This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH v2] GAS/MIPS: Add `-mfix-r5900' option for the R5900 short loop errata
- From: Fredrik Noring <noring at nocrew dot org>
- To: "Maciej W. Rozycki" <macro at linux-mips dot org>
- Cc: Nick Clifton <nickc at redhat dot com>, Chenghua Xu <paul dot hua dot gm at gmail dot com>, binutils at sourceware dot org, Jürgen Urban <JuergenUrban at gmx dot de>
- Date: Tue, 27 Nov 2018 20:30:45 +0100
- Subject: Re: [PATCH v2] GAS/MIPS: Add `-mfix-r5900' option for the R5900 short loop errata
- References: <17c89c52e61cf1f6c4f365c3c25e89d28f096062.1540632453.git.noring@nocrew.org> <alpine.LFD.2.21.1811271626120.32615@eddie.linux-mips.org>
Thank you for your review, Maciej!
> Please prepare a suitable ChangeLog entry. I have missed its absence in
> the initial review somehow (sorry about that), but since we need another
> iteration anyway, please include it alongside.
Done!
> However please set `mips_fix_r5900'
> appropriately in `mips_after_parse_args' instead, i.e.:
>
> if (!mips_fix_r5900_explicit)
> mips_fix_r5900 = file_mips_opts.arch == CPU_R5900;
>
> so that it's the only setting referred in determination as to whether to
> enable the workaround.
Done!
> This will affect temporary `.set arch=' overrides,
> but I think they are not supposed to override the global `-mfix-r5900'
> setting, whether specified or inferred.
Losing -mfix-r5900 inadvertently is fatal to the R5900 and will likely
cause corruption and any other imaginable error, so that must not happen.
This is also the main motivation to implement a warning for unfixable
cases involving for example the noreorder directive.
> This is otherwise OK, so please resubmit with just this issue addressed.
Great, will post v3 soon!
Fredrik