(* fmt ).pad, ...... addr:0x47b730 name: fmt .(*pp).badArgNum, addr:0x47b940 name: fmt .(*pp).missingArg, addr:0x47bb50 name: fmt .(*pp).doPrintf, addr:0x47cf70 name: fmt .glob..func1, addr:0x47cfd0 name: fmt .init, addr:0x47d170 name: type .. hash . fmt . fmt , addr:0x47d1f0 name: type .. eq . fmt . fmt , addr:0x47d2a0 name:main.main, addr:0x47d310 name:main.init, |
如上可以看到,有很多函数是以fmt.(*pp)、strconv.*、sync.*、reflect.*、unicode.*等开头的,后面对应的函数名,也与 golang 的包里对应的包中函数名一致。。。用 IDA来确认一遍
,果然在 .gopclntab 段里有很多 reflect.*开头的函数。
这就很奇怪了,golang 编译时,默认把 runtime 包编译进来就好了,应该不会把strconv\sync\reflect\unicode等包包含进来啊。程序中,只写了一句fmt.Println(),莫非是fmt包import了其他几个包导致的?回去搜了下代码,果然…
嗯,应该是这里问题,改用 go 的内置函数print试试。
1
2
3
4
5
|
package main func main() { print( "hello, world!\n" ) } |
编译后,对比大小
01
02
03
04
05
06
07
08
09
10
11
12
13
|
root@cnxct: /home/cfc4n/go_vs_c # go build -o go_print.out main_print.go root@cnxct: /home/cfc4n/go_vs_c # go build -o go_print_sw.out -ldflags="-s -w" main_print.go root@cnxct: /home/cfc4n/go_vs_c # du -sh * 12K c.out 820K c_static_gs.out 888K c_static.out 1.5M go_fmt.out 1012K go_fmt_sw.out 940K go_print.out 624K go_print_sw.out 4.0K main.c 4.0K main_fmt.go 4.0K main_print.go |
看如上结果,go_print_sw.out 变成了 624K , c_static_gs.out为820K,不光没比C的静态编译的大,还比它小呢。。。 不过呢,这也不能说明什么问题,只是因为其包含的函数内容不一样。
好了,至此已经知道为什么 golang 编译的文件比 C 的大了,因为 go 语言是静态编译的,而 C 的编译(比如 gcc编译器)都是动态链接库形式编译的。所以,导致了 go 编译的文件稍微大的问题。其次,跟其他语言比较字符串输出的话,用print内置函数就好了,就不要使用fmt包下的函数来比较了,因为 fmt 包引入了好多其他的包。。。这也增加编译后的二进制文件的体积。
其实呢,golang 的编译(不涉及 cgo 编译的前提下)默认使用了静态编译,不依赖任何动态链接库,这样可以任意部署到各种运行环境,不用担心依赖库的版本问题。只是体积大一点而已,存储时占用了一点磁盘,运行时,多占用了一点内存。早期动态链接库的产生,是因为早期的系统的内存资源十分宝贵,由于内存紧张的问题在早期的系统中显得更加突出,因此人们首先想到的是要解决内存使用效率不高这一问题,于是便提出了动态装入的思想。也就产生了动态链接库。在现在的计算机里,操作系统的硬盘内存更大了,尤其是服务器,32G、64G 的内存都是最基本的。可以不用为了节省几百 KB 或者1M,几 M 的内存而大大费周折了。而 golang 就采用这种做法,可以避免各种 so 动态链接库依赖的问题,这点是非常值得称赞的。
原文: http://www.cnxct.com/why-golang-elf-binary-file-is-large-than-c/