This is the mail archive of the
mailing list for the binutils project.
Re: [Patch, AArch64, BFD, LD] [2.24] Cortex-A53 erratum 835769 linker workaround.
- From: Marcus Shawcroft <marcus dot shawcroft at gmail dot com>
- To: Tejas Belagod <tejas dot belagod at arm dot com>, gingold at adacore dot com
- Cc: "binutils at sourceware dot org" <binutils at sourceware dot org>, Marcus Shawcroft <Marcus dot Shawcroft at arm dot com>, Richard Earnshaw <Richard dot Earnshaw at arm dot com>
- Date: Fri, 24 Oct 2014 11:38:36 +0100
- Subject: Re: [Patch, AArch64, BFD, LD] [2.24] Cortex-A53 erratum 835769 linker workaround.
- Authentication-results: sourceware.org; auth=none
- References: <5447BA3A dot 9070308 at arm dot com>
On 22 October 2014 15:07, Tejas Belagod <email@example.com> wrote:
> Some early revisions of the Cortex-A53 have an erratum (835769) whereby
> it is possible for a 64-bit multiply-accumulate instruction in AArch64 state
> to generate an incorrect result. The details are quite complex and hard to
> determine statically, since branches in the code may exist in some
> circumstances, but all cases end with a memory (load, store, or
> prefetch) instruction followed immediately by the multiply-accumulate
> We employ a linker patching technique, whereby we move the potentially
> affected multiply-accumulate instruction into a patch region and
> replace the original instruction with a branch to the patch.
> This patch achieves the linker patch. Also, a patch to turn this on from the
> compiler driver has been submitted independently to the gcc lists.
> The fix is disabled by default and can be turned on by the
> --fix-cortex-a53-835769 linker option.
> Tested on aarch64-none-linux-gnu (and bootstrap with the option) and built
> and run various large benchmarks with it.
> Ok for 2.24?
This should be back ported. However, with 2.25 imminent I want to
make sure 2.25 gets back ported before/at the same time as 2.24. I'd
like an explicit OK from Tristan for the back port to 2.25.