On Wed, Feb 04, 2026 at 09:59:22AM -0300, Gustavo Sousa wrote:
> Matt Roper <[email protected]> writes:
> 
> > On Mon, Feb 02, 2026 at 06:43:19PM -0300, Gustavo Sousa wrote:
> >> From: Shekhar Chauhan <[email protected]>
> >> 
> >> Add platform definition along with device IDs for NVL-P.
> >
> > Not sure if you wanted to add bspec references to this one like you had
> > for the Xe3p_LPG descriptor, but the flags you have here look correct to
> > me for now and the PCI IDs look correct so,
> >
> > Reviewed-by: Matt Roper <[email protected]>
> 
> Thanks!
> 
> I'll send v2 with the list of fields and Bspec references. I'm
> copying/pasting them here as well, just in case you can spot any errors
> before I post v2:
> 
>       .dma_mask_size (Bspec 74198)

This matches the way we've justified this field on past platforms,
although we've discovered that the "physical address range" value on
that page doesn't necessarily directly translate to the dma_mask_size we
need to use in software; there are cases like ADL-S where it isn't an
exact match due to other characteristics of the base platform (MKTE and
such).  But that extra information isn't really available in the
graphics bspec, so I don't think we have any better reference that can
be given.

>       .has_cached_pt (Bspec 71582)
>       .has_display (Bspec 74196)
>       .has_flat_ccs (Bspec 65255)

Not sure if this is the best page for FlatCCS since it doesn't really
indicate that a platform does/doesn't support it (e.g., the page doesn't
drop out when the bspec is filtered to platforms like CRI that lack
compression).

A better reference might be 74110 since it has a yes/no on compression
support by platform.


Matt

>       .has_page_reclaim_hw_assist (Bspec 73451)
>       .max_gt_per_tile (Bspec 74196)
>       .va_bits (Bspec 74198)
>       .vm_max_level (Bspec 59507)
> 
> --
> Gustavo Sousa

-- 
Matt Roper
Graphics Software Engineer
Linux GPU Platform Enablement
Intel Corporation

Reply via email to