This is the mail archive of the
libc-ports@sources.redhat.com
mailing list for the libc-ports project.
Re: [PATCH] power-cpu add-on for --with-cpu
- From: Daniel Jacobowitz <drow at false dot org>
- To: Steven Munroe <munroesj at us dot ibm dot com>
- Cc: Roland McGrath <roland at redhat dot com>, libc-ports at sources dot redhat dot com
- Date: Wed, 30 Nov 2005 17:57:20 -0500
- Subject: Re: [PATCH] power-cpu add-on for --with-cpu
- References: <438E2E79.2030000@us.ibm.com>
On Wed, Nov 30, 2005 at 04:58:01PM -0600, Steven Munroe wrote:
> Updated patch with working HWCAP based libraries search. This took a bit
> of effort as I did find an endian problem. Turns out the the bits are
> numbered differently in big-endian as well (bit-0 is the high order
> bit). But the dl-sysdep.c code is counting from the low order bit of the
> dl_hwcap. This required some extra padding and rearranging of the
> _dl_powerpc_cap_flags array.
>
> Also work out an agreement with David Edelsohn for gcc to generate
> appropriate _ARCH_? defines to match the -mcpu setting. Use this to make
> the ./bits/mathinline.h safe for the default case.
>
> Also also added an new optimized version of memcpy.S for power4/power5.
>
> While there are more optimizations in the works, I think we have the
> critical mass to commit this code as an add-on/port in cvs.
I don't get it. This is a set of optimizations for supported CPUs for
the PowerPC port, which is not an add-on. Why should it live in ports?
The nofpu bits were moved here because they were "insufficiently
mainstream".
--
Daniel Jacobowitz
CodeSourcery, LLC