On 10/25/19 5:22 PM, Rémi Bernon wrote:
On 10/25/19 5:14 PM, Hans-Kristian Arntzen wrote:
It does indeed sound like something is off with the RS 1.1 support, however, it's not clear to me what is going wrong or what the app is doing differently if it detects 1.1. As far as I know, the only difference with RS 1.1 is the addition of descriptor/data volatility information.
I didn't hit any issues with RS 1.1 on SOTTR when I tried it, fwiw.
Cheers, Hans-Kristian
I clearly have a crash on SOTTR with NVIDIA GPU. The game requests root signature 1.0 and we reply with 1.1, it then skips a bunch of CreateRootSignature calls that I can see when we return 1.0 only.
Then I have a couple of messages like:
warn:d3d12_pipeline_state_init_graphics:Root signature is NULL.
And quickly followed by a crash because of the assert in:
libs/vkd3d/command.c:3973
which tries to access root_signature->parameters with a NULL pointer.
MSDN says that this feature request should not reply with a higher value than requested. I didn't feel like it required a test, but I can add one.
Right, CheckFeatureSupport does specify it returns minimum version based on input. I missed that.
https://docs.microsoft.com/en-us/windows/win32/api/d3d12/ns-d3d12-d3d12_feat...
Cheers, Hans-Kristian