Open billhollings opened 2 years ago
Isn't supporting Vulkan Memory model optional? The model wasn't formalized at Vk-1.0 time. So a Vulkan-1.0 portability implementation shouldn't need to pass the memory model tests at all.
Isn't supporting Vulkan Memory model optional?
Not all the tests under dEQP-VK.memory_model
concern VK_KHR_vulkan_memory_model
. (I actually pointed this concern out when Bill started working on fixing these tests, and that was his reply.) Believe it or not, the permuted_index
tests are among those that don't require it.
The problem AIUI is that Metal atomics currently only support relaxed
ordering, when SPIR-V specifies stronger orderings acquire
, release
, acq_rel
, and seq_cst
.
Interesting. Should this flag be called shaderAtomicOrdering
then (trying to come up with something closer to the semantics)?
Some CTS tests fail on shader releasing and acquiring shared variables.
This is most likely fixable with effort within the implementations.
In the meantime, these use cases can be disabled in CTS and reported as Not Supported, by adding a
VkPhysicalDevicePortabilitySubsetFeaturesKHR:: shaderSharedVariableReleaseAcquire
capability.This affects 2 CTS tests: