[Bug 87131] [PNV regression] igt/gem_exec_lut_handle takes more than 10 minutes

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=87131

Chris Wilson <chris at chris-wilson.co.uk> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEEDINFO                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #15 from Chris Wilson <chris at chris-wilson.co.uk> ---
Two major improvements landed:

commit 4ff4b44cbb70c269259958cbcc48d7b8a2cb9ec8
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date:   Fri Jun 16 15:05:16 2017 +0100

    drm/i915: Store a direct lookup from object handle to vma

commit 2889caa9232109afc8881f29a2205abeb5709d0c
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date:   Fri Jun 16 15:05:19 2017 +0100

    drm/i915: Eliminate lots of iterations over the execobjects array

which used to be enough to not only fix the original regression but make a
substantial improvement.

However, since the regression was noticed, I added two more iterations to the
test (to exercise different paths through reloc) and we now interact with
dma-fence and reservation_objects which are a huge step backwards.

Resolving as the original regression is fixed. I've some patches to improve
scalabiltiy of reservation_objects, but the main cost of them are the atomics
we so painstakingly avoided previously (because of this bug highlighting how
expensive they were along this path).

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are on the CC list for the bug.
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20170616/04b5f068/attachment-0001.html>