| Age | Commit message (Collapse) | Author |
|
This reverts commit 94dd6081959103c1c9b0bcfda173d4149d5555b9.
Change-Id: Ie5b437d822c5efe6baf0baf2830591f50bf3ee4d
|
|
Fixes #667.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Validation was assuming LOAD_OP_CLEAR for some attachments resulting
in bad validation messages.
Change-Id: I8ff75ca63f591e3002b1ee2b61a6844d0cf1c07d
|
|
In core_validation attachment_references_compatible(), having array
elements set to VK_ATTACHMENT_UNUSED was not handled resulting in
invalid memory references.
Change-Id: If78aea1467d191db905272bece16cc9f56643136
|
|
Nothing used this after rationalizing the state tracking.
Change-Id: I67cc267336b5d64bea9645a8e0a072885db58c26
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: I8f719c6e57cbab836b90f30414e6494a0643580e
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: I3e5a5d08763c0bf26c6c69aecf4bcef17d055a48
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: Iceda3b28474e289189ef1f97b271f9e9bb7b2083
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Instead of having a combination of firstTimeFlag, in_use, and
needsSignaled, let's just have an enum with the three states a fence can
actually be in.
Change-Id: I6a40cd709717c75740c9977b149c4f71a02a27ab
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: Ib8f9a917115323d8d43b971593b213ff04138a35
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: I8351405299f2f109489fdddfc111412164aded77
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: Id3e0f09254197119b821710f0e9ce27d3458afc9
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: I46f4ef4c225e199c0415507bd19dc4ffc98a77bc
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: I2b62475951315c6b231342be6f7c74865e3ea2bb
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
|
|
If a DS image is used in a descriptor, regardless of the underlying
image layout, we need to validate that BOTH DEPTH and STENCIL aspect
bits are NOT set. Only one of the two bits is allowed.
|
|
|
|
Correctly handle null resolve attachment case. When verifying attachment
compatibility, matching null resolves is ok.
|
|
At draw time, the active render pass must be compatible with the render
pass that the gfx pipeline was created with.
|
|
|
|
In the core_validation layer, much of the validation was occurring
AFTER the API call had been made. Separated validation and tracking.
Change-Id: I6f8502d52dd2861a18254de48141a3736ced257f
|
|
When creating descriptor set layout, we were only incrementing the dynamic descriptor
count by 1 regardless of the number of descriptors in a binding. This fix increments
the count by the number of descriptors in the binding.
|
|
Change-Id: Ifdce033092a72980e1ff685792d486d2a101f35d
|
|
Stencil attachment load/storeOp settings were being ignored. Added
format appropriate checks for these flags.
Change-Id: I6a917fc8a28cbb0d0441152e8d6630defb1b81e3
|
|
Added entries for new attachment index tests and enums.
Change-Id: Ida1de0bffc544718e63ccf3641a09dbe577f3c4d
|
|
Framebuffer attachment images are created with IMAGE_USAGE flags. These
flags are now validated against their attachment location in subpass
description structures. Also added a check for a preserve attachment
being set to UNUSED and fixed a few typos in the attachment index
checking stuff.
Change-Id: Id62d3539c490ce8f381d7ae606402980f11946fa
|
|
Added new error enum for memory index type mismatches.
Change-Id: I6f857facc4209f103da5cef9f6561a0e4ec3c334
|
|
Bind[Image|Buffer]Memory did not validate that the memoryPropertyType
matched the memory_requirements memory type flags for the buffer
or image.
Change-Id: I7007588d78408689c34df66a17cc9201a37569b1
|
|
A) Adding the WriteDescriptorSetIntegrityCheck test, which verifies
some requirements of chapter 13.2.3 of the Vulkan Spec
1) A uniform buffer update must have a valid buffer index.
2) When using an array of descriptors in a single WriteDescriptor,
the descriptor types and stageflags must all be the same.
3) Immutable Sampler state must match across descriptors
The above tests also exercise cases where descriptors are updated with
consecutive bindings both correctly and incorrectly
B) adding a MissingClearAttachment test that points to a wrong
colorAttachment index in a VkClearAttachment structure passed to
vkCmdClearAttachments
C) update vk_validation_layer_details.md to reflect above changes.
Change-Id: I3f4f2233371b68caf23b5a4111d63a94765121c3
|
|
Fix a recent change broke windows builds that don't use the VC IDE or
decorate build trees like VC does
Change-Id: I93c8a559c4255408a6fa890b2a3e4d81822d1cea
|
|
Change-Id: I129b4d074f0604bbdf6d81c052333229ece99ac3
|
|
Enable output of a short message header containing definitions of
debug message types.
Change-Id: Ia84fb0e79d599b8d022b7d78267b7b6a0307823e
|
|
Was previously statically linked. Converting to a DLL will allow
data sharing and use of utils in samples repo.
Change-Id: I63ccc1f0f57dfad40f7068c60ac623bce5bd714d
|
|
One of the event tracking lists was not getting cleared upon a reset,
resulting in an invalid reference on reuse.
Change-Id: I45d932822f27197266b363634d6cdccc785209d3
|
|
This test was accidentally triggering the fence reuse case as well, and
blindly continuing past that error. It doesn't need to use the fence at
all, so don't supply one.
Also tweak the layer to not run into the assert if the caller elected to
skip -- QueueSubmit still has validation and updates tangled together,
but this improves it slightly.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
When a command buffer is simultaneously inflight multiple times, each
can have different semaphores associated. Storing the set of semaphores
on the GLOBAL_CB_NODE caused us to get confused and emit bogus errors.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
We'd like to be able to get back to them from the tracking struct.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Unused.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: I7468c088d99f7bebabc05d1a84529f6e9e00d424
|
|
Added new cases, removed old enums, moved info from devlimits
to parameter validation.
Change-Id: Iddf6a679dc8596dcf411a6b880a63a26999607ec
|
|
Was ignoring VK_WHOLE_SIZE. Moved validation cases for Update/FillBuffer
to Parameter Validation layer (as they were stateless) and added the
remaining additional stateless valid usage checks.
Change-Id: I0f41c6ba55f218a6db9fb34736a2abc52f856edf
|
|
Swapchain layer had stale references -- rename things still using
imageArraySize instead of imageArrayLayers.
Change-Id: Iaa4500ab8cf55e50b96636a50136605dd3e6d83e
|
|
Added device_limits MISSING_QUERY_COUNT
Change-Id: I5b129ec12930129f5d5fe46bbe2622977dc2d992
|
|
Replace call sequence errors with warnings for
vkEnumeratePhysicalDevices and GetPhysicalDeviceQueueFamilyProperties.
Change-Id: I1505f4a2b339ac7ef435da8b3ec2f85612026a42
|
|
The validation layer checks for dependency between subpasses using the
same attachments, but it also issues warnings if there are "only"
implicit dependencies between two subpasses. This is explicitly
allowed by the spec and should not cause validation warnings.
Change-Id: I1039955d6d445e1ee234116c6da2e4dbaf9c7dd9
|