This is the mail archive of the
libc-ports@sources.redhat.com
mailing list for the libc-ports project.
Re: [PATCH roland/arm-memcpy] ARM: Make armv7 memcpy implementations SFI-friendly
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: Roland McGrath <roland at hack dot frob dot com>
- Cc: <libc-ports at sourceware dot org>
- Date: Tue, 18 Jun 2013 22:20:17 +0000
- Subject: Re: [PATCH roland/arm-memcpy] ARM: Make armv7 memcpy implementations SFI-friendly
- References: <20130618220149 dot CB94E2C0E4 at topped-with-meat dot com>
On Tue, 18 Jun 2013, Roland McGrath wrote:
> * sysdeps/arm/arm-features.h (ARM_BX_NINSNS): New macro.
> * sysdeps/arm/armv7/multiarch/memcpy_impl.S: Macroize the
> computed-jump dispatch sections. Use sfi_breg throughout.
> [ARM_ALWAYS_BX]: Define a different version of the dispatch macros
> that uses bx rather than add-to-pc, and respects ARM_BX_ALIGN_LOG2.
> [!USE_NEON] (D_l, D_h): Use r10, r11 rather than r8, r9.
> (tmp2): Use r8 rather than r10.
This is OK with the comments referring to an assembler bug amended to give
a binutils bug number (and preferably an indication of for what version it
was fixed) to make it easier in future to tell whether the workaround can
be removed because of a higher minimum binutils version requirement.
--
Joseph S. Myers
joseph@codesourcery.com