| Age | Commit message (Collapse) | Author |
|
|
|
This VU exists now in recent specs.
|
|
Fixes #1910
|
|
Likely benign on anything reasonable, but still UB.
|
|
Codegen creates its own prototypes, no longer need them in the header
file.
Change-Id: Ibaab124a26391dc4925e49f095ddac83523aa725
|
|
- updated desktop/android build
- removed layers/object_tracker.cpp
- updated doc validator script for new location
- updated error database for 124 new checks
- added gen'd VU to GetSwapchainImages
Change-Id: If1b218dfcf62b203a8df96a0f12cbeb8bdcaf826
|
|
Change-Id: Ia97d2141a54263a9f034958f416661146dd953e2
|
|
Change-Id: I9419cdfdc39d33f5cc9f769c7498fde0230b144b
|
|
Change-Id: I704df0840801734ca35d5137a9d79d5a202b4fb9
|
|
These two checks were opportunistic and won't automatically be
covered by codegen. Removing now for clarity.
Change-Id: I3ada65c80094e8fa732a4ffc0da8c9ddeb0e11e0
|
|
Change-Id: I57ae325f27b365cd3b4e0f24d033af76ceeaeec7
|
|
One of the DebugMarkerSetObjectName parameters changes to const.
Change-Id: I23455957b83a84983bc7217a2a0d937ae03055e2
|
|
Special case function must avoid code generation.
Change-Id: I7c03cb2c0b4053f3d2b0579d532f1da82be80bd4
|
|
Special case function must avoid code generation.
Change-Id: I786d4ef4a8855f10e9cfe1234a5d885056f50a93
|
|
Moved to manually-generated, it has an odd corner case that's a one-
off and not worth code-generating.
Change-Id: I2631dd2107fffcd7e641b5a98bee126502d61ba8
|
|
Change-Id: Id77d5560479cbbc6f5453a6362c5a20dd721fdd0
|
|
Change-Id: I6acebbcbb9a10d86f3b7ec56148d9a680d085921
|
|
|
|
Fixes #1931.
|
|
|
|
Added validation checks, tests, and updated error database.
Change-Id: Id4dfe66bc325ae0060b7668572026c8a40f83d64
|
|
Change-Id: I7b9c729a90571d2888d7a22a9899546c547e6153
|
|
This change arises from an issue reported by Alon at Samsung. He
reports that a driver bug is modifying the descriptor set state on
UpdateDescriptorSets() call, so the data is getting corrupted in
validation. Since UpdateDescriptorSets() is a void function, it
doesn't make a difference in validation if we update state before or
after the call, since the call itself won't change the state being
updated. It actually saves us a lock call to update pre-call.
|
|
Test case where VkBufferMemoryBarrier is incorrectly issued within a
renderPass.
Kill old code that was doing the same check/test w/o the VUID.
|
|
This is VUID VALIDATION_ERROR_1b800932. Case where barrier
dependencyFlags mis-matches subpass dependency dependencyFlags.
Also update VU DB.
|
|
Complete remaining work for VALIDATION_ERROR_1b80092e and
VALIDATION_ERROR_1b800930. Make sure the image memory barriers src/dst
access masks are a subset of corresponding subpass self-dependency
access masks.
|
|
This is VU VALIDATION_ERROR_1b800934.
For a barrier in renderPass, bufferMemoryBarrierCount must be zero.
|
|
This is VUID VALIDATION_ERROR_1b800932. Verify that an barrier in a
renderPass has a dependencyFlags value that matches the corresponding
subpass self-depenency dependencyFlags value.
|
|
Change-Id: Ibc5028f660c91f66d986af19f6c7959eb9b55359
|
|
The device layers should list the names of the original layers
enabled during vkCreateInstance. We were expanding the
meta-layers which caused there to be a mismatch in names.
NOTE: that this list still include Implicit and Environment
variable enabled layers since those are techincally enabled at
vkCreateInstance as well.
Change-Id: I3ef88623e96a12a8f18f8f04e95a6f5444f4be56
|
|
Change-Id: Iacc0cb2676cd2346879d7d51e91b9b75232d5ad4
|
|
Moved to-be-generated source to object_tracker.cpp, manually-
written and helper functions to object_tracker_utils.cpp (new
file), and reworked header. Updated doc validator to search
new file.
Change-Id: I0fd3bcc4f6d123db07c7f8e2633e6fe16833e231
|
|
Update barrier validation to handle VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT
and VK_PIPELINE_STAGE_ALL_COMMANDS_BIT cases in the subpass dependency.
Expand GFX bit as it's an alias for a fixed set of bits. If
ALL_COMMANDS is specified then skip validation because we know that
barrier mask must be a subset of subpass mask.
|
|
Move renderPass barrier w/o self-dependency test from InvalidBarriers
to RenderPassBarrierConflicts test case. This prevents having to
re-created the default renderPass in InvalidBarriers test.
Also update framework subpass with self-dependency to set a group of
gfx pipe/access masks instead of EVERY mask, which is not allowed.
|
|
Add accessMask tests to RenderPassBarrierConflicts. Hit the two cases
where src & dstAccessMasks from pMemoryBarriers are mis-matched from
VkSubpassDependency.
|
|
Verify that src/dstAccessMask from pMemoryBarriers is a subset of the
corresponding mask from the relevant VkSubpassDependency instance.
|
|
Fix this copy-and-pasted error message by changing src->dst.
|
|
Migrate validation for renderpass barrier in CmdPipelineBarrier to its
own function. Refactor some device and command buffer params to avoid
duplicate look-ups. Add some const decls to command buffer state that
shouldn't change down the pre-call path.
|
|
- updated vulkan.h
- updated generator.py
- updated reg.py
- updated vk.xml
- updated Win/Lin json files
- fixed and updated vulkan.hpp
- updated validation error messages and database
- applied validextensionstruct fixes from 1.0.54 LVL vk.xml file
- fixed parameter_validation DebugMarker prototypes
- fixed unique_objects DebugMarker prototypes
- fixed NV_FILL_RECTANGLE in error DB
- updated error database with new covered VUs
Change-Id: I4eec816a6dc9756105a357beab6af4abb303994d
|
|
This patch adds a check to vkEnumerateInstanceExtensionProperties to
omit extensions provided by disabled implicit layers. Per section 31.2
(1.0.51), "When pLayerName parameter is NULL, only extensions provided
by the Vulkan implementation or by implicitly enabled layers are
returned." Arguably, implicit layers that have been disabled are not
"implicitly enabled".
|
|
Verify that the src & dst stage masks for a subpass barrier are a
subset of the subpass dependency's src & dst masks respectively.
Add RenderPassBarrierConflicts test to hit these cases. This test hits
the two cases of an image barrier src/dst masks not being a subset of
their corresponding subpass dependency masks.
Update database file to reflect that these checks are implemented and
that this test covers them.
|
|
VALIDATION_ERROR_1b800928 is implemented and tested in InvalidBarriers.
|
|
This is VALIDATION_ERROR_1b800928. Add the ID and slightly touch up the
error language. Also, this check should be general for any barrier
that's inserted for a renderPass so remove extra memBarrierCount
condition.
|
|
When an image descriptor with a destroyed image view was used we hit an
assert. This is not caught upstream and can occur if a descriptor is
re-used following image view deletion. Add code to catch this case
and report the destroyed image view.
|
|
Add check for Maintenance1 and negative_viewport_height extensions being
enabled simultaneously, along with test and a database update.
Change-Id: Id44f2a322fd32883d8b6d7ff1dab56996dec76c7
|
|
Change-Id: I82cd3f004410dea1d173ef8022091d6f4463d0f5
|
|
Change-Id: I73dac6489e41a94fc53a1c7942888a3a6479a805
|
|
Change-Id: I314ad9bc3e297da93f4d5c1d44289c9cca3a8fde
|
|
Moved 'Demo' declaration to near the top of the file to avoid referencing
problems.
Change-Id: I084231ada5683971baa073f1b761ac214fa02cfd
|
|
This fixes an MSVC compiler warning that's treated as an error
(and breaks the external sources build on Windows).
|