On 11 July 2017 at 23:49, Alex Smith <[email protected]> wrote: > On 11 July 2017 at 14:27, Alex Smith <[email protected]> wrote: >> >> On 10 July 2017 at 05:59, Dave Airlie <[email protected]> wrote: >>> >>> From: Dave Airlie <[email protected]> >>> >>> This patch uses addrlib to workout the tile swizzles according >>> to the surface index. It seems to produce the same values as >>> amdgpu-pro for the deferred test. >>> >>> v2: don't apply swizzle to CMASK. the eg docs don't mention >>> it, and we clearly don't align cmask for that. >>> v3: disable surf index for dedicated images, as these will >>> most likely be shared, and I don't think the metadata has >>> space for this info in it yet. >> >> >> FWIW, disabling this for images marked as dedicated means this won't get >> any improvements for render targets on our games. We create all render >> targets as dedicated when NV_dedicated_allocation is available since this >> gets us significant perf improvement on NVIDIA. >> >> If it's not currently possible to have this enabled for dedicated images >> we could avoid using it on AMD, though I'm curious if there's likely to be >> any other perf benefits to marking RTs as dedicated we'd then be missing out >> on? I've not done any testing to see if there's any benefit from using it. > > > Realised this possibly didn't sound clear - what I'm asking is does using > NV_dedicated_allocation give any perf benefit on RADV at all like it does > for NV? If not we could avoid it to get the benefits from this patch.
No on radv we really only want dedicated allocation for shared resources i.e. external memory ones. I don't think you'd get any speed up from using it. I'd like to figure out a way to make it work with this feature, but I think it means enhancing the metadata a bit. Dave. _______________________________________________ mesa-dev mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/mesa-dev
