Michael Lentine [Tue, 24 May 2016 14:49:25 +0000 (09:49 -0500)]
Clear priorFences on reset.
Michael Lentine [Fri, 20 May 2016 22:45:02 +0000 (17:45 -0500)]
layers: Fix query tracking across multiple command buffers on the same queue
Michael Lentine [Fri, 20 May 2016 15:14:00 +0000 (10:14 -0500)]
layers: Fix event tracking to differentiate reading and writing
Chris Forbes [Sun, 22 May 2016 23:29:51 +0000 (11:29 +1200)]
layers: Move and rename validate_descriptor_availability_in_pool
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 20 May 2016 06:27:28 +0000 (18:27 +1200)]
layers: Add getPipelineLayout helper, use it
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 20 May 2016 05:38:44 +0000 (17:38 +1200)]
layers: Use pipeline layout -> ds layout direct map
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 20 May 2016 05:14:32 +0000 (17:14 +1200)]
layers: Use getDescriptorSetLayout helper in more places
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 20 May 2016 05:12:39 +0000 (17:12 +1200)]
layers: Remove more AllocateDescriptorSets noise
Can argue about whether this stuff at INFO level is worthwhile...
Replace the comments with a clearer set of comments.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 20 May 2016 05:04:07 +0000 (17:04 +1200)]
layers: Further rework to AllocateDescriptorSets
- operator new does not return null, it throws.
- do lookups once, at the start
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 20 May 2016 04:34:01 +0000 (16:34 +1200)]
layers: Get rid of pre-call updates in AllocateDescriptorSets
Moving all the updates later requires that we totalize the required
descriptors by type first, rather than testing each binding's
requirement as we go.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chia-I Wu [Sun, 15 May 2016 23:48:14 +0000 (07:48 +0800)]
parameter_validation: avoid invalid GIPA call in vkCreateDevice
Save VkInstance in layer_data so that we query vkCreateDevice with a valid
instance.
Chia-I Wu [Sun, 15 May 2016 23:45:43 +0000 (07:45 +0800)]
parameter_validation: handle device commands in GetInstanceProcAddr
Call intercept_core_device_command in GetInstanceProcAddr.
Chia-I Wu [Sun, 15 May 2016 23:43:38 +0000 (07:43 +0800)]
parameter_validation: no interface functions in layer functions
Move handling of interface functions in
parameter_validation::GetInstanceProcAddr to v0's vkGetInstanceProcAddr.
Chia-I Wu [Sun, 15 May 2016 23:41:17 +0000 (07:41 +0800)]
parameter_validation: refactor GetInstanceProcAddr
Split command intercepting out to intercept_core_instance_command. Assert
that instance is valid if intercept_core_instance_command returns nullptr.
Chia-I Wu [Sun, 15 May 2016 23:37:41 +0000 (07:37 +0800)]
parameter_validation: add intercept_core_device_command
It returns the function pointers for all intercepted core device commands.
Call intercept_core_device_command from GetDeviceProcAddr. Assert that
device is valid in GetDeviceProcAddr.
Chia-I Wu [Sun, 15 May 2016 23:34:09 +0000 (07:34 +0800)]
parameter_validation: improve EnumerateDeviceExtensionProperties
Call down the next layer unless pLayerName is parameter_validation.
Chia-I Wu [Sun, 15 May 2016 23:30:58 +0000 (07:30 +0800)]
parameter_validation: add parameter_validation::global_layer
Replace pc_global_layers with global_layer. There is no plan to define
more than one layer in the same namespace.
Chia-I Wu [Fri, 13 May 2016 06:37:49 +0000 (14:37 +0800)]
parameter_validation: drop vk prefix for layer functions
Drop VK_LAYER_EXPORT as well.
Chia-I Wu [Fri, 13 May 2016 06:07:36 +0000 (14:07 +0800)]
parameter_validation: put layer functions into a namespace
Put all layer fucntions/data into parameter_validation namespace. I had
to add some wrappers to make everything work.
Chia-I Wu [Fri, 13 May 2016 06:06:08 +0000 (14:06 +0800)]
parameter_validation: move vkEnumerate*Properties around
Move them toward the end of the file.
David Pinedo [Fri, 20 May 2016 21:05:44 +0000 (15:05 -0600)]
winrtinstaller: remove dependency on msvcrt redistributables
Loader and demos now use static version of msvcrt.
Runtime Installer no longer installs msvcrt.
Mark Lobodzinski [Thu, 19 May 2016 23:06:56 +0000 (17:06 -0600)]
layers: GH7, Replace bad var names in obj_tracker
C++ reserves variable names that start with an underscore followed
by a capital letter. This patch series replaces or removes these
invalid variable names from the LVL codespace.
Change-Id: I71368cf30e267841cedcc3861afa3444689d71da
Mark Lobodzinski [Thu, 19 May 2016 23:25:59 +0000 (17:25 -0600)]
layers: GH7, Fix bad var names in layer utils
Change-Id: Iec47953ba0611de5c62f4fa876eb02fefca32427
Mark Lobodzinski [Thu, 19 May 2016 23:22:25 +0000 (17:22 -0600)]
demos: GH7, Fix invalid variable names
Change-Id: Iabb217f70d0068284c682d9da6b4c2834902256e
Mark Lobodzinski [Thu, 19 May 2016 23:17:58 +0000 (17:17 -0600)]
headers: GH7, Fix bad variable names in vk_icd.h
Change-Id: Ic78792a71ebedbdb5bae68c47856db1c89a714af
Mark Lobodzinski [Thu, 19 May 2016 23:23:38 +0000 (17:23 -0600)]
layers: GH7, Fix bad var names in core_validation
Change-Id: I4fd1b8bb1cb83af633e04fc561771b6d76732588
Mark Lobodzinski [Thu, 19 May 2016 23:10:58 +0000 (17:10 -0600)]
layers: GH7, Fix bad var names in device_limits
Change-Id: I6dec3fd87f62ac171886562206f3fc528f24bc29
Mark Lobodzinski [Thu, 19 May 2016 23:10:01 +0000 (17:10 -0600)]
layers: GH7, Replace bad variables in image layer
Change-Id: I6c90dafe83bf4e9674b4e32ed867d77b5820006d
Mark Lobodzinski [Thu, 19 May 2016 23:08:10 +0000 (17:08 -0600)]
layers: GH7, Fix bad var names in threading layer
Change-Id: Iabfd7900a934162a425008593ecc0f0b4d56476e
Mark Lobodzinski [Thu, 19 May 2016 23:01:48 +0000 (17:01 -0600)]
layers: GH7, Fix bad var names in swapchain layer
Change-Id: I46331c530654334abacbfb3ffd3656809f6b7d5d
Tobin Ehlis [Thu, 19 May 2016 17:08:34 +0000 (11:08 -0600)]
tests: GH535 Add tests for clear image errors
Add test that clears a depth/stencil image with CmdClearColorImage()
and one that clears color image with CmdClearDepthStencilImage()
Tobin Ehlis [Thu, 19 May 2016 16:10:09 +0000 (10:10 -0600)]
layers: Replace is_depth() in image layer with util function
vk_layer_utils has function vk_format_is_depth_or_stencil() that is
exactly the same as local is_depth() function so kill local function
and use utility function.
Tobin Ehlis [Thu, 19 May 2016 15:59:30 +0000 (09:59 -0600)]
layers: GH535 Add checks for Clear image errors
This adds two new checks to make sure that CmdClearColorImage is
not called with depth/stencil format image and that
CmdClearDepthStencilImage is not called with color format image.
Also add getImageState() helper function to image layer to
simplify/unify code for lookup of IMAGE_STATE from imageMap.
Tobin Ehlis [Thu, 19 May 2016 19:08:11 +0000 (13:08 -0600)]
doc: Add "Pending" string back in for shader checker
"Pending" is used as a trigger keyword for document validator to know
when one section is ending and a new section is starting so adding it
back in to get doc validation passing.
Chia-I Wu [Thu, 19 May 2016 02:45:02 +0000 (10:45 +0800)]
docs: update v0 languages for device layer deprecation
Clarify that a layer's
vkEnumerateInstanceLayerProperties
vkEnumerateInstanceExtensionProperties
- should enumerate the layer itself
vkEnumerateDeviceLayerProperties
- is deprecated
vkEnumerateDeviceExtensionProperties
- must handle all layers by chaining
vkCreateInstance
- `pNext` handling is covered by the spec
vkCreateDevice
- allow validation layers to validate layer and extension names
Clarify that the layer library's
vkEnumerateInstanceLayerProperties
vkEnumerateInstanceExtensionProperties
- are not used by the desktop loader
- can be aliases to the layer's versions when the layer library
contains only one layer
vkEnumerateDeviceLayerProperties
vkEnumerateDeviceExtensionProperties
- are not used by the desktop loader
vkGetInstanceProcAddr
- mention that the special cases for vkCreateDevice and device commands
are for compatibility
Mark Lobodzinski [Thu, 19 May 2016 17:50:32 +0000 (11:50 -0600)]
layers: Updated validation layer details doc
Some rewording, pulled out pending work sections and pointed readers
to Github for open/pending issues or feature requests.
Change-Id: I1ad55777a064e7f26a3f5cce48fb63d9e12d4b32
Tobin Ehlis [Thu, 19 May 2016 14:00:00 +0000 (08:00 -0600)]
layers: DescriptorSet class clean-up
Minor formatting improvements, removed some useless "const" decls
and unused vars and updated a few error codes to be more appropriate.
Tobin Ehlis [Thu, 19 May 2016 13:56:18 +0000 (07:56 -0600)]
layers: Verify that src of copy update is updated
Before verifying source update comments make sure that the source
descriptors have been updated. This guards against some obscure
corner cases where a bad source descriptor could slip through the
content verification.
Tobin Ehlis [Wed, 18 May 2016 19:43:26 +0000 (13:43 -0600)]
layers: Refactor DescriptorSet update interface
Mainly refactor and moving code in order to provide an interface to
DescriptorSet class that matches top-level vkUpdateDescriptorSets()
function.
Split the validation of an update as a separate task from performing
the update. This allows validation prior to calling down the chain
and then only update the state if validation is clean.
Hoisted all of the update validation into the DescriptorSet class
which prevents having to copy all of the maps into the individual
Descriptor classes. This simplifies both their creation and updating
their contents.
Updated the top-level core_validation UpdateDescriptorSets() code
to match Vulkan Validation Layer Authoring Guidelines. As this is
an initial POC for the architecture, I kept the Pre* & Post* functions
in the core_validation.cpp file, but they should eventually be spun
out.
Tobin Ehlis [Tue, 17 May 2016 14:01:41 +0000 (08:01 -0600)]
layers: Migrate types from core_validation.h
This change just moves a few types from core_validation.h to
core_validation_types.h. This is preparation for migrating DescriptorSet
class to store GLOBAL_CB_NODE ptrs directly.
Chris Forbes [Tue, 17 May 2016 06:47:09 +0000 (18:47 +1200)]
layers: Simplify command buffer pool cleanup
We don't need to carefully pick through the pool's buffers, removing
them. We're about to throw away the whole container.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Tue, 17 May 2016 05:42:27 +0000 (17:42 +1200)]
layers: More slight tidying
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Tue, 17 May 2016 03:45:31 +0000 (15:45 +1200)]
use getFramebuffer more
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Tue, 17 May 2016 03:31:00 +0000 (15:31 +1200)]
layers: simplify TransitionSubpassLayouts
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Tue, 17 May 2016 03:27:58 +0000 (15:27 +1200)]
layers: Add getFramebuffer() helper
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Tue, 17 May 2016 02:59:22 +0000 (14:59 +1200)]
layers: Simplify CmdEndRenderPass
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Mon, 16 May 2016 23:36:23 +0000 (11:36 +1200)]
layers: Add getRenderPass helper
Various similar functions exist, but this one was missing. Returns the
RENDER_PASS_NODE ptr if it exists, or nullptr otherwise.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Mon, 16 May 2016 23:02:24 +0000 (11:02 +1200)]
layers: Unify cb lookup in CmdBeginRenderPass
We already have a GLOBAL_CB_NODE ptr here, and we know it's not null.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Tobin Ehlis [Thu, 19 May 2016 12:58:13 +0000 (06:58 -0600)]
layers: Remove lastBound state reset in destructor
No need to reset this state right before GLOBAL_CB_NODE is destroyed.
Tobin Ehlis [Tue, 17 May 2016 20:23:46 +0000 (14:23 -0600)]
layers: Improve DescriptorSet cleanup
Create private helper function InvalidateBoundCmdBuffers() within the
DescriptorSet class to unify invalidate cases due to set being updated
or freed.
Add a destructor for DescriptorSet to make sure that no bound cmd buffers
hang on to deleted set references.
Tobin Ehlis [Tue, 17 May 2016 16:41:55 +0000 (10:41 -0600)]
layers: Reduce DescriptorSet map look-ups
Migrate some data structures to refer to DescriptorSet class
instead of raw VkDescriptorSet. This saves map look-ups.
Tobin Ehlis [Tue, 17 May 2016 14:31:46 +0000 (08:31 -0600)]
layers: Refactor DescriptorSet Update code
This migrates code for DescriptorSet in_use and flagging bound
command buffers as invalid from core_validation.cpp into DescriptorSet
class.
Tobin Ehlis [Tue, 17 May 2016 16:38:45 +0000 (10:38 -0600)]
layers: Fix intercepts of parameter_validation
Make parameter_validation register vkFreeDescriptorSets() and
vkUpdateDescriptorSets() functions as intercepted.
Tobin Ehlis [Tue, 17 May 2016 14:01:41 +0000 (08:01 -0600)]
layers: Migrate types from core_validation.h
This change just moves a few types from core_validation.h to
core_validation_types.h. This is preparation for migrating DescriptorSet
class to store GLOBAL_CB_NODE ptrs directly.
Tobin Ehlis [Tue, 17 May 2016 13:58:01 +0000 (07:58 -0600)]
layers: Const correctness for debug_report_data param
debug_report_data pointer passed to various log functions can be const.
Jon Ashburn [Wed, 18 May 2016 20:07:47 +0000 (14:07 -0600)]
doc: update layer and loader doc for device lyaer deprecation
Change-Id: I53ff750a25fd5ea390c22f5ded2247dfbedce2aa
Pavol Klacansky [Tue, 10 May 2016 09:08:19 +0000 (03:08 -0600)]
Remove Xlib dependency when XCB is used
Change-Id: I3626f299aae8acc2640e3f233a46d84cc819fd63
Tobin Ehlis [Tue, 17 May 2016 20:50:07 +0000 (14:50 -0600)]
layers: Update cmds allowed in 2nd CB subpass
If subpassContents are of type VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS
then CmdNextSubpass and CmdEndRenderPass are accepatable cmds to be added to
the secondary cmd buffer.
Spec language is from section 7.4 Render Pass Commands: If contents is
VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS, the contents are recorded in
secondary command buffers that will be called from the primary command
buffer, and vkCmdExecuteCommands is the only valid command on the command
buffer until vkCmdNextSubpass or vkCmdEndRenderPass.
Tobin Ehlis [Mon, 16 May 2016 17:23:01 +0000 (11:23 -0600)]
layers: GH544 Allow INVALID cmd buffers to be reset
BeginCommandBuffer performs an implicit reset of a command buffer.
This allows command buffers in the INVALID state to undergo this
implicit reset.
In order to make sure that an invalid command buffer in the INVALID
state has had EndCommandBuffer called, this commit adds a CMD_END
command that's inserted at the end of a command. If a command buffer
in the INVALID state has had EndCommandBuffer called and it is now
having BeginCommandBuffer called, it's ok to implicitly reset the
command buffer.
buffer's vector of commands.
Mark Lobodzinski [Tue, 17 May 2016 20:52:43 +0000 (14:52 -0600)]
layers: Update layer README.md for release
Updated language surrounding device/instance layer changes, fixed
up some markdown formatting issues.
Change-Id: I66353eeeef6789ab15c4e08c21f9da149f9ce09e
Ian Elliott [Tue, 17 May 2016 18:03:35 +0000 (12:03 -0600)]
demos: GH#559 fix for new Android windows
On Android, treat a 1st-time new window differently from an nth-time
new window.
Tobin Ehlis [Tue, 17 May 2016 12:43:48 +0000 (06:43 -0600)]
layers: Add correct IsStorage() to TexelDescriptor
Similar to Norbert's previous fix for ImageDescriptor, TexelDescriptor
was also missing correct IsStorage() override function and so would always
return false instead of actual storage_ value.
Change-Id: I9006cdb614c5b78db0a68b71dcc4300f4c5c0d7f
Dustin Graves [Mon, 16 May 2016 17:07:59 +0000 (11:07 -0600)]
layers: Ignore 'noautovalidity' params w/codgen
Add special handling to the parameter_validation layer's code
generator for vk.xml parameters with the 'noautovalidty' attribute:
- Ignore vk.xml parameters with the 'noautovalidity' attribute
when generating code for the parameter_validation layer.
- Implement custom validation routines for parameters with
'noautovalidity' tags.
Issues-Addressed: GitHub 515
Change-Id: I1bd28cc79233fb8b040d6e107ed5e2176a84edb5
Dustin Graves [Tue, 17 May 2016 00:35:55 +0000 (18:35 -0600)]
layers: Fix transposed messages
Fix parameter_validation::get_result_rescription utility
function, which had the messages for VK_ERROR_DEVICE_LOST
and VK_ERROR_INITIALIZATION_FAILED transposed.
Change-Id: I47126389cbfafce663ec474fa04b899e6f2f663c
Jeremy Hayes [Tue, 17 May 2016 15:32:12 +0000 (09:32 -0600)]
loader: update linux so version
Change-Id: I84656754952e7488712ca5bcd4701e2e85d92cce
Jon Ashburn [Tue, 17 May 2016 14:43:38 +0000 (08:43 -0600)]
loader: Treat GLOBAL layers same as INSTANCE layers
Also change validation layers type to be GLOBAL so they work with old loaders
prior to device layer deprecation.
Change-Id: I32788cb7788c8ad840ced15236c5ed792edfbbff
Mark Lobodzinski [Tue, 17 May 2016 14:44:09 +0000 (08:44 -0600)]
layers: GH485, Remove redundant format props check
With new checks in place for CreateImage, old check in CreateRenderPass
is redundant and can be removed.
Change-Id: I53e5201f453041b3c8ba17edc193c1f3b142f4cb
Mark Lobodzinski [Tue, 17 May 2016 14:42:03 +0000 (08:42 -0600)]
layers: GH485, Strengthen image format validation
CreateImage was checking that format caps weren't null, but was not
validating against color or depth/stencil attachment types.
Change-Id: Ic8996829033f552f6ef7477bdefdaba702ae2403
Tony Barbour [Mon, 16 May 2016 21:47:56 +0000 (15:47 -0600)]
demos: Make sure shaderClipDistance is available before requesting
Change-Id: Id0b58d54a553fce4914ddc1097f1e1c8ac3c337f
Jon Ashburn [Mon, 16 May 2016 23:35:43 +0000 (17:35 -0600)]
loader: reduce message level if GDPA of a layer is not found
Change-Id: Ic4dd99015f5f3a07a6657f196fde89181d66c0f6
GregF [Mon, 16 May 2016 23:26:26 +0000 (17:26 -0600)]
update glslang and spirv-tools revisions
Rene Lindsay [Fri, 13 May 2016 16:19:41 +0000 (09:19 -0700)]
git: Ignore Qt generated files
Change-Id: I5970f83e8316e3f13e79e103cc17c96e31fef95e
Jon Ashburn [Mon, 16 May 2016 20:40:36 +0000 (14:40 -0600)]
misc: Update to version 1.0.13, header, xml, JSON files
Change-Id: Id70c1b709d35b2f508e0dde5f862102c87c80521
Jon Ashburn [Mon, 16 May 2016 20:01:18 +0000 (14:01 -0600)]
loader: deprecate device layers
Change-Id: I43c279e36368bf1ef9a2f446007e34366bfff777
Cody Northrop [Mon, 16 May 2016 15:09:15 +0000 (09:09 -0600)]
windows: Fix android-generate.bat
Mirror changes made to android-generate.sh in
cb9ce9e0
Jamie Madill [Wed, 11 May 2016 20:11:47 +0000 (16:11 -0400)]
layers: Fix VkBool32 return in vk-layer-generate.py.
This regressed in
f4542e618, but is only visible when defining
VK_USE_PLATFORM_WIN32_KHR, so might have been missed. Fixes a
warning in MS 2015, as well as an incorrect return value.
Jamie Madill [Tue, 10 May 2016 14:36:20 +0000 (07:36 -0700)]
layers: Add output file option to vk-generate scripts.
With some build systems, it's not easy to pipe stdout into a file.
This solution allows the scripts to use the last argument as an
output file option. This is a problem in partiuclar for GN.
Chris Forbes [Mon, 16 May 2016 02:09:35 +0000 (14:09 +1200)]
layers: Track dynamic descriptor offsets per set index
One big pool of dynamic offsets didn't work well:
- CmdBindDescriptorSets just appended, without regard for unaffected sets,
whether offsets were already recorded for the affected sets, etc. A
command buffer which bound descriptor sets many times would just
accumulate junk.
- DescriptorSet::ValidateSetNode was always passed the base of the
offset pool, when the set's first dynamic offset may have been some
distance into the pool.
Instead, track a vector of dynamic offsets per set index, alongside
the tracking of the bound descriptor set, which makes both correct
tracking of updates, and correct validation at draw time easier.
Fixes lots of weird behavior with more complex binding sequences.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Sun, 15 May 2016 23:19:50 +0000 (11:19 +1200)]
layers: Narrow scopes of some locals for clarity
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Norbert Nopper [Sun, 15 May 2016 17:19:41 +0000 (19:19 +0200)]
Fix allows to return actual storage_ and not default false value.
Tobin Ehlis [Thu, 12 May 2016 22:57:14 +0000 (16:57 -0600)]
layers: Clean up old mem_tracker merge remnants
Clean up most of the remaining MTMERGESOURCE defines.
Many of these come straight out. A number of copy functions
were able to merge some code to provide a slight efficiency gain.
Tobin Ehlis [Mon, 16 May 2016 14:36:40 +0000 (08:36 -0600)]
layers: Fix android build errors
Tobin Ehlis [Thu, 12 May 2016 14:42:09 +0000 (08:42 -0600)]
tests: Update doc validation
Print a single warning line identifying the number of missing tests.
Tobin Ehlis [Thu, 12 May 2016 14:32:31 +0000 (08:32 -0600)]
tests: Add InvalidMemoryMapping test
Tobin Ehlis [Thu, 12 May 2016 14:31:32 +0000 (08:31 -0600)]
layers: Fixes to memory mapping checks
MapMemory w/ size 0 is now invalid in the spec so remove comment and
change callback to ERROR.
Update some error messages for clarity.
Only update the mapped memory range state if MapMemory succeeds.
Tobin Ehlis [Wed, 11 May 2016 20:57:49 +0000 (14:57 -0600)]
tests: Update document validation related to tests
If a test is not done, standardize on "TODO" as the only allowed
text. Otherwise, the txt is considered to be a valid test and an
error will be flagged if the test cannot be found in
layer_validation_tests.cpp.
Tobin Ehlis [Wed, 11 May 2016 20:40:44 +0000 (14:40 -0600)]
layers: Kill MISSING_MEM_BINDINGS error
The MEMTRACK_MISSING_MEM_BINDINGS enum was dead code. There's no way
that the check can be hit in current code so added an assert() for
future safety and killed the checks and their documentation.
Tobin Ehlis [Wed, 11 May 2016 19:52:53 +0000 (13:52 -0600)]
tests: Add tests for INVALID_ALIASING cases
Chris Forbes [Fri, 13 May 2016 04:23:58 +0000 (16:23 +1200)]
layers: Move `fb` from RENDER_PASS_NODE to GLOBAL_CB_NODE
This is part of a renderpass instance, not of the renderpass itself.
If stored where MT put it, things go horribly wrong when recording
command buffers in parallel.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 13 May 2016 01:42:12 +0000 (13:42 +1200)]
layers: Finish unhooking layer_data from SC pipeline validation
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 13 May 2016 01:39:17 +0000 (13:39 +1200)]
layers: Use pipeline layout ptr from pipeline node
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 13 May 2016 01:35:20 +0000 (13:35 +1200)]
layers: Add PIPELINE_LAYOUT_NODE * to PIPELINE_NODE
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 13 May 2016 01:27:58 +0000 (13:27 +1200)]
layers: No longer require layer_data for stage validation
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 13 May 2016 01:17:42 +0000 (13:17 +1200)]
layers: Pass in shaderModuleMap to SC
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 13 May 2016 01:01:24 +0000 (13:01 +1200)]
layers: use RENDER_PASS_NODE* from pipeline in SC
Breaks one more dep from SC guts -> layer_data
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Chris Forbes [Fri, 13 May 2016 01:01:02 +0000 (13:01 +1200)]
layers: Add RENDER_PASS_NODE* to PIPELINE_NODE*
Signed-off-by: Chris Forbes <chrisforbes@google.com>
Mark Lobodzinski [Thu, 12 May 2016 21:14:07 +0000 (15:14 -0600)]
layers: LX518, Fix ObjectTracker memory leaks
Several places in object_tracker were not properly deleting data
structures, which caused signicant memory leaks.
Change-Id: Ie64ab18efecc45838e103dd7a8cb5f75c92155c4
Ian Elliott [Mon, 9 May 2016 19:54:01 +0000 (13:54 -0600)]
layers: PR512 Update some of the swapchain errors that are now tested.
Ian Elliott [Mon, 9 May 2016 18:33:47 +0000 (12:33 -0600)]
layers: PR512 Fix swapchain's validateCreateSwapchainKHR()
It used to dereference pCreateInfo, even when it was NULL.
Karl Schultz [Thu, 12 May 2016 14:11:23 +0000 (08:11 -0600)]
tests: GH431 PR 510 Apply suggested changes
- Convert error to warning for overlapping ranges
- Consider overlapping ranges in range checks
- User a reference to avoid a copy
- Many more additional tests
Change-Id: I18ad43b98684f16a978461a1a089ebff5437a563