On 8/26/19 11:10 AM, Chris Wilson wrote:
Quoting Daniele Ceraolo Spurio (2019-08-26 18:56:53)
On 8/24/2019 1:54 AM, Chris Wilson wrote:
Note that this requires us to fix the tagging so that we don't perform a
lite-restore from the reloc instance to the user instance.
What's wrong with lite-restoring in this case? It's not something we
stop now AFAICS.
I have patches to fix the regression. But the question is whether or not
the pre-parser is happy to cross the lite-restore boundary. If it is a
light lite-restore and there is no cache invalidation...
-Chris
On pre-gen12 the parser doesn't cross the BB_START, so we're safe from
that POV.
Daniele
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
