[cfe-users] Machine code footprint

[David Blaikie via cfe-users]

Looks like most of the difference can be accounted for by the fact that A
is not default constructible - which probably causes some more complicated
code in the standard library which the compiler isn't currently able to see
through. I haven't looked at the specifics of what that might be:

$ cat vec.cpp

#include <vector>


template<class T>

std::vector<T> f()

{

  std::vector<T> a;

  a.resize(10000, 1);

  return a;

}



class A

{

public:

  long a;



  *A() = default;*

  A(long a) : a(a) {}

};


template std::vector<long> f();

template std::vector<A> f();

$ clang++-tot -O3 vec.cpp -c && nm vec.o --size-sort -Ctd

0000000000000079 W std::vector<A, std::allocator<A> > f<A>()

0000000000000079 W std::vector<long, std::allocator<long> > f<long>()

0000000000001802 W std::vector<A, std::allocator<A>
>::_M_fill_insert(__gnu_cxx::__normal_iterator<A*, std::vector<A,
std::allocator<A> > >, unsigned long, A const&)

0000000000001898 W std::vector<long, std::allocator<long>
>::_M_fill_insert(__gnu_cxx::__normal_iterator<long*, std::vector<long,
std::allocator<long> > >, unsigned long, long const&)

On Sun, Jan 10, 2021 at 11:11 AM Marcel Keller via cfe-users <
cfe-users at lists.llvm.org> wrote:

> Hi,
>
> When I compile the attached code, I find that one instantiation of
> vector::_M_fill_insert is double the size of the other even though the
> semantics should be the same:
>
> $ clang++ -O3 -g -c insert.cpp
> $ nm insert.o  --size-sort -Ctd
> 0000000000000079 W std::vector<A, std::allocator<A> > f<A>()
> 0000000000000079 W std::vector<long, std::allocator<long> > f<long>()
> 0000000000001765 W std::vector<long, std::allocator<long>
>  >::_M_fill_insert(__gnu_cxx::__normal_iterator<long*, std::vector<long,
> std::allocator<long> > >, unsigned long, long const&)
> 0000000000003638 W std::vector<A, std::allocator<A>
>  >::_M_fill_insert(__gnu_cxx::__normal_iterator<A*, std::vector<A,
> std::allocator<A> > >, unsigned long, A const&)
>
> What is the reason for this? I would expect that a class trivially
> containing one member would be treated the same as just the member.
>
> I'm using clang 10 on Ubuntu 20.04.
>
> Best regards,
> Marcel
>
> _______________________________________________
> cfe-users mailing list
> cfe-users at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-users/attachments/20210110/4c9851a8/attachment.html>