Alexey,
Hopefully the gcc developers might manage to get the LTO
branch merged for gcc 4.4. They have some benchmarks in figure 7
of the article "Interprocedural optimization framework in GCC"
which suggests that whole program optimizations currently can
yield as much as ~30% speed improvements.
Probably it won't be merged, but llvm-project already provides ways to
build the backend for the current gcc.
I actually tried it locally — it works. But, I guess, I am not the
right person to package the beast for fink
On 9/12/07, Jack Howarth <[EMAIL PROTECTED]> wrote:
> Alexey,
>The proble
Alexey,
The problem is that llvm is unlikely to be merged into
gcc for 4.3 (and I'm not even sure it will make gcc 4.4).
Actually a more interesting change slated for gcc 4.4
is the replacement of the haifa scheduler. It is described
in the gcc 2007 summit proceedings...
https://ols2006.108.red
Another interesting thing to test would be llvm-based gcc ;)
On 9/11/07, Jack Howarth <[EMAIL PROTECTED]> wrote:
>A release candidate of gcc 4.2.2 has been posted by
> the gcc maintainer. I have posted packaging for
> gcc42-4.2.2-1000 to the fink tracker using the RC
> tarball for gcc 4.2.2. T
A release candidate of gcc 4.2.2 has been posted by
the gcc maintainer. I have posted packaging for
gcc42-4.2.2-1000 to the fink tracker using the RC
tarball for gcc 4.2.2. The final release tarballs are
slated for a week from now at which time I'll update
the packaging to those.
J
