guenthert on Cosmopolitan Libc: your build-once run-anywhere C library

Comment on Cosmopolitan Libc: your build-once run-anywhere C library

"That's a huge improvement in generated code size. The above two compiles used the same gcc flags"

It would have been awfully nice to state the version of the compiler and the flags being used. With gcc 10 on Linux/AMD64, libc 4.15.0, I get with '-Os':

  --8<--
  strlcpy:
  .LFB5:
        .cfi_startproc
        pushq   %rbp
        .cfi_def_cfa_offset 16
        .cfi_offset 6, -16
        movq    %rdi, %rbp
        movq    %rsi, %rdi
        pushq   %rbx
        .cfi_def_cfa_offset 24
        .cfi_offset 3, -24
        movq    %rdx, %rbx
        subq    $24, %rsp
        .cfi_def_cfa_offset 48
        movq    %rsi, 8(%rsp)
        call    strlen
        testq   %rbx, %rbx
        movq    8(%rsp), %rsi
        je      .L1
        leaq    -1(%rbx), %rdx
        movq    %rbp, %rdi
        cmpq    %rax, %rdx
        cmova   %rax, %rdx
        movq    %rdx, %rcx
        movq    %rdx, %rcx
        rep movsb
        movb    $0, 0(%rbp,%rd  
  .L1ubq    $24, %rsp
        .cfi_def_cfa_offset 48
        movq    %rsi, 8(%rsp) 
        call    strlen
        testq   %rbx, %rbx
        movq    8(%rsp), %rsi
        je      .L1
        leaq    -1(%rbx), %rdx
        movq    %rbp, %rdi
        cmpq    %rax, %rdx
        cmova   %rax, %rdx
        movq    %rdx, %rcx
        cmova   %rax, %rdx
        movq    %rdx, %rcx
        addq    $24, %rsp
        .cfi_def_cfa_offset 24
        popq    %rbx
        .cfi_def_cfa_offset 16
        popq    %rbp
        .cfi_def_cfa_offset 8
        ret
        .cfi_endproc
  -->8--

Which doesn't seem so bad.

gary_0 5y

GCC optimizations are also smart enough to remove memcpy calls for small numbers of bytes entirely and just output the unrolled movs/loads inline. If you're memcpy'ing a struct from a byte buffer and reading some fields, the optimizer is usually smart enough to output the minimum instructions necessary. As I recall, MSVC's optimizations perform similarly.

Comments