Re: [PATCH v3] eventdev: ensure 16-byte alignment for events

[Bruce Richardson <bruce.richardson@intel.com>]

On Fri, Oct 06, 2023 at 12:35:26PM +0200, Morten Brørup wrote:
> > From: Bruce Richardson [mailto:bruce.richardson@intel.com]
> > Sent: Friday, 6 October 2023 12.17
> ev: ensure 16-byte alignment for events
> > 
> > On Fri, Oct 06, 2023 at 12:13:54PM +0200, Morten Brørup wrote:
> > > > From: Bruce Richardson [mailto:bruce.richardson@intel.com]
> > > > Sent: Friday, 6 October 2023 11.45
> > > >
> > > > The event structure in DPDK is 16-bytes in size, and events are
> > > > regularly passed as parameters directly rather than being passed as
> > > > pointers. To help compiler optimize correctly, we can explicitly request
> > > > 16-byte alignment for events, which means that we should be able
> > > > to do aligned vector loads/stores (e.g. with SSE or Neon) when working
> > > > with those events.
> > > >
> > > > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > > > Acked-by: Morten Brørup <mb@smartsharesystems.com>
> > > > Acked-by: Jerin Jacob <jerinj@marvell.com>
> > > >
> > > > ---
> > >
> > > [...]
> > >
> > > > +_Static_assert(sizeof(struct rte_event) == 16, "Event structure size is
> > not 16-bytes in size");
> > >
> > > Thank you for adding this extra check. We should have more of these.
> > >
> > > NB: _Static_assert is deprecated in C23 [1], so for forward compatibility,
> > you could use static_assert (which is available in <assert.h>) instead. Nice
> > to have; feel free to ignore this comment.
> > >
> > Is the availability in assert.h backward compatible with C11, since the
> > link you posted seems to imply that "static_assert" is only from C23
> > onwards?
> 
> Yes, the link mentions "static_assert" being available for C11 as a convenience macro in assert.h.
> 
> I had to read the link very carefully to get this. I guess I'm not the only one. :-)
> 

I missed that in the link, but I did have a read of the header file itself
to see that there was a macro there with C11 guards. :-)

> I don't object to moving it to the .c file. However, I think it's convenient for readability to have the static_assert close to the thing it checks, and/or close to any code that relies on the assumption it checks.
> 
I'm ok either way, and happy enough to have the check in the C file. I
suppose it saves cluttering up the public header file with checks that
should not be relevant to the end-user, and are only for DPDK developers.

Anyway, v4 sent, hopefully with all concerns addressed.

/Bruce