From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jason Gunthorpe Subject: Re: [PATCH RFC] libibverbs: add ARM64 memory barrier macros Date: Tue, 24 May 2016 11:22:50 -0600 Message-ID: <20160524172250.GD8037@obsidianresearch.com> References: <20160518220302.81260E09E9@smtp.ogc.us> <20160518222857.GB23835@obsidianresearch.com> <060701d1b1e7$dc7e1ff0$957a5fd0$@opengridcomputing.com> <20160519180552.GA26130@obsidianresearch.com> <065c01d1b1ff$e1e149f0$a5a3ddd0$@opengridcomputing.com> <20160519192805.GA32668@obsidianresearch.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Gabriele Svelto Cc: Steve Wise , dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: linux-rdma@vger.kernel.org On Fri, May 20, 2016 at 11:44:14AM +0200, Gabriele Svelto wrote: > On 19/05/2016 21:28, Jason Gunthorpe wrote: > > Yes, we probably should have had a 'memcpy to wc' verbs helper that > > did the best possible wc copy and prompt flush rather than the mess > > with wc_wmb and open coded memcpy. IIRC there is some way to do this > > with xmm non temporal cache line stores that is even better??? > > Non-temporal stores will go straight to memory IIRC. With many modern > HCAs being able to read data directly out of the last-level cache of the > processor they could end up being a suboptimal choice. We are not talking about memory backed write for DMA, but batching a MMIO write into a 64 byte burst. non-temporal stores would seem to be the best way to do that? Jason -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html