[Intel-gfx] [PATCH v3] drm/i915: Enable WaProgramMgsrForCorrectSliceSpecificMmioReads for Gen9

[Tvrtko Ursulin]

On 11/03/2021 01:27, Chiou, Cooper wrote:
>> From: Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com>
>> Hi,
>>
>> On 08/03/2021 17:32, Chiou, Cooper wrote:
>>> I've tested on GLK, KBL, CFL Intel NUC devices and got the following
>> performance results, there is no performance regression per my testing.
>>>
>>> Patch: [v5] drm/i915: Enable
>>> WaProgramMgsrForCorrectSliceSpecificMmioReads for Gen9 Test suite:
>>> phoronix-test-suite.supertuxkart.1024x768.Fullscreen.Ultimate.1.GranPa
>>> radisoIsland.frames_per_second
>>> Kernel version: 5.12.0-rc1 (drm-tip)
>>>
>>> a. Device: Intel NUC kit NUC7JY Gemini Lake Celeron J4005 @2.7GHz (2
>> Cores)
>>>       Without patch, fps=57.45
>>>       With patch, fps=57.49
>>> b. Device: Intel NUC kit NUC8BEH Coffee Lake Core i3-8109U @3.6GHz(4
>> Cores)
>>>       Without patch, fps=117.23
>>>       With patch, fps=117.27
>>> c. Device: Intel NUC kit NUC7i3BNH Kaby Lake Core i3-7100U @2.4GHz(4
>> Cores)
>>>       Without patch, fps=114.05
>>>       With patch, fps=114.34
>>>
>>> Meanwhile, Intel lkp team has validated performance on lkp-kbl-nuc1 and
>> no regression.
>>> f69d02e37a85645a  d912096c40cdc3bc9364966971
>>> testcase/testparams/testbox
>>> ----------------  -------------------------- ---------------------------
>>>             %stddev      change         %stddev
>>>                 \          |                \
>>>         29.79                       29.67
>>> phoronix-test-suite/performance-true-Fullscreen-Ultimate-1-
>> Gran_Paradiso_Island__Approxima-supertuxkart-1.5.2-ucode=0xde/lkp-kbl-
>> nuc1
>>>         29.79                       29.67        GEO-MEAN phoronix-test-
>> suite.supertuxkart.1280x1024.Fullscreen.Ultimate.1.GranParadisoIsland.fram
>> es_per_second
>>>
>>
>> CI results are green so that is good.
>>
>> Do the machines used for performance testing include unusual fusing?
> [Chiou, Cooper] Yes, this performance test included fusing sku as following NUC GLK Celeron Linux device,
> Intel NUC kit NUC7JY Gemini Lake Celeron J4005 @2.7GHz (2 Cores)
>      Without patch, fps=57.45
>      With patch, fps=57.49

I was referring to the original performance regression report which came 
from LKP which was on "i7-7567U" so Kabylake. Can yo find such machine 
to test on?

>> Worrying thing is that we were never able to reproduce the reported regression in house due lack of identical machine, right?
> [Chiou, Cooper] Yes, if device is core-i3/5/7 GT2, then hang issue isn’t able to reproduce due to there is no fused/disabled subslice0.
> 
> But VP8 hw encoding GPU hang issue can be 100% reproduced on CML/KBL Pentium/Celeron GT1 sku,
> explain root cause as the following,
> on Chrome CML Pentium 6450u GT1 sku, we observed when system exit PC7 power state, MGSR(0xFDC) is reset to 0x10000000,
> read its reg(fuse2:0x9120)=0x02988480 bit 20-23=1001 means subslice0/3 are fused disabled, it's defined in bspec as well.
> on CML core-i5 sku, reg(fuse2)=0x22889140 bit 20-23=1000 only subslice3 is fused.
> 
> so we have to skip this "fused/disabled" subslice 0/3 on GT1 sku in i915, then use sublice 1 or 2 and re-program 0xFDC=0x11000000 or 0x12000000
> since reg[0xfdc]=0x10000000 to use subslice0, but sublice0 is fused/disabled, so once go this fused disabled subslice0 then GPU hang happened.
> 
> In some OEM Linux projects, they only have core-i sku no Pentium/Celeron sku as I knew and vp8 hw encoding is rare to use on Ubuntu,
> as this two reasons then Linux didn't reproduce this fused subslice0 gpu hang issue before, but it does happen on Chrome projects.

Understood on the hang part.

Regards,

Tvrtko