platform/upstream/Vulkan-Tools.git
8 years agolayers: Fixup OT error messages
Mark Lobodzinski [Mon, 18 Jul 2016 23:06:52 +0000 (17:06 -0600)]
layers: Fixup OT error messages

Print object type explicitly; also fixup a couple of format
specifiers.

Change-Id: I6f5036f70cae05bd5f7ee9320e24b2e16bdc811b

8 years agolayers: Fix thread-safety fails in OT FreeXxx APIs
Mark Lobodzinski [Wed, 13 Jul 2016 17:29:00 +0000 (11:29 -0600)]
layers: Fix thread-safety fails in OT FreeXxx APIs

Tracking data was cleaned up after calling down the chain in
FreeCommandBuffers and FreeDescriptorSets -- this left a hole that
another thread could walk through and corrupt the tracking data.

Change-Id: I7cc9c9bfae9f1ea179122125dd09952650f14d81

8 years agolayers: Update object_tracker layer
Mark Lobodzinski [Fri, 1 Jul 2016 16:53:31 +0000 (10:53 -0600)]
layers: Update object_tracker layer

Removed dead code, added per-instance and per-device object tracking,
(this was broken in codegen'd version with no cross-device
validation), updated for new layer architectures and coding standards,
removed OT-related codegen -- it is now a standalone cpp file.

Change-Id: I64464b855e1b4841c8e3a581387e0e9065b006f7

8 years agolayers : gh682 move gipa dispatch table init
Mark Young [Tue, 5 Jul 2016 22:41:50 +0000 (16:41 -0600)]
layers : gh682 move gipa dispatch table init

Move the GIPA setup of the dispatch tables for layers into
the layer_init_device_dispatch_table call since we're already
doing a majority of it there.  This removes the need to separately
setup the WSI extension entry-points.

Additionally, memset the table to 0, just to make sure anything
that gets added in the future is obvoiusly not set.

Change-Id: I63c7c107cd9f7957e2766fdbeb8a06bc1ae0eda6

8 years agodocs: Remove extra line in doc headers
Mark Young [Thu, 21 Jul 2016 15:45:32 +0000 (09:45 -0600)]
docs: Remove extra line in doc headers

GitHub markdown adds a line after H1 and H2 entries.  Normal
markdown doesn't do this.  So my WSYIWYG editor incorrectly showed
the results of my previous edits.

Change-Id: I3c197c1fb1f364bf4b1050d1eaa1141affefe9d1

8 years agomisc: Support for cross-platform trace/replay
David Pinedo [Thu, 21 Jul 2016 16:56:27 +0000 (10:56 -0600)]
misc: Support for cross-platform trace/replay

Change-Id: If77890356911fa90d22af28a449a05284ad2097f

8 years agolayers: Check BUFFER_CREATE_SPARSE_BINDING_BIT for buffers
Tobin Ehlis [Thu, 21 Jul 2016 12:46:12 +0000 (06:46 -0600)]
layers: Check BUFFER_CREATE_SPARSE_BINDING_BIT for buffers

8 years agolayers: Validation buffer descriptor updates for bound memory
Tobin Ehlis [Wed, 20 Jul 2016 20:04:20 +0000 (14:04 -0600)]
layers: Validation buffer descriptor updates for bound memory

When updating buffer descriptors, make sure that there is memory bound to any
non-sparse buffer.
Add comment that we don't need to do this for image descriptor updates
because they require the imageView (not just image) and we already validate
at imageView creation that memory is bound to image.

8 years agolayers: Refactor buffer/image validation functions
Tobin Ehlis [Wed, 20 Jul 2016 19:27:33 +0000 (13:27 -0600)]
layers: Refactor buffer/image validation functions

Add some consts and put fwd decls in core_validation_types.h so that these
can be used by descriptor_set class.

8 years agoloader: Clean up some things in the MD file
Mark Young [Wed, 20 Jul 2016 17:38:53 +0000 (11:38 -0600)]
loader: Clean up some things in the MD file

Clean up some documentation in the LoaderAndLayerInterface
markdown.  Over the next few weeks, I'll be working on
cleaning up the format and language of this doc.

Change-Id: I7858981293de9befb34701c31542d88c664cd6d8

8 years agolayers: Change pNext validation error to warning
Dustin Graves [Tue, 19 Jul 2016 19:43:53 +0000 (13:43 -0600)]
layers: Change pNext validation error to warning

Change the severity level of the parameter_validation layer's pNext
parameter validation messages from error to warning.  Also add an
explanation that the warning is derived from the Valid Usage
language from version X of the Vulkan spec/header, which may be out
of date:

This warning is based on the Valid Usage documentation for version %d
of the Vulkan header.  It is possible that you are using a struct
from a private extension or an extension that was added to a later
version of the Vulkan header, in which case your use of pNext is
perfectly valid but is not guaranteed to work correctly with
validation enabled

Issues-Addressed: LunarXchange 560
Change-Id: I97f55cba7538506d246e45ef603a86db9ad1adcb

8 years agolayers: Add codegen header version
Dustin Graves [Tue, 19 Jul 2016 19:17:35 +0000 (13:17 -0600)]
layers: Add codegen header version

Add an integer constant to the parameter_validation layer's generated
code to specify the header/XML version from which the code was
generated.

Change-Id: Id690962856130007a24732a478ffa9c28f774b8f

8 years agocodegen: Sync with Vulkan-Docs vk.xml
Dustin Graves [Mon, 18 Jul 2016 20:54:37 +0000 (14:54 -0600)]
codegen: Sync with Vulkan-Docs vk.xml

Update the local copy of vk.xml with header 21 changes from the
original Vulkan-Docs vk.xml.

Change-Id: I48af9f50ad31c79be2ef2c78c4a33392f7d64d51

8 years agolayers: Fix Windows build warning
Mark Lobodzinski [Wed, 20 Jul 2016 16:02:23 +0000 (10:02 -0600)]
layers: Fix Windows build warning

Change-Id: Ic40ff6cb79882be92abbb2d8c02298962c73e46f

8 years agolayers: Fix an MSVS error and a warning.
Jamie Madill [Wed, 6 Jul 2016 17:37:33 +0000 (13:37 -0400)]
layers: Fix an MSVS error and a warning.

Fatal error C1017: invalid integer constant expression
warning C4389: '!=': signed/unsigned mismatch

Change-Id: I13e06f1578e47a0076241e1a344f4f05ae6d50c2

8 years agolayers: Clarify std_validation usage in README
Mark Lobodzinski [Wed, 20 Jul 2016 15:15:53 +0000 (09:15 -0600)]
layers: Clarify std_validation usage in README

Change-Id: Ia99542ca3f96d719f60d5dccab115340bd88c3f3

8 years agolayers: Relax SC descriptor binding uniqueness to allow split variables
Chris Forbes [Tue, 19 Jul 2016 03:08:38 +0000 (15:08 +1200)]
layers: Relax SC descriptor binding uniqueness to allow split variables

This allows the separate sampler & image variables case from 13.1.4 to
work correctly.

Signed-off-by: Chris Forbes <chrisforbes@google.com>
8 years agolayers: Allow consuming COMBINED_IMAGE_SAMPLER as separate pieces
Chris Forbes [Tue, 19 Jul 2016 02:50:04 +0000 (14:50 +1200)]
layers: Allow consuming COMBINED_IMAGE_SAMPLER as separate pieces

Signed-off-by: Chris Forbes <chrisforbes@google.com>
8 years agotests: Add test exceptions to doc validator
Tobin Ehlis [Mon, 18 Jul 2016 23:11:29 +0000 (17:11 -0600)]
tests: Add test exceptions to doc validator

Don't require test cases for internal layer error cases.

8 years agotests: Remove old '_NA' doc check code
Tobin Ehlis [Mon, 18 Jul 2016 23:04:15 +0000 (17:04 -0600)]
tests: Remove old '_NA' doc check code

There are no "_NA" checks remaining so kill exception code for it.

8 years agolayers: Kill redundant DRAWSTATE_PUSH_CONSTANTS_ERROR case
Tobin Ehlis [Mon, 18 Jul 2016 22:22:03 +0000 (16:22 -0600)]
layers: Kill redundant DRAWSTATE_PUSH_CONSTANTS_ERROR case

This check was really an INVALID_PIPELINE check that is already
handled by object_tracker so killing it in core_validation.

8 years agotests: Update doc for PUSH_CONSTANTS_ERROR
Tobin Ehlis [Mon, 18 Jul 2016 22:19:42 +0000 (16:19 -0600)]
tests: Update doc for PUSH_CONSTANTS_ERROR

The DRAWSTATE_PUSH_CONSTANTS_ERROR cases are all hit by InvalidPushConstants
test.

8 years agolayers: Replace INVALID_EVENT check with OBJECT_INUSE
Tobin Ehlis [Mon, 18 Jul 2016 20:26:18 +0000 (14:26 -0600)]
layers: Replace INVALID_EVENT check with OBJECT_INUSE

If an event being destroyed is in use, the correct error enum to use
is DRAWSTATE_OBJECT_INUSE.

8 years agolayers: Replace INVALID_SEMAPHORE with OBJECT_INUSE check
Tobin Ehlis [Mon, 18 Jul 2016 20:19:12 +0000 (14:19 -0600)]
layers: Replace INVALID_SEMAPHORE with OBJECT_INUSE check

DRAWSTATE_INVALID_SEMAPHORE was bad name for what is really another
instance of DRAWSTATE_OBJECT_INUSE check. Updated the check to use
the new enum and removed the doc entry for the old check which is
no longer used.

8 years agotests: Remove duplicate check column from doc
Tobin Ehlis [Mon, 18 Jul 2016 19:49:22 +0000 (13:49 -0600)]
tests: Remove duplicate check column from doc

8 years agotests: Updated test doc
Tobin Ehlis [Mon, 18 Jul 2016 19:39:24 +0000 (13:39 -0600)]
tests: Updated test doc

Clarified cases to be checked for OBJECT_INUSE, INVALID_QUERY and
INVALID_FENCE cases.

8 years agotests: Update doc for CANT_FREE_FROM_NON_FREE_POOL
Tobin Ehlis [Mon, 18 Jul 2016 19:25:08 +0000 (13:25 -0600)]
tests: Update doc for CANT_FREE_FROM_NON_FREE_POOL

This case is correctly verified by test FreeDescriptorFromOneShotPool.

8 years agotests: Update doc for DRAWSTATE_INVALID_DESCRIPTOR_SET
Tobin Ehlis [Mon, 18 Jul 2016 19:23:23 +0000 (13:23 -0600)]
tests: Update doc for DRAWSTATE_INVALID_DESCRIPTOR_SET

Capture another affected API and update Notes.

8 years agolayers: Remove DRAWSTATE_INVALID_POOL check
Tobin Ehlis [Mon, 18 Jul 2016 19:14:01 +0000 (13:14 -0600)]
layers: Remove DRAWSTATE_INVALID_POOL check

This check was only flagged when Allocating descriptor sets or resetting
a descriptor pool, and in both cases object_tracker will be the first layer
to identify and flag the error.

Removing this as a redundant check and updating documentation.

8 years agotests: Update doc for DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH
Tobin Ehlis [Wed, 13 Jul 2016 20:54:17 +0000 (14:54 -0600)]
tests: Update doc for DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH

All of these cases are hit by various tests so added those to the doc.

8 years agotests: Update doc for DRAWSTATE_INVALID_COMMAND_BUFFER
Tobin Ehlis [Wed, 13 Jul 2016 20:48:32 +0000 (14:48 -0600)]
tests: Update doc for DRAWSTATE_INVALID_COMMAND_BUFFER

A few of these cases are already hit so added those test names.
Most of the cases are still not hit by testing, those, so broke
those out in the Notes column.

8 years agotests: Remove DRAWSTATE_INVALID_PIPELINE_LAYOUT enum value
Tobin Ehlis [Wed, 13 Jul 2016 20:42:50 +0000 (14:42 -0600)]
tests: Remove DRAWSTATE_INVALID_PIPELINE_LAYOUT enum value

This was no longer used anywhere so remove from enum and doc.

8 years agotests: Update doc for CB_SIMULTANEOUS_USE check
Tobin Ehlis [Wed, 13 Jul 2016 20:37:49 +0000 (14:37 -0600)]
tests: Update doc for CB_SIMULTANEOUS_USE check

There are 3 cases of this check in validation. The first case is hit by
CommandBufferTwoSubmits test so added that.
The other 2 cases are both in CmdExecuteCommands() and are not hit to made
a note of that for future work.

8 years agoloader: Clarify in docs about WSI and inst ext.
Mark Young [Tue, 19 Jul 2016 17:49:45 +0000 (11:49 -0600)]
loader: Clarify in docs about WSI and inst ext.

Clarify in the LoaderAndLayerInterface markdown file what WSI
extensions are available by default in the desktop loader.
Additionally, clarify that instance extensions are not supported
by the Loader unless they have been specifically added to the
Loader.

Change-Id: I21bd2a63ed1e4a7e50d27e7a6b136eeeaea4d68c

8 years agolayers: GH773 Check improper renderpass layout for only first use
Tobin Ehlis [Tue, 19 Jul 2016 01:01:43 +0000 (19:01 -0600)]
layers: GH773 Check improper renderpass layout for only first use

Only verify that a LOAD_OP attachment does not have first layout of *READ_ONLY*
type on its first use. The spec language this is checking is from section
"7.1 Render Pass Creation" :

The first use of an attachment must not specify a layout equal to
VK_IMAGE_LAYOUT_DEPTH_STENCIL_READ_ONLY_OPTIMAL or VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL
if the attachment specifies that the loadOp is VK_ATTACHMENT_LOAD_OP_CLEAR.

Previously we were checking this condition on all uses of an attachment
which would incorrectly flag errors on proper uses beyond the first use by
later subpasses.

Note that validation for CreateRenderPass has a fair amount of duplication and
still needs to be cleaned up to consolidate work that's currently being done
before and after the call down the chain. This is simply a singular bug fix.

8 years agomisc: update spirv tools revision
Karl Schultz [Mon, 18 Jul 2016 22:40:49 +0000 (16:40 -0600)]
misc: update spirv tools revision

Change-Id: I8abb31de604ae4a50bb4900e9498de9953c5646b

8 years agolayers: NV_dedicated_allocation support for unique_objects
Dustin Graves [Thu, 14 Jul 2016 23:28:11 +0000 (17:28 -0600)]
layers: NV_dedicated_allocation support for unique_objects

Add ID substitution support for the
VkDedicatedAllocationMemoryAllocateInfoNV extension structure to the
unique_objects layer.  The current implementation is specific to the
case where the vkAllocateMemory pAllocateInfo parameter references
a single VkDedicatedAllocationMemoryAllocateInfoNV struct.

Issues-Addressed: GitHub #755
Change-Id: I33c3f00d32149044694ecafe808673e5d90107b4

8 years agomisc: Update to Vulkan header version 21
Karl Schultz [Mon, 18 Jul 2016 18:04:39 +0000 (12:04 -0600)]
misc: Update to Vulkan header version 21

Change-Id: I112bb3d53daf067b231532620f45cd8891537cbb

8 years agobuild: Update glslang to e4821e
Karl Schultz [Mon, 18 Jul 2016 15:40:37 +0000 (09:40 -0600)]
build: Update glslang to e4821e

This includes applying a patch to glslang to revert glslang
commit a5c33d, which fixes a texel fetch problem that many IHV
drivers are not ready to cope with yet.

8 years agolayers: Fix minor completeness/performance issue
Mark Mueller [Wed, 13 Jul 2016 20:49:35 +0000 (14:49 -0600)]
layers: Fix minor completeness/performance issue

The first time the condition is discovered it will
be the same as every subsequent time, so there is
no need to continue. Also, the whole structure must
be compared, so memcmp assures that will happen
with good efficiency

Change-Id: I71f608c71f83ebcd02212bb391a30bdf15279ff6

8 years agowindows: Return exit code from build script if any step fails
Cody Northrop [Wed, 13 Jul 2016 21:59:22 +0000 (15:59 -0600)]
windows: Return exit code from build script if any step fails

8 years agolayers: Update clearValueCount check
Tobin Ehlis [Fri, 15 Jul 2016 22:01:13 +0000 (16:01 -0600)]
layers: Update clearValueCount check

The pClearValues array is indexed by the attachment index which
means that it must be at least as large as the last LOAD_OP_CLEAR
attachment in the renderpass. Updating this check to reflect this
requirement.

8 years agolayers: PR767, Fix crash exiting smoketest
Mark Lobodzinski [Fri, 15 Jul 2016 19:53:44 +0000 (13:53 -0600)]
layers: PR767, Fix crash exiting smoketest

Commit a86b57c caused a Windows crash when shutting down smoketest.
Command buffer cleanup for object bindings wasn't happening when
calling DestroyCommandPool.

Change-Id: Ie233763e37ed9bc6988dddbf932aa3eda2960940

8 years agolayers: LX555/GH663, Update details doc
Mark Lobodzinski [Thu, 14 Jul 2016 18:09:12 +0000 (12:09 -0600)]
layers: LX555/GH663, Update details doc

Add entry for new buffer/image memory binding checks.

Change-Id: Ic02c97cd8dd7e054ca67a66d82fc5728bf9167be

8 years agolayers: LX555/GH663, Add BindXxxMemory validation
Mark Lobodzinski [Thu, 14 Jul 2016 15:54:11 +0000 (09:54 -0600)]
layers: LX555/GH663, Add BindXxxMemory validation

If the user neglected to call BindImageMemory or BindBufferMemory,
validation gave not indication of the failure which then results in
a driver or layer crash. This commit extends and generalizes the
checks done for CreateImageView and CreateBufferView to cover all
the bound memory checks for the 'recording commands in command
buffer' case in spec section 11.6

Change-Id: I3b4a3d130e9b6c242413fc0ad14c1624295b33b3

8 years agolayers: Handle binding between cmd buffer and query pools
Tobin Ehlis [Fri, 8 Jul 2016 20:22:01 +0000 (14:22 -0600)]
layers: Handle binding between cmd buffer and query pools

For the CmdBind* functions that involve queryPool object, add a binding between
the cmd buffer and the queryPool.

In the event that a queryPool is destroyed, invalidate any bound cmd buffers
and make sure invalid reporting is handled correctly for query pools.

When a cmd buffer is destroyed that has query pool bindings, those bindings
will be cleared under existing object_binding cleanup in resetCB.

8 years agolayers: Set cmd buffer invalid on event destroy
Tobin Ehlis [Thu, 7 Jul 2016 22:47:10 +0000 (16:47 -0600)]
layers: Set cmd buffer invalid on event destroy

When an event is destroyed, set any cmd buffers that it was bound
to to the CB_INVALID state.

Also added getEventNode() helper and made use of it where appropriate.

8 years agolayers: Rework Destroy/Free data structure cleanup
Tobin Ehlis [Wed, 13 Jul 2016 15:41:09 +0000 (09:41 -0600)]
layers: Rework Destroy/Free data structure cleanup

When destroying dispatchable objects we must cleanup internal data
structures prior to calling down the chain in the event that another
thread were to prempt execution and re-allocate a freed handle prior
to the layer cleanup code executing.

For non-dispatchable handles this shouldn't be a requirement due to
the fact that layers only see unique handles. However, moved some
non-dispatchable cleanup before the call to ICD as well for consistency
and to be extra safe.

8 years agolayers: threading with cmd buffer handle reuse
Mike Stroyan [Wed, 13 Jul 2016 17:07:27 +0000 (11:07 -0600)]
layers: threading with cmd buffer handle reuse

A driver may reuse a command buffer handle in another thread.
The threading layer needs to remove command buffer information
before calling FreeCommandBuffers.
This addresses part of github issue #719.

8 years agolayers: Remove unused vars from vk-layer-generate.py
Tobin Ehlis [Thu, 14 Jul 2016 20:45:54 +0000 (14:45 -0600)]
layers: Remove unused vars from vk-layer-generate.py

8 years agolayers: On Free/Destroy fix unique_objects map cleanup
Tobin Ehlis [Thu, 14 Jul 2016 20:37:51 +0000 (14:37 -0600)]
layers: On Free/Destroy fix unique_objects map cleanup

When an object is freed or destroyed, unique_objects must clean up its
map entry prior to calling down the chain. This is becase the ICD may
be preempted just after object deletion, and the same object handle
could then be allocated before returning to validation. In that case,
unique_objects would delete the entry for the newly-created object.

8 years agolayers: faster threading layer for single thread
Mike Stroyan [Wed, 13 Jul 2016 16:10:25 +0000 (10:10 -0600)]
layers: faster threading layer for single thread

Add a check to the threading layer to detect a single thread case.
If the application has only called vulkan from one thread at a time,
then skip access counters for externally synchronized parameters.
This greatly reduces the overhead of the layer for applications
that don't truly use vulkan in a multi-threaded way.

8 years agolayers: Use type-specific mutexes in thread layer
Mike Stroyan [Thu, 30 Jun 2016 19:21:37 +0000 (13:21 -0600)]
layers: Use type-specific mutexes in thread layer

Use type-specific mutexes to reduce lock contention.

8 years agoloader: mr739 - Link VKstatic against shlwapi
Nekotekina [Fri, 8 Jul 2016 12:49:13 +0000 (15:49 +0300)]
loader: mr739 - Link VKstatic against shlwapi

Change-Id: I40345f39ece1d580139cb09ae01c1a36d43016dd

8 years agolayers: Fix core_validation Win32 build
Mark Lobodzinski [Wed, 13 Jul 2016 20:59:41 +0000 (14:59 -0600)]
layers: Fix core_validation Win32 build

Change-Id: Ib15d5c6fad234c0d657b59613293f38d28969697

8 years agolayers: Add binding between cmd buffer and pipeline
Tobin Ehlis [Fri, 8 Jul 2016 21:42:38 +0000 (15:42 -0600)]
layers: Add binding between cmd buffer and pipeline

Track all pipelines bound to a cmd buffer and if any of them are destroyed
set the cmd buffer as invalid.

8 years agolayers: Handle binding between cmd buffer and query pools
Tobin Ehlis [Fri, 8 Jul 2016 20:22:01 +0000 (14:22 -0600)]
layers: Handle binding between cmd buffer and query pools

For the CmdBind* functions that involve queryPool object, add a binding between
the cmd buffer and the queryPool.

In the event that a queryPool is destroyed, invalidate any bound cmd buffers
and make sure invalid reporting is handled correctly for query pools.

When a cmd buffer is destroyed that has query pool bindings, those bindings
will be cleared under existing object_binding cleanup in resetCB.

8 years agolayers: Add addCommandBufferBinding() utility function
Tobin Ehlis [Fri, 8 Jul 2016 19:27:48 +0000 (13:27 -0600)]
layers: Add addCommandBufferBinding() utility function

Created addCommandBufferBinding() utility function that creates a link between the
cb_binding data struct of an object and the object_binding data struct of the cmd
buffer.

Currently updated events to use this utility function, but will expand to other
object types in the future.

8 years agolayers: Set cmd buffer invalid on event destroy
Tobin Ehlis [Thu, 7 Jul 2016 22:47:10 +0000 (16:47 -0600)]
layers: Set cmd buffer invalid on event destroy

When an event is destroyed, set any cmd buffers that it was bound
to to the CB_INVALID state.

Also added getEventNode() helper and made use of it where appropriate.

8 years agodemos: Fix android build
Cody Northrop [Wed, 13 Jul 2016 16:01:10 +0000 (10:01 -0600)]
demos: Fix android build

8 years agodocs: GH422 Addition of Buffer Validation Tests
Mark Mueller [Thu, 7 Jul 2016 21:10:06 +0000 (15:10 -0600)]
docs: GH422 Addition of Buffer Validation Tests

Reflect new tests in documentation

Change-Id: Ib54179e4a1cea6aa38d2cf616bcf944c2e65eb08

8 years agolayers: GH422 Support addition of Buffer Validation Tests
Mark Mueller [Thu, 7 Jul 2016 17:57:59 +0000 (11:57 -0600)]
layers: GH422 Support addition of Buffer Validation Tests

Remove dead code by getting rid of deleteMemObjInfo function,
add check for freeMemObjInfo failure, and fix an incorrect
validation test

Change-Id: Ie84d16883abfb6d4d41b06d6fa210c7011861ecd

8 years agolayers: Handle special layerCount case
Dustin Graves [Tue, 12 Jul 2016 19:57:02 +0000 (13:57 -0600)]
layers: Handle special layerCount case

Make parameter validation logic for VkImageSubresource ignore
layerCount when it is set to VK_REMAINING_ARRAY_LAYERS.

Change-Id: Ib2e50e678c79f00850e085fe39c22631f9d00636
Issues-Addressed: GitHub VulkanTools #58

8 years agolayers: Don't use instance pointer after it has been destroyed
Cody Northrop [Tue, 12 Jul 2016 17:43:32 +0000 (11:43 -0600)]
layers: Don't use instance pointer after it has been destroyed

In unique_objects' explicit_DestroyInstance(), the recently introduced
second get_dispatch_table() was using a destroyed instance as a lookup
key, failing to find an entry (since the driver controls that memory
and probably freed it), and was crashing consistently on PixelC.

Instead, use the instance pointer to cleanup instanceExtMap before
calling down to the ICD's DestroyInstance.

Additionally, use the existing VkLayerInstanceDispatchTable* instead
of calling get_dispatch_table() twice.

8 years agolayers: Added return check for freeMemObjInfo call
Mark Mueller [Mon, 11 Jul 2016 21:03:44 +0000 (15:03 -0600)]
layers: Added return check for freeMemObjInfo call

In FreeMemory, the return value of freeMemObjInfo was
being ignored, which incorrectly allowed calls which
failed validation to still be passed to lower levels.
This change corrects that.

Change-Id: Iddbffad22f9502061441e961daef2e7fa26a5fcf

8 years agomisc: Update to Vulkan header version 20
Karl Schultz [Mon, 11 Jul 2016 21:09:57 +0000 (15:09 -0600)]
misc: Update to Vulkan header version 20

Changes:
vulkan.h
json files
linux SO version
loader and tests to reflect API change in vkCmdUpdateBuffer
vk.xml to match Vulkan-Docs

Change-Id: I1763bca6fccf619ad8e76c0363f68fa38ad69900

8 years agolayers: Fix WIN32 build
Dustin Graves [Fri, 8 Jul 2016 21:03:39 +0000 (15:03 -0600)]
layers: Fix WIN32 build

Fix a uint64_t to VkImage reinterpret_cast for 32-bit VS2013.  I
Missed this when reviewing the previous change.

Change-Id: I4a0865ab546b45792b8f4d841f6c6c04ff22501b

8 years agolayers: Remove cb_bindings when CB is reset
Tobin Ehlis [Fri, 8 Jul 2016 18:33:45 +0000 (12:33 -0600)]
layers: Remove cb_bindings when CB is reset

When a CB is reset (or freed) make sure to remove it from any image/buffer
cb_bindings that it was tied to.

8 years agodemos: Remove VkCommandBufferInheritanceInfo
Rene Lindsay [Thu, 7 Jul 2016 23:00:14 +0000 (16:00 -0700)]
demos: Remove VkCommandBufferInheritanceInfo

VkCommandBufferInheritanceInfo is only required for secondary buffers.
Remove from cube and tri.

Change-Id: I0000000000000decaf15bad00000000000000028

8 years agowinsdk: Fix error when installing from syswow64
Lenny Komow [Thu, 7 Jul 2016 22:45:34 +0000 (16:45 -0600)]
winsdk: Fix error when installing from syswow64

8 years agolayers: Hold lock during pipelineLayoutMap.erase()
Tobin Ehlis [Thu, 7 Jul 2016 19:40:29 +0000 (13:40 -0600)]
layers: Hold lock during pipelineLayoutMap.erase()

8 years agolayers: Remove entries from pipelineLayoutMap
Tobin Ehlis [Thu, 7 Jul 2016 18:04:20 +0000 (12:04 -0600)]
layers: Remove entries from pipelineLayoutMap

When VkPipelineLayout objects are destroyed, remove their corresponding
map entries.

8 years agolayers: Update layout compatibility to use stored state
Tobin Ehlis [Thu, 7 Jul 2016 17:06:26 +0000 (11:06 -0600)]
layers: Update layout compatibility to use stored state

When checking VkPipelineLayout compatibility, it's possible that the
VkPipelineLayout handle that was bound may have been subsequently deleted.

To account for this, store the PIPELINE_LAYOUT_NODE at bind time and
use that to do the compatibility check.

8 years agolayers: Rename vars
Tobin Ehlis [Thu, 7 Jul 2016 15:23:38 +0000 (09:23 -0600)]
layers: Rename vars

Migrate PIPELINE_LAYOUT_NODE members from camelcase.

8 years agolayers: Simplify PIPELINE_LAYOUT_NODE
Tobin Ehlis [Thu, 7 Jul 2016 15:20:13 +0000 (09:20 -0600)]
layers: Simplify PIPELINE_LAYOUT_NODE

Just use setLayout vector and kill duplicate descriptorSetLayout vector.

8 years agolayers: Store PIPELINE_LAYOUT_NODE directly in PIPELINE_NODE
Tobin Ehlis [Thu, 7 Jul 2016 14:14:48 +0000 (08:14 -0600)]
layers: Store PIPELINE_LAYOUT_NODE directly in PIPELINE_NODE

The layout and its associated node may be deleted after pipeline is
created so store a copy of the PIPELINE_LAYOUT_NODE in the
PIPELINE_NODE to make sure data is available when it's needed.

8 years agolayers: Rename pipelineLayout to pipeline_layout
Tobin Ehlis [Thu, 7 Jul 2016 14:12:03 +0000 (08:12 -0600)]
layers: Rename pipelineLayout to pipeline_layout

8 years agobuild: patch glslang to fix g++ 5.3 compilation issue
Karl Schultz [Thu, 7 Jul 2016 16:41:25 +0000 (10:41 -0600)]
build: patch glslang to fix g++ 5.3 compilation issue

Add GLSLANG_REVISION-sensitive step to update_external_sources
to patch glslang to allow compilation on Ubuntu 16.10,
Arch Linux and other platforms using g++ 5.3.

This patch won't happen if the GLSLANG_REVISION is changed.

Change-Id: I84a501cb44ee527f45f499d89b2745683e0c8e70

8 years agolayers: GH689 Disable perf warning for linear tiled textures
Tobin Ehlis [Wed, 6 Jul 2016 17:32:12 +0000 (11:32 -0600)]
layers: GH689 Disable perf warning for linear tiled textures

Linear tiled textures should use GENERAL layout so perf warning doesn't apply.

8 years agolayers: Minor refactor to save duplicate look-ups
Tobin Ehlis [Wed, 6 Jul 2016 17:23:05 +0000 (11:23 -0600)]
layers: Minor refactor to save duplicate look-ups

Pass layer_data* and GLOBAL_CB_NODE* to Verify[Dest|Source]ImageLayout() to avoid
a second look-up of the same data.

8 years agolayers: Fix handling of component qualifier on VS inputs
Chris Forbes [Wed, 6 Jul 2016 00:18:26 +0000 (12:18 +1200)]
layers: Fix handling of component qualifier on VS inputs

Interface matcher assumed no location aliasing, but this is allowed.

Signed-off-by: Chris Forbes <chrisforbes@google.com>
8 years agolayers: Migrate to validateImageUsageFlags function
Tobin Ehlis [Wed, 6 Jul 2016 14:36:19 +0000 (08:36 -0600)]
layers: Migrate to validateImageUsageFlags function

At ImageView creation time use validateImageUsageFlags function that avoids an
additional look-up for IMAGE_NODE. Kill previous validate_image_usage_flags()
function.

8 years agolayers: Remove validate_buffer_usage_flags() function
Tobin Ehlis [Wed, 29 Jun 2016 15:07:52 +0000 (09:07 -0600)]
layers: Remove validate_buffer_usage_flags() function

Replaced with validateBufferUsageFlags()

8 years agolayers: Invalidate Cmd Buffers on DestroyBuffer
Tobin Ehlis [Tue, 28 Jun 2016 22:36:47 +0000 (16:36 -0600)]
layers: Invalidate Cmd Buffers on DestroyBuffer

When buffer is destroyed, any command buffers that it's bound to must
be invalidated.

8 years agolayers: Invalidate Cmd Buffers on DestroyImage
Tobin Ehlis [Tue, 28 Jun 2016 22:24:17 +0000 (16:24 -0600)]
layers: Invalidate Cmd Buffers on DestroyImage

When image is destroyed, any command buffers that it's bound to must
be invalidated. Also clean-up DestroyImage() a bit.

8 years agolayers: Add binding between images and cmd buffer
Tobin Ehlis [Tue, 28 Jun 2016 21:52:55 +0000 (15:52 -0600)]
layers: Add binding between images and cmd buffer

For cmds that use images, update cb_bindings to tie image to cmd
buffer. This will be used to track invalid cmd buffers due to
destroying images.

The function structure here is the same but I've updated all of the
calls to use mem bindings from IMAGE_NODE with a single call rather
than the old two-call mechanism.

The assumption is that parameter_checker should be above
core_validation so I haven't replicated a bunch of its checks in the
case of invalid objects. I've added some asserts as a last line of
defense to make debugging invalid object issues in core_validation
simpler, but if running with full validation stack and responding to
validation errors from parameter_checker, then the asserts should
never be hit.

8 years agolayers: Add image to IMAGE_NODE
Tobin Ehlis [Tue, 28 Jun 2016 20:52:11 +0000 (14:52 -0600)]
layers: Add image to IMAGE_NODE

Store image inside the image ptr so that once node is looked-up
all relevent data is available when just node ptr is passed around.

8 years agolayers: Add binding between buffer and cmd buffer
Tobin Ehlis [Tue, 28 Jun 2016 20:04:54 +0000 (14:04 -0600)]
layers: Add binding between buffer and cmd buffer

Updated call mechanism to create binding between a VkBuffer
and a VkCommandBuffer so that it occurs in a single call.
Removed old mem_tracker mechanism for looking up memory from
a buffer and just use mem in BUFFER_NODE instead.
Also simplified call to check valid buffer use flags in order
to avoid a redundant map look-up.

Will make similar changes for images next.

8 years agolayers: Bulk rename skipCall -> skip_call
Tobin Ehlis [Tue, 28 Jun 2016 16:28:02 +0000 (10:28 -0600)]
layers: Bulk rename skipCall -> skip_call

Trying to migrate to consistent lowercase variable naming and
skip_call was different all over the place to unifying in a
single commit.

8 years agoloader: Fix 2 warnings detected in VS2015.
Jamie Madill [Wed, 6 Jul 2016 15:19:42 +0000 (11:19 -0400)]
loader: Fix 2 warnings detected in VS2015.

warning C4456: declaration of 'i' hides previous local declaration
warning C4244: '=': conversion from 'int' to 'uint16_t', possible loss of data

Change-Id: I651de053b206afa36fce27af92ffbe0a608e89d0

8 years agolayers: Remove device_limits from Android build
Mark Lobodzinski [Wed, 29 Jun 2016 19:58:52 +0000 (13:58 -0600)]
layers: Remove device_limits from Android build

Change-Id: I4d63b740f046ee8378267082abc3792aa7a85418

8 years agoscripts: Fix device_limits details doc validation
Mark Lobodzinski [Wed, 29 Jun 2016 20:24:50 +0000 (14:24 -0600)]
scripts: Fix device_limits details doc validation

Change-Id: Idd383c1a28fa7007449fe24a205c2164b21dfe09

8 years agoloader: Remove device_limits layer
Mark Lobodzinski [Wed, 29 Jun 2016 20:18:30 +0000 (14:18 -0600)]
loader: Remove device_limits layer

Change-Id: I2660be2eb099ab6300b78e364991ca23f0abf26d

8 years agolayers: Deprecate device_limits layer
Mark Lobodzinski [Wed, 29 Jun 2016 20:15:10 +0000 (14:15 -0600)]
layers: Deprecate device_limits layer

Change-Id: Idd440935f3329d017ba4ee21d9ce87e432eec98e

8 years agodemos: Remove device_limits layer from cube and tri
Mark Lobodzinski [Wed, 29 Jun 2016 20:01:14 +0000 (14:01 -0600)]
demos: Remove device_limits layer from cube and tri

Change-Id: I834133e5054668433c355408886f46c042a2ec6f

8 years agotests: Add tests for invalid VkDescriptorBufferInfo
Tobin Ehlis [Sat, 2 Jul 2016 00:23:58 +0000 (18:23 -0600)]
tests: Add tests for invalid VkDescriptorBufferInfo

Hit 3 new checks just added for VkDescriptorBufferInfo:
1. offset larger than buffer
2. range of 0
3. range larger than buffer size minus offset

Also update descriptor write and copy update error enums to be more clear,
along with also updating documentation.

8 years agolayers: Add validation for VkDescriptorBufferInfo struct
Tobin Ehlis [Fri, 1 Jul 2016 23:36:48 +0000 (17:36 -0600)]
layers: Add validation for VkDescriptorBufferInfo struct

For a descriptor write update to a buffer type, verify that the contents of
VkDescriptorBufferInfo struct are valid according to valid usage guidelines
of the spec.

Spun off the usage bit validation into its own function ValidateBufferUsage().
Previous ValidateBufferUpdate() function now calls that function as well as
performing all of the checks for the buffer, range, and offset members of
VkDescriptorBufferInfo struct.

8 years agolayers: Skip UNUSED attachments in pColorAttachments array
Tobin Ehlis [Fri, 1 Jul 2016 19:54:41 +0000 (13:54 -0600)]
layers: Skip UNUSED attachments in pColorAttachments array

8 years agolayers: Use VkRenderPassCreateInfo to check compatibility
Tobin Ehlis [Thu, 30 Jun 2016 15:29:18 +0000 (09:29 -0600)]
layers: Use VkRenderPassCreateInfo to check compatibility

Make the two renderpass compatibility functions only use VkRenderPassCreateInfo.
At the call site, check renderPass handles and if they don't match, proceed to
call the compatibility function with valid VkRenderPassCreateInfo.

Note that saved renderPass handles are valid at the call site, as they are
stored in framebuffer, command buffer, or pipeline state. However it's possible
that renderPass itself was destroyed, so VkRenderPassCreateInfo is pulled from
the framebuffer or pipeline state in those cases.

8 years agolayers: GH690 Save renderPass create info in PSO
Tobin Ehlis [Mon, 27 Jun 2016 18:57:05 +0000 (12:57 -0600)]
layers: GH690 Save renderPass create info in PSO

The renderPass that a pipeline is created with may be destroyed before
the pipeline is used in a draw. In order to make sure that the renderPass
data will live as long as the pipeline, save the VkRenderPassCreateInfo
using a safe_VkRenderPassCreateInfo in the PIPELINE_NODE struct.