> -----Original Message----- > From: Karol Lewandowski [mailto:k.lewandowsk@xxxxxxxxxxx] > Subject: Re: [PATCH v6 4/4] libusbg: Add inline to > gadget_write_string(). > > On 01/23/2014 11:13 AM, Krzysztof Opasiak wrote: > >> -----Original Message----- > >> From: David Laight [mailto:David.Laight@xxxxxxxxxx]> > >> From: Of Stanislaw Wadas > >>> Add inline to gadget_write_string(). > >> ... > >>> -static void usbg_write_string(char *path, char *name, char > >> *file, char *buf) > >>> +static inline void usbg_write_string(char *path, char *name, > >> char *file, char *buf) > >>> { > >>> usbg_write_buf(path, name, file, buf); > >>> } > >> > >> Why? > >> The compiler is very unlikely not to inline it anyway. > >> > >> If it doesn't inline it, then specifying 'inline' probably > >> won't make any difference either. > >> > > > > I agree that new version of gcc can do this without explicit > inline, but > > please think about some legacy systems and compilers. Older > versions of > > gcc or other compilers may not inline this function if there is > no > > keyword. We should write portable code and don't depend on any > version > > of compiler. So to do it good we should pass as much information > to > > compiler as possible. > > Given different optimization flags (-Os vs -O3) compiler can > generate > completely different code. Both will be "optimal" but by > completely > different criteria. > > We should not try to outsmart compiler, it's usually way smarter > than > us. As far as I know , compiler is not smart but only has a good heuristic. In my opinion this heuristic is based also on inline keyword. Let's take an example 2 simple functions: void func(int *a, int b) { int i =1; for(i=1; i < b; ++a); } static inline void func2(int* a, int b) { func(a, b); } Then in main we have a set of 7 call to this functions. Using my gcc 4.6.3 with -01 flag I get different asm if I use inline keyword or not. When using inline keyword I got func2 optimized out but if I don't use inline keyword func2 is still present in code and called each time. Both source and asm results you may found below this message. I hope that you see now that using or not using inline may influent the result of the compilation, so specifying inline makes a difference. -- BR's Krzysztof Opasiak Samsung R&D Institute Poland Samsung Electronics --- test.c: >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> void func(int *a, int b) { int i =1; a++; for(i=1; i < b; ++a); } static inline void func2(int* a, int b) { func(a, b); } int main(int argc, char** argv) { int a =(int) (long)argv[0]; int b =(int) (long)argv[1]; func2(&a, b); func2(&a, b); func2(&a, b); func2(&a, b); func2(&a, b); func2(&a, b); func2(&a, b); a++; return a; } >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> gcc -S test.c -O1 without func2 inline >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> .file "test.c" .text .globl func .type func, @function func: .LFB0: .cfi_startproc cmpl $1, %esi jle .L4 .L3: jmp .L3 .L4: rep ret .cfi_endproc .LFE0: .size func, .-func .type func2, @function func2: .LFB1: .cfi_startproc call func rep ret .cfi_endproc .LFE1: .size func2, .-func2 .globl main .type main, @function main: .LFB2: .cfi_startproc pushq %rbp .cfi_def_cfa_offset 16 .cfi_offset 6, -16 pushq %rbx .cfi_def_cfa_offset 24 .cfi_offset 3, -24 subq $16, %rsp .cfi_def_cfa_offset 40 movq %rsi, %rbx movq 8(%rsi), %rbp movl %ebp, %esi leaq 12(%rsp), %rdi call func2 movl %ebp, %esi leaq 12(%rsp), %rdi call func2 movl %ebp, %esi leaq 12(%rsp), %rdi call func2 movl %ebp, %esi leaq 12(%rsp), %rdi call func2 movl %ebp, %esi leaq 12(%rsp), %rdi call func2 movl %ebp, %esi leaq 12(%rsp), %rdi call func2 movl %ebp, %esi leaq 12(%rsp), %rdi call func2 movq (%rbx), %rax addl $1, %eax addq $16, %rsp .cfi_def_cfa_offset 24 popq %rbx .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2: .size main, .-main .ident "GCC: (Ubuntu/Linaro 4.6.3-1ubuntu5) 4.6.3" .section .note.GNU-stack,"",@progbit >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> gcc -S test.c -O1 with func2 inline >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> .file "test.c" .text .globl func .type func, @function func: .LFB0: .cfi_startproc cmpl $1, %esi jle .L4 .L3: jmp .L3 .L4: rep ret .cfi_endproc .LFE0: .size func, .-func .globl main .type main, @function main: .LFB2: .cfi_startproc pushq %rbp .cfi_def_cfa_offset 16 .cfi_offset 6, -16 pushq %rbx .cfi_def_cfa_offset 24 .cfi_offset 3, -24 subq $16, %rsp .cfi_def_cfa_offset 40 movq %rsi, %rbx movq 8(%rsi), %rbp movl %ebp, %esi leaq 12(%rsp), %rdi call func movl %ebp, %esi leaq 12(%rsp), %rdi call func movl %ebp, %esi leaq 12(%rsp), %rdi call func movl %ebp, %esi leaq 12(%rsp), %rdi call func movl %ebp, %esi leaq 12(%rsp), %rdi call func movl %ebp, %esi leaq 12(%rsp), %rdi call func movl %ebp, %esi leaq 12(%rsp), %rdi call func movq (%rbx), %rax addl $1, %eax addq $16, %rsp .cfi_def_cfa_offset 24 popq %rbx .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2: .size main, .-main .ident "GCC: (Ubuntu/Linaro 4.6.3-1ubuntu5) 4.6.3" .section .note.GNU-stack,"",@progbits >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
