| Age | Commit message (Collapse) | Author |
|
Change-Id: I88defbb803818af3e7242d3c587a8b8decb42ffd
|
|
Change-Id: I539b8248ef8fb151de702c7d5e1773d149ce5313
|
|
vk-generate.py generated the vk_dispatch_table_helper.h file
based on the vulkan.py file. Added dispatch_table_generator.py,
which is an vk.xml-based generator to automatically pick up
API changes.
Change-Id: I93adae00566823ce8fc653a68f4b1b4079a998fc
|
|
Change-Id: I60ee7f881cfd2bd78de21fa14146ef5667e20cc6
|
|
Update Valid Usage enums in CmdExecuteCommands() and their
status in the VU database.
This partially implements Jira task VL-65
Change-Id: Ia31dbf78ed93418ced6597401d77f3fa6488935e
|
|
Add VALIDATION_ERROR_UNDEFINED to UNIQUE_VALIDATION_ERROR_CODE enum
with a value of "-1". This had been custom-defined in
object_tracker.cpp so removing it from there.
|
|
|
|
Various valid usage updates and additions.
Cmd to do update was:
python spec.py -update -remap 168-2059:1378-1203:1383-1206:1385-1207:\
1380-1204:1447-1268:1449-1269:1505-1310:1507-1311:1510-1313:1512-1314:\
1542-1753:1544-1755:1547-1758:1549-1760
|
|
Change-Id: Iafaf5c45a53ae702ecad8a387b11710ad60358f3
|
|
Change-Id: I9dcc1df2d979f1c36a88bb2d6181636c543431a1
|
|
Change-Id: I8b7ff75f334e8c2f2e07c04b6183abcd96e662df
|
|
This option was not used in any repo.
Change-Id: I133a1aaf34bd3e1a842150fc06f795eef8d1e606
|
|
Some updates based on review feedback.
Fix VerifyQueueStateToSeq to still validate across semaphore pairs in
the same fashion that RetireWorkOnQueue did.
Use a deque iterator rather than local copy.
Update some formatting.
|
|
Added validation flags to allow for disabling of validation for:
vkWaitForFences
vkGetFenceStatus
vkDeviceWaitIdle
vkQueueWaitIdle
|
|
Update QueueWaitIdle in core_validation to use Pre/Post pattern.
|
|
Update DeviceWaitIdle in core_validation to use Pre/Post pattern.
|
|
RetireWorkOnQueue() had one check that could cause a validation error.
The check makes sure that an unsignalled event does not precede a query
pool reset in a cmd buffer being retired.
This change splits out the check so that it can be hoisted prior to
calls down the chain, and the actual retirement of work on a queue then
only needs to take place after a call down the chain succeeds.
This initial commit separates these operations for fence retirement
and a follow-on commit will separate remaining operations for queue
retirement.
|
|
Refactor GetFenceStatus to use Pre/Post pattern.
|
|
Refactor WaitForFences to use Pre/Post pattern. This function currently
has a Post effect that can cause a validation error. Planning to remove
this in a follow-on commit.
|
|
This was an inadvertent change from 3f3748ad
|
|
|
|
|
|
|
|
|
|
This initial revision only supports builds for Linux.
Pull Request #1082
Change-Id: I0200eb7ae33ff23af2014624ee399bfe1929aaa8
|
|
Change-Id: I152ce109cada7bee85678217f39819ffe51580b9
|
|
Modify the 2 debug marker terminators we use so that they don't call
into the ICD unless the ICD commands exist.
Change-Id: I72a96475f21d0fca572f3c5e6f5676782a8a4bf3
|
|
STL min/max functions were causing windows build failures.
Change-Id: Ia153a5d6a3d245b645b5f829c3f076c023d865f1
|
|
Fixes #1165
According to spec descriptor updates should roll over to the next
binding number. If bindings were out of order, this was broken in
validation.
This fix corrects validation and state update for out-of-order
descriptor bindings by checking consecutive updates based on correct
binding count for consecutive bindings and by performing updates
on a per-binding basis, making sure to roll update over to the
correct next binding.
Also update the error message, related negative test, and database
file.
|
|
Fixes #1074
Object_tracker and parameter_checker intercept DebugMarker extension
functions which may not be supported by underlying device and
therefore could terminate with a NULL function call.
Updating these layers to handle this case by making sure that their
downstream function ptr is not null before calling down the chain.
If the downstream function ptr is null, considering that VK_SUCCESS
in the layers for functions with VkResult return code.
|
|
Killing the printCB* functions that would dump details of command
buffers for a number of functions.
|
|
Killing the print_mem_list function that would dump details of memory
allocations for a number of functions.
This function is a vestige of the original validation layers when
more debug info was useful. Now there are alternate tools and layers
that can produce this type of info so it's extraneous in the validation
layers.
|
|
This could also switch to using gnproc from brew if needed in the future.
|
|
readlink -f is not supported on macOS
brew is already a requirement to build for macOS
|
|
|
|
Update glslang and spirv-tools repo commit IDs to the latest version.
spirv-headers was still at the latest, and therefore, didn't need to
be updated.
Change-Id: I9165868cc43d077e2936ed53fdd7dd0171844453
|
|
Change-Id: I38ddea801f3efb887eac9da469428d3476f7adc3
|
|
Change-Id: I4e15f423964d101b1c0518bda9305f84b9ed35f9
|
|
Change-Id: Iebd6eeadeb31c0947314c388043a21d11a900b26
|
|
Change-Id: Iaa32b64423a7df3e8533d3fcd3a38f07e4a94051
|
|
Change-Id: I2ea237e46b028c0d2afa3a0f559eef42d8639a25
|
|
For the tests that trigger object_tracker messages with
error enums/messages:
- Update DB with names of these tests
- Update tests to use the error enum message strings
Change-Id: I75a122fe4864c798fa8401340c64dc5e9ed6dcb7
|
|
Change the testname column from Unknown to None for all
error enums referenced in object tracker. This is a good
starting point because there are vitually no tests explicitly
written to test object tracker checks.
A later commit will update the few rows in the database
with the names of the tests that currently exist.
Change-Id: Ifffb5ee09a74e8e2b7b8b2e70e7e9f9b4d12afbc
|
|
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>
|
|
Compute pipeline creation can partially fail in the same way as graphics
pipeline creation. Handle it exactly the same way in CV.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Previously we'd end up with a VK_NULL_HANDLE -> junk PIPELINE_STATE
mapping, which confuses other things.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|