This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: Performance of copy algorithm
- From: jlquinn at optonline dot net
- To: Benjamin Kosnik <bkoz at redhat dot com>
- Cc: libstdc++ at gcc dot gnu dot org
- Date: Tue, 04 Feb 2003 15:52:32 -0500
- Subject: Re: Performance of copy algorithm
----- Original Message -----
From: Benjamin Kosnik <bkoz@redhat.com>
Date: Tuesday, February 4, 2003 1:15 pm
Subject: Re: Performance of copy algorithm
>
> >The case B above apparently makes use of the
memcpy builtin.
> Changing>the innards of __copy_trivial makes calls
to the library
> (verified>with ltrace).
> >
> >Suggestions on how I should proceed here?
>
> Perhaps this is obvious, but did you try calling
__builtin_memcpy
> directly?
Yup. Made absolutely no difference. I started
trying to look at the implementation of the memcpy
builtin. I'm not really sure what's going on, but
it seems to ask if the arguments are pointers and if
not, bail out to the library. But, it's an internal
notion of pointer and I'm not familiar with the
internal representation, so this comment could be a
red herring.
Jerry