This is the mail archive of the
mailing list for the GCC project.
Re: [6/7] Explicitly classify vector loads and stores
- From: Richard Sandiford <rdsandiford at googlemail dot com>
- To: Richard Biener <richard dot guenther at gmail dot com>
- Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>, richard dot sandiford at arm dot com
- Date: Sun, 03 Jul 2016 18:10:48 +0100
- Subject: Re: [6/7] Explicitly classify vector loads and stores
- Authentication-results: sourceware.org; auth=none
- References: <87d1nin8hz dot fsf at e105548-lin dot cambridge dot arm dot com> <87lh26ltp4 dot fsf at e105548-lin dot cambridge dot arm dot com> <CAFiYyc26DTF2i27QUKBeD_S-2q13oqqOdB-Xx=ziiBkhHvg7Mg at mail dot gmail dot com>
Richard Biener <email@example.com> writes:
> On Wed, Jun 15, 2016 at 10:52 AM, Richard Sandiford
> <firstname.lastname@example.org> wrote:
>> This is the main patch in the series. It adds a new enum and routines
>> for classifying a vector load or store implementation.
>> Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
> Why's the setting and checking of the memory access type conditional on !slp?
> I'd rather avoid doing this :/
For loads we need it for hybrid SLP, since we can vectorise the
same load twice, once for SLP and once not. (See e.g. pr62075.c.)
For stores it was unnecessary cut-&-paste.
Is it OK with the !slp restricted to vectorizable_load?