I can't replay the trace locally so i won't be able to debug the issue with that. Can you record one with dxvk directly? Since this is probably a shader issue it should be fine.
That said, I can't spot any obvious error in the commit that introduced this regression. I don't want to revert it either because it'll break other games.
This one starts to replay at least. Unfortunately the replay crashes for me before showing the issue. Hopefully it's more useful to you. 2¾ hours to record this thing. :persevere:
I should also note I recorded this trace with a "bad" commit, 0bf9799, rather than a working one. I didn't realize until 20 minutes in (when I reached the benchmark..) and then figured I'd just make another with a working commit when it finished. Didn't expect it to take quite so long.
I'll make another if needed.
Thanks. I'm currently trying to debug another regression caused by that commit in Kingdom Come: Deliverance, which may or may not be related. I'll give you an update once I figure this out.
I should also note I recorded this trace with a "bad" commit, 0bf9799, rather than a working one
In this case, this doesn't matter, because as mentioned, it's a shader-related issue. Nothing has changed on the D3D11 side of things.
2f20a61342472cdb218f47ad8b94c129109d0337 should fix this. Essentially this is a game bug, it binds a multisampled image to a slot that expects a non-multisampled image. It's pure luck that this ever worked in the first place.
Confirming fixed by 2f20a61.
Nothing extracted yet.
As of dcb5b2a game world is black.
Software information
Dragon Age: Inquisition
Settings: Ultra/High mixture
System information
Apitrace file(s)
Trace is from Windows 7 as the game will not run under wined3d.
It's also ridiculously large at 1.7GiB compressed, 3.2GiB uncompressed.
I'm unable to replay the trace so I don't know how useful it will be.
Log files
DragonAgeInquisition_d3d11.log
DragonAgeInquisition_dxgi.log