The Vulkan spec says:
The queueFamilyIndex member of each element of pQueueCreateInfos must be unique
within pQueueCreateInfos, except that two members can share the same
queueFamilyIndex if one is a protected-capable queue and one is not a
protected-capable queue.
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
It was needed for use cases like DXVK winelib builds. It was never
really adopted and it's problematic for PE builds.
Signed-off-by: Jacek Caban <jacek@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Previously
"thunk" : True, "private_thunk" : True
and
"thunk" : False, "private_thunk" : True
had the same effect and make_vulkan used both variant.
Replacing the two bools with an enum should reduce confusion.
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
When using the windows vulkan loader vk_icdGetPhysicalDeviceProcAddr
is used for unknown physical device functions.
Fixes Red Dead Redemption 2 with vkd3d-proton. The game replaces wine's
vulkan-1.dll with a windows version that doesn't know functions used by
vkd3d-proton.
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Currently, this function is passed through winevulkan to the system's
Vulkan loader, which causes the loader to write the system's device
layers properties in pProperties.
Unreal Engine 4 then calls vkEnumerateDeviceExtensionProperties, using
VkLayerProperties::layerName as the pLayerName.
winevulkan's implementation of vkEnumerateDeviceExtensionProperties then
returns VK_ERROR_LAYER_NOT_PRESENT, crashing UE4.
This issue was found while debugging Project Wingman (Steam appid
895870), but it should also happen to any UE4 application using the
Vulkan RHI.
Signed-off-by: Victor Hermann Chiletto <v@hnn.net.br>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
The section for vkEnumerateInstanceLayerProperties states [1]:
On success, this command returns
* VK_SUCCESS
* VK_INCOMPLETE
On failure, this command returns
* VK_ERROR_OUT_OF_HOST_MEMORY
* VK_ERROR_OUT_OF_DEVICE_MEMORY
Always setting the layer count to zero and returning VK_SUCCESS is
valid, as even if the function is given an array in pProperties, it's
filled with zero layers
[1]: https://www.khronos.org/registry/vulkan/specs/1.2-extensions/man/html/vkEnumerateInstanceLayerProperties.html
Signed-off-by: Victor Hermann Chiletto <v@hnn.net.br>
Signed-off-by: Joshua Ashton <joshua@froggi.es>
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Fixes one of the issue related to the changes for VK_KHR_synchronization2.
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Previously bitmask types were always treated as 32-bit values, now the basetype of each bitmask must
be checked.
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Piers Daniell <pdaniell@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Needed to skip VK_NV_acquire_winrt_display in the next version bump.
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
In the past this was done in DllMain, the removal broke winelib applications.
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
We shouldn't do that according to the Vulkan xml maintainer.
https://github.com/KhronosGroup/Vulkan-Docs/pull/1379
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Messages from the driver/layers won't be reported twice.
Messages during instance creation by the host loader won't be reported,
since the information is partly wrong and the windows loader reports the
correct information.
Other messages by the loader will be reported twice if lunarg's vulkan-1.dll
is used. This shouldn't be a problem because the messages are just used
for logging/debugging.
Needed for Quake2RTX.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=49813
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This eases development and of Vulkan extensions by allowing developers
to locally override the specification XML file used by winevulkan.
Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>
Signed-off-by: Piers Daniell <pdaniell@nvidia.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Jacek Caban