[Intel-gfx] [PATCH v4 08/38] drm/i915: Prepare retire_requests to handle out-of-order seqnos

[Jesse Barnes]

On 01/11/2016 02:10 PM, Chris Wilson wrote:
> On Mon, Jan 11, 2016 at 06:42:37PM +0000, John.C.Harrison at Intel.com wrote:
>> From: John Harrison <John.C.Harrison at Intel.com>
>>
>> A major point of the GPU scheduler is that it re-orders batch buffers
>> after they have been submitted to the driver. This leads to requests
>> completing out of order. In turn, this means that the retire
>> processing can no longer assume that all completed entries are at the
>> front of the list. Rather than attempting to re-order the request list
>> on a regular basis, it is better to simply scan the entire list.
> 
> This is a major misstep. Just think in terms of per-context timelines,
> and retirment order within those timelines being consistent..

I think you're just re-iterating the desire for per-context timelines.  We all want this, but after already implementing several big reworks, I don't think it's fair to make John do this.  When we move that direction after this lands, we can simplify/drop code where possible (sure maybe it's more total churn, but my guess is we'll probably find other things to refactor as well, so it doesn't matter too much).

Jesse