This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
optimization/4131: Why the C++ compiler don't place a const class object to ".rodata" section?
- To: gcc-gnats at gcc dot gnu dot org
- Subject: optimization/4131: Why the C++ compiler don't place a const class object to ".rodata" section?
- From: k_satoda at f2 dot dion dot ne dot jp
- Date: 26 Aug 2001 16:36:55 -0000
- Reply-To: k_satoda at f2 dot dion dot ne dot jp
>Number: 4131
>Category: optimization
>Synopsis: Why the C++ compiler don't place a const class object to ".rodata" section?
>Confidential: no
>Severity: non-critical
>Priority: medium
>Responsible: unassigned
>State: open
>Class: change-request
>Submitter-Id: net
>Arrival-Date: Sun Aug 26 09:46:01 PDT 2001
>Closed-Date:
>Last-Modified:
>Originator: Zutoshi
>Release: unknown-2.9
>Organization:
>Environment:
various
>Description:
I want to use a 'fixed-pointed-value' as a replacement of 'floating-point-value'.
But the const instance of my class never be placed to ".rodata" section.
They seem to need ".ctors" , however the constructing code has only a constant instructions to store a certain value.
I think it dosen't need any codes , and needs only few bytes of ".rodata" section.
I tried same test on some compilers , but no one generate the codes I want.
These can be a headache on machines that have little RAM.
Please excuse my poor english typing.(I am a Japanese.)
>How-To-Repeat:
// compile the following with maximam optimization.
class T
{
int raw;
public:
enum { BASE = (1<< 8) };
template<class value_type> T(value_type opr) : raw(static_cast<int>(opr * BASE)) {}
template<class value_type> operator value_type() const { return static_cast<value_type>(raw) / BASE; }
};
const T t1 = 1.0;
const int i1 = static_cast<int>(1.0 * T::BASE);
int main(void)
{
return i1 ^ static_cast<int>(t1);
}
>Fix:
>Release-Note:
>Audit-Trail:
>Unformatted: