When checking for unknown physical device functions, check the first layer that supports
vk_layerGetPhysicalDeviceProcAddr in the chain starting with the layer closest to the
application. This prevents unecessary work being done, and if any layer wraps VkInstance
will properly call through the wrapping layer first without calling into any other layer.
Various situations could cause an OOM to turn into a hard crash due to double freeing
of memory that was improperly cleaned up. Also fixed memory leaks when layers would
report OOM.
Make it possible for layers to declare that they support unknown functions
and set up the layer dispatch table appropriately. A very niche use case
but helpful when developing validation for functions not yet out in the
public headers.
My previous commit to enable physical device modifications in a layer
required new compiled layers to do the work. Since this can more
cleanly be done by setting layer flags (as suggested by @charles-lunarg)
I have gone back and done that.
The loader trampoline previously would query all devices every time
vkEnumeratePhysicalDevices was called.
To do this, it would make two calls every time:
- First, it would ignore the passed in user values
- Second, it would query the total number of available devices.
- Third, it would query the values for every available device
This resulted in layers reporting 2 vkEnumeratePhysicalDevices call for
every 1 the application made which could get very polluted in output.
It didn't break any functionality, just made things messy.
This change removes that behavior and adds a bunch of test cases to verify
nothing broke in the move.
Add tests for unknown device functions queried with both vkGetDeviceProcAddr
and vkGetInstanceProcAddr. Also loops enough times to exhaust the available
unknown function supply.
Had to fix the vk_layerGetPhysicalDeviceProcAddr in TestLayer to make the tests
pass.
MacOS currently skips tests which use vkGetInstanceProcAddr. This is due to the
need for unknown functions being lower priority and so do not warrent the time
needed to write the assembly necessary to resolve the issue fully.
Add tests to verify that the 3 pre-instance functions are properly
intercepted by the layer when everything is well-defined.
Also include two fixes:
1) Disable environment variable value of 0 should be treated the same as not present
2) The vkEnumerateInstanceVersion override path was broken, missing a NULL check
The VK_MAKE_API_VERSION macro and the VK_API_VERSION_1_X macros are better to
use. This commit switches the tests over to the non-deprecated versions of the
macros.
By adding builder style usage to most of the layer interface, it is easier to
create and use layers in tests. This commit refactors the layer interface and
updates most of the tests to use this style.
Made InstanceWrapper and DeviceWrapper simpler to use by moving them to test_environment
then making the Create() member functions and doing the gtest assertions in the function
itself.
Additionally:
Removed the `detail` namespace in the test_environment.
Renamed get_new_test_icd/layer to just `reset_icd/layer`. This is clearer about what it
is doing and makes the interface more expressive with code such as
`env->reset_icd().SetMinInterfaceVersion(5);`
* `get_instance_func` needed to have __stdcall (VKAPI_ATTR) added
* Several .def files were incorrect
* vk_layer.h forward declares vkNegotiateLoaderLayerInterfaceVersion. This is
problomatic because it doesn't have __declspec(dllexport), a token added by
FRAMEWORK_EXPORT on windows. We have to special case the export macro here to
prevent that mismatched definition.
Functions needed in the various layer interface versions are now
exported as well as guarded behind macros configurable from CMake.
The export_definitions folder was populated as well.
Change-Id: Iebfdd8ae02062eb81b8aa9f705885c9384c95be9
andrew0229