On 2026-10-05 18:57:38+02:00, Boris Brezillon wrote: > On Mon, 5 Oct 2026 17:05:14 +0100 > Steven Price <[email protected]> wrote: > > > On 05/10/2026 16:37, Boris Brezillon wrote: > > > > Ah, yes that's true :) Although there's no need for us to do an > > invalidate on the unmap path... > > My bad, I'm confusing Panfrost and Panthor here. There's indeed no L2 > flush+inval on unmap in Panfrost. We do have an L2 flush+inval on map > (FLUSH_PT) that should cover this case though.
But we're doing panfrost_mmu_flush_range() both at the end of maps and unmaps, so I wonder whether when you say we do L2 flush+inval on map, what you meant was when submit a new job, we force a flush+invalidate for the AS the job runs against.
