Matt Roper <[email protected]> writes:

> On Mon, Feb 02, 2026 at 02:11:04PM -0800, Matt Roper wrote:
>> On Mon, Feb 02, 2026 at 06:43:07PM -0300, Gustavo Sousa wrote:
>> > From: Shekhar Chauhan <[email protected]>
>> > 
>> > Add graphics IP versions 35.10. Since there will be features enabled
>> > that are not present in graphics_xe2, already create a new
>> > graphics_xe3p, although currently it only enables the same features as
>> > graphics_xe2.
>> > 
>> > Here is a list of fields, associated Bspec references and eventual
>> > comments:
>> > 
>> >  .va_bits (Bspec 74198)
>> >  .vm_max_level (Bspec 59507)
>> >     The spec says "The PPGTT is always a 5-level structure even when the
>> >     virtual address space is less than 57 bits."
>> >  .vram_flags
>> >    - ~XE_VRAM_FLAGS_NEED64K
>> >      That limitation that does not exist for Xe3p_LPG.
>> 
>> These three aren't part of the graphics descriptor anymore, so they're
>> not relevant to this patch.
>> 
>> >  .hw_engine_mask (Bspec 60149)
>> >  .tile_gsm_size (Bspec 52961)
>> >     This is queried directly from the hardware.
>> 
>> I don't think this flag has ever existed, outside of some temporary
>> development patches that never moved forward.
>> 
>> >  .has_asid (Bspec 71132)
>> >  .has_atomic_enable_pte_bit (Bspec 59510, 74675)
>> >  .has_flat_ccs (Bspec 65255)
>> 
>> This one also isn't part of the graphics descriptor anymore.
>> 
>> >  .has_indirect_ring_state (Bspec 67296)
>> >  .has_range_tlb_inval (Bspec 71126)
>> >  .has_usm (Bspec 59651)
>> > 
>> 
>> We should probably add a reference for has_64bit_timestamp (bspec 60318)
>> since it's set in XE2_GFX_FEATURES.
>> 
>> Is there any reason not to go ahead and squash in
>> multi_queue_engine_class_mask from patch #12 here?
>> 
>> 
>> Matt
>> 
>> > Signed-off-by: Shekhar Chauhan <[email protected]>
>> > Signed-off-by: Gustavo Sousa <[email protected]>
>> > ---
>> >  drivers/gpu/drm/xe/xe_pci.c | 5 +++++
>> >  1 file changed, 5 insertions(+)
>> > 
>> > diff --git a/drivers/gpu/drm/xe/xe_pci.c b/drivers/gpu/drm/xe/xe_pci.c
>> > index b5e8935fff1d..edca562462fa 100644
>> > --- a/drivers/gpu/drm/xe/xe_pci.c
>> > +++ b/drivers/gpu/drm/xe/xe_pci.c
>> > @@ -106,6 +106,10 @@ static const struct xe_graphics_desc graphics_xe2 = {
>> >    XE2_GFX_FEATURES,
>> >  };
>> >  
>> > +static const struct xe_graphics_desc graphics_xe3p = {
>
> One more comment: we may want to name this "graphics_xe3p_lpg" to help
> distinguish it from graphics_xe3p_xpc below.

Incorporated all of the suggestions and they will be present in v2.

Thanks!

>
>
> Matt
>
>> > +  XE2_GFX_FEATURES,
>> > +};
>> > +
>> >  static const struct xe_graphics_desc graphics_xe3p_xpc = {
>> >    XE2_GFX_FEATURES,
>> >    .has_indirect_ring_state = 1,
>> > @@ -148,6 +152,7 @@ static const struct xe_ip graphics_ips[] = {
>> >    { 3003, "Xe3_LPG", &graphics_xe2 },
>> >    { 3004, "Xe3_LPG", &graphics_xe2 },
>> >    { 3005, "Xe3_LPG", &graphics_xe2 },
>> > +  { 3510, "Xe3p_LPG", &graphics_xe3p },
>> >    { 3511, "Xe3p_XPC", &graphics_xe3p_xpc },
>> >  };
>> >  
>> > 
>> > -- 
>> > 2.52.0
>> > 
>> 
>> -- 
>> Matt Roper
>> Graphics Software Engineer
>> Linux GPU Platform Enablement
>> Intel Corporation
>
> -- 
> Matt Roper
> Graphics Software Engineer
> Linux GPU Platform Enablement
> Intel Corporation

Reply via email to