[Intel-gfx] [PATCH 17/32] drm/i915: Remove the lazy_coherency parameter from request-completed?

[Chris Wilson]

On Mon, Dec 14, 2015 at 02:59:30PM +0000, Tvrtko Ursulin wrote:
> 
> Hi,
> 
> On 11/12/15 11:33, Chris Wilson wrote:
> >Now that we have split out the seqno-barrier from the
> >engine->get_seqno() callback itself, we can move the users of the
> >seqno-barrier to the required callsites simplifying the common code and
> >making the required workaround handling much more explicit.
> 
> What bothers me about this patch, and the one preceding it, is that
> I don't see a tangible improvement for the programmer who still has
> to know when to read the seqno and when to "read it harder, read for
> real".

In earlier patches, I called it irq_barrier.

It's not reading it harder. It's just that there is a ordering issue
with receiving an interrupt and the seqno write being visible.

> Barrier in this sense has a relation to the state of things but
> somehow feels too low level to me when used from the code. But to be
> fair I am not sure how to better define it.
> 
> Would ring->get_seqno paired with ring->read_seqno perhaps make
> sense? Implementation for ring->read_seqno would just be a flush
> followed by ring->get_seqno then. Or maybe keep the barrier and add
> ring->read_seqno which would be ring->seqno_barrier +
> ring_get_seqno?

No.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre