[libfabric-users] [ofiwg] Shared Rx and Connected

[Xiong, Jianxin]

That's also what in my mind. The "source address" reference merely refers to the name of the argument in the fi_cq_readfrom() call.

With the special flag bit set, the src_addr would contain the endpoint handle that the message is received on.

> -----Original Message-----
> From: Doug Oucharek <dougso at me.com>
> Sent: Tuesday, September 10, 2019 12:25 PM
> To: Hefty, Sean <sean.hefty at intel.com>
> Cc: Xiong, Jianxin <jianxin.xiong at intel.com>; libfabric-
> users at lists.openfabrics.org
> Subject: Re: [ofiwg] [libfabric-users] Shared Rx and Connected
> 
> Speaking as a user of this proposed feature, I’d be more interested in the EP
> than the address.  With the address, I’d have to hash it to get a local context
> representing the remote side. I can get that from the EP without hashing making
> it a better approach.
> 
> Doug
> 
> On Sep 10, 2019, at 12:18 PM, Hefty, Sean <sean.hefty at intel.com> wrote:
> 
> >> One possibility is to define a new bit in the "flags" field of the
> >> returned CQ entries to indicate the source address is for connected endpoints.
> >
> > This is possible.  Though maybe rather than identifying the source of the
> message, we could identify the destination EP that processed the message.