Just about every d3d9 program I could find (except the most stupid of samples) crashes if I disable my 950M from device manager.
Normal d3d9, nor vkcube or basemarkgpu seems to be affected by this behaviour.
Yes I have also verified the dxvk dll isn't getting hijacked like in #1904
Just about everything, but the following log is from unigine sanctuary.
Unigine.exe caused an Access Violation at location 14E938BA in module igvk32.dll Reading from location 00000004.
Registers:
eax=00000000 ebx=232cfdd8 ecx=00000000 edx=00000000 esi=232cfdd8 edi=00000000
eip=14e938ba esp=1d69fa80 ebp=1d69fabc iopl=0 nv up ei pl nz na po nc
cs=0023 ss=002b ds=002b es=002b fs=0053 gs=002b efl=00010206
AddrPC Params
14E938BA 00000001 1D69FB30 1D69FC30 igvk32.dll!SetDriverStoreDir
14E935DD 232CFDD8 00000000 00000001 igvk32.dll!SetDriverStoreDir
624B00E3 00000001 00000000 158DE6E8 d3d9.dll!Direct3D9EnableMaximizedWindowedModeShim
624BED13 00000006 00000001 00000000 d3d9.dll!Direct3D9EnableMaximizedWindowedModeShim
62448F34 00000006 00000003 00000000 d3d9.dll!Direct3D9EnableMaximizedWindowedModeShim
62447959 62555370 62555388 0ABB7BAC d3d9.dll!Direct3D9EnableMaximizedWindowedModeShim
62555388 0ABB7BA8 BDF61EA0 00000000 d3d9.dll!Direct3D9EnableMaximizedWindowedModeShim
777875F4 FFFFFFFF 777A735F 00000000 ntdll.dll!__RtlUserThreadStart
777875C4 62555370 0ABB7BA8 00000000 ntdll.dll!__RtlUserThreadStart@8
Intels Vulkan driver on Windows is notoriously bad. DXVK works fine on Intel GPUs on Linux.
I mean you could always try running these demos with Vulkan validation layers enabled to see if we're hitting some weird edge case on Intel's Windows driver (also, for the love of god, please post log files as requested by the issue template, why does nobody do that anymore?), but we have no plans to support it, especially since it's known to cause nothing but issues for just about any popular Vulkan project out there.
Because I think this problem is unrelated and preceding any actual operation with the vulkan driver itself (I understand every windows driver but perhaps nvidia's is bugged AF, but crashing immediately sounds something else)
Indeed, I have been fiddling with vkconfig for the past 10 minutes, and even if I enable every validation area the log isn't any longer than the normal one by dxvk.
Unigine_d3d9_intel.log
Unigine_d3d9_nvidia.log
Though I guess I was also misdiagnosing the problem too, now that I tried to force the intel gpu from the nvidia control panel.
To be clear, validation layer output would not be inside the DXVK logs, but rather in stdout (i.e. console output), or a separate file if you instruct the layer to dump it somewhere (should also be doable with vkconfig).
But at least the log tells us that there's nothing obviously failing.
Yes, the logs I have attached are dxvk's.
I'm just saying the ones from the vulkan configurator don't contain any additional information besides that.
Nothing for us to fix here then.
I very much suspect you are missing some d3d9 entry point..
https://github.com/apitrace/apitrace/issues/704
EDIT: I would have tried on W7 just to check, but as you may be aware the vulkan driver there hasn't been updated since early 2017
If some part of the Intel driver stack or any other Windows component calls undocumented D3D9 apis, that's not our problem.
@mirh
Windows -> closed source -> it's your choice. Happy using :)