[Intel-xe] [PATCH] drm/xe/: add gt tuning for indirect state

[Lionel Landwerlin]

On 15/04/2023 01:11, Lucas De Marchi wrote:
> On Fri, Apr 14, 2023 at 02:59:55PM -0700, Lucas De Marchi wrote:
>> On Fri, Apr 14, 2023 at 12:33:55PM -0700, Matt Atwood wrote:
>>> Force indirect state sampler data to only be in the dynamic state pool,
>>> which is more convienent for UMD. Behavior change mirrors windows
>>> behavior.
>>
>> Better than the last sentence would be something like:
>>
>> Behavior change mirrors similar change for i915 in commit 16fc9c08f0ec
>> ("drm/i915: disable sampler indirect state in bindless heap").
>>
>>
>> BSpec: 46052
>>
>>>
>>> Signed-off-by: Matt Atwood <matthew.s.atwood at intel.com>
>>> ---
>>> drivers/gpu/drm/xe/regs/xe_gt_regs.h | 1 +
>>> drivers/gpu/drm/xe/xe_tuning.c       | 5 +++++
>>> 2 files changed, 6 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/xe/regs/xe_gt_regs.h 
>>> b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
>>> index f174758d8307..b7324aede6b6 100644
>>> --- a/drivers/gpu/drm/xe/regs/xe_gt_regs.h
>>> +++ b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
>>> @@ -286,6 +286,7 @@
>>> #define   ENABLE_SMALLPL            REG_BIT(15)
>>> #define   SC_DISABLE_POWER_OPTIMIZATION_EBB    REG_BIT(9)
>>> #define   GEN11_SAMPLER_ENABLE_HEADLESS_MSG    REG_BIT(5)
>>> +#define   INDIRECT_STATE_BASE_ADDR_OVERRIDE    REG_BIT(0)
>>>
>>> #define GEN9_HALF_SLICE_CHICKEN7        MCR_REG(0xe194)
>>> #define   DG2_DISABLE_ROUND_ENABLE_ALLOW_FOR_SSLA REG_BIT(15)
>>> diff --git a/drivers/gpu/drm/xe/xe_tuning.c 
>>> b/drivers/gpu/drm/xe/xe_tuning.c
>>> index 27cf1330facd..5abef45a10fc 100644
>>> --- a/drivers/gpu/drm/xe/xe_tuning.c
>>> +++ b/drivers/gpu/drm/xe/xe_tuning.c
>>> @@ -26,6 +26,11 @@ static const struct xe_rtp_entry gt_tunings[] = {
>>>       XE_RTP_RULES(PLATFORM(DG2)),
>>>       XE_RTP_ACTIONS(SET(XEHP_SQCM, EN_32B_ACCESS))
>>>     },
>>> +    { XE_RTP_NAME("Tuning: Set Indirect State Override"),
>>> +      XE_RTP_RULES(GRAPHICS_VERSION_RANGE(1200, -1)),
>>
>> as noted by José, since we have that define let's use it:
>> XE_RTP_END_VERSION_UNDEFINED
>>
>> with those changes:
>>
>>     Reviewed-by: Lucas De Marchi <lucas.demarchi at intel.com>
>
> taking that back, sorry. Just noticed this setting needs to be
> re-applied on engine reset, so gt_tunings is not the right place.
> Please add
> 1) engine_tunings[] array
> 2) xe_tuning_process_engine (to be called everywhere
> xe_wa_process_engine() is called.
>
> This can be done as a preparatory commit so we don't mix the addition of
> this workaround with the additional infra.
>
> 3) Make sure to add FUNC(xe_rtp_match_first_render_or_compute) to the
> the rtp rules
>
> Lucas De Marchi


Do we need to fix i915 for this too?


Thanks,


-Lionel



>
>>
>> Lucas De Marchi
>>
>>> + XE_RTP_ACTIONS(SET(GEN10_SAMPLER_MODE, 
>>> INDIRECT_STATE_BASE_ADDR_OVERRIDE,
>>> +                 XE_RTP_ACTION_FLAG(MASKED_REG)))
>>> +    },
>>>     {}
>>> };
>>>
>>> -- 
>>> 2.39.2
>>>