The new init mechanism always uses GetInstanceProcAddr
to query the next layer's CreateInstance or CreateDevice
function. Thus no reason to cache those function pointers
in the instance dispatch table.
New layer init method requires the construction of
Link information for CreateInstance and CreateDevice
that is accessible to layers via the CreateInfo.pNext pointer.
The layer can then use the Get*ProcAddr from the Link
structure to initialize their dispatch table if the
call down the chain returns successfully.
This removes the need to do special initialization work
at Get*ProcAddr time.
Layer Get*ProcAddr now return their internal function
pointers regardless of the value of instance or device.
Only need to have valid instance & device when looking
up extensions or when passing the request down the chain.
This mechanism allows us to remove object wrapping used
by the loader previously. Also simplifies the dispatch table
setup.
Conflicts:
layers/device_limits.cpp
layers/draw_state.cpp
loader/loader.c
loader/trampoline.c
These structures will be used to extend CreateInstance
and CreateDevice to support layers. The loader will
create a chain of these structures - one per layer -
that tells the layer the Get*ProcAddr for the next
lower part.
It took a few commits (squashed down into 1) to get it correct. Includes:
Use loader_heap_{alloc|free}().
Try to deal correctly with initializing the vkCreate*SurfaceKHR()'s.
GetInstancePorcAddr() is specified to return trampoline entrypoints for all
Vulkan core and extension entrypoints that are dispatched on an instance object
or a child of that instance object. However, typically, device extensions would
be unknown to the loader (don't want to rev the loader everytime an IHV creates
a new device extension).
This patch allows loader to dynamically discover device extension entrypoints
and configure generic trampoline code for these discovered device extensions.
Add a check to determine if the API version being requested can be
handled by the loader. If it can't, then we return incompatible driver.
If we can, we continue along our merry way.
An addendum is that we also bumped the Loader/ICD interface version.
This was requested by Nvidia so that the ICDs know that we pay attention
to the version in the loader. If they don't see the new Loader/ICD
interface version, they will handle the failing on all non-1.0 API
requests.
Change-Id: Icb7dd45e754c9f6a6c8186198333bacc68077b93
Add new extensions for 1.0.39. Also, updated layers to include
minimal set of functionality for 1.0.39 extensions. Extensions include:
- VK_KHR_get_physical_device_properties2
- VK_KHR_shader_draw_parameters
- VK_EXT_direct_mode_display
- VK_EXT_display_surface_counter
- VK_EXT_display_control
Also, redo the LoaderAndLayerIf document.
Change-Id: I10412086da7a798afe832a3892e18f606259b5af
Need to keep VkIcdSurface private to the loader.
The definition may conflict with some redefinitions of
VK_DEFINE_NON_DISPATCHABLE_HANDLE and it's not necessary to make public
anyway.
Change-Id: I30e166f0ecce2fd3ea36c745dc073b705973d75f
Use the ICD to create and destroy VkSurfaceKHR objects instead
of just performing the work in the ICD. This only occurs if the ICD
exports the appropriate entry-points, and exposes version 3 of the
loader/icd interface.
Change-Id: I5e7bf9506318823c57ad75cf19d3f53fdfa6451e
Per Khronos Bugzilla Bug 15077, on Windows and Linux, VkSurfaceKHR is treated
as a pointer to platform-specific structs that contain the platform-specific
connection and surface/window info. The Vulkan loader vkCreate*SurfaceKHR()
functions will fill in the struct. ICDs and layers will cast VkSurfaceKHR to a
pointer to the appropriate VkIcdSurface* struct.
