[Bug 103517] 10% perf drop in CPU bound tests with "i965: Call gen6_upload_push_constants() even when the stage is disabled"

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=103517

            Bug ID: 103517
           Summary: 10% perf drop in CPU bound tests with "i965: Call
                    gen6_upload_push_constants() even when the stage is
                    disabled"
           Product: Mesa
           Version: git
          Hardware: Other
                OS: All
            Status: NEW
          Severity: normal
          Priority: medium
         Component: Drivers/DRI/i965
          Assignee: intel-3d-bugs at lists.freedesktop.org
          Reporter: eero.t.tamminen at intel.com
        QA Contact: intel-3d-bugs at lists.freedesktop.org

ezBench bisected following commit:
-------------------------------------------------------
commit 66b4a7a79e3da03e11233acd32b6e37bf3969dd8
Author:     Kenneth Graunke <kenneth at whitecape.org>
AuthorDate: Fri Oct 20 15:53:50 2017 -0700
Commit:     Kenneth Graunke <kenneth at whitecape.org>
CommitDate: Tue Oct 24 16:14:04 2017 -0700

    i965: Call gen6_upload_push_constants() even when the stage is disabled.

    This properly sets stage_state->push_constant_dirty = true, so that we
    emit 3DSTATE_CONSTANT_XS to disable the constant buffer for the shader
    stage.  It also sets stage_state->push_const_size = 0.

    Reviewed-by: Topi Pohjolainen <topi.pohjolainen at intel.com>
-------------------------------------------------------

As the cause for ~10% performance drop in following CPU bound 3D tests:
- GfxBench v4 Driver2 onscreen/offscreen
- SynMark v7 Batch6, Batch7 & Multithread

Most visible the drop was on SKL i5 6600K (it's never TDP limited), then SKL
GT4e & BDW GT2/GT3.  On other platforms it's visible in trends, but impact is a
bit smaller and variance is too large to bisect it well.

There were no related performance improvements visible.

When looking at the system state before and after the change, main difference
is that CPU usage seems to have moved slightly from user-space to something
requested from kernel side, which makes sense for this commit.

-- 
You are receiving this mail because:
You are the assignee 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-3d-bugs/attachments/20171030/ed313460/attachment.html>