| Age | Commit message (Collapse) | Author |
|
KHR_surface, KHR_display extension functions were not exposed properly.
Specifically, added LOOKUP_GIPA entries for:
* GetPhysicalDeviceDisplayPropertiesKHR
* GetDisplayModePropertiesKHR
* CreateDisplayPlaneSurfaceKHR
* GetPhysicalDeviceDisplayPlanePropertiesKHR
* GetDisplayPlaneSupportedDisplaysKHR
* CreateDisplayModeKHR
* GetDisplayPlaneCapabilitiesKHR
* DestroySurfaceKHR
Added missing function declaration DestroySurfaceKHR in loader_icd
struct.
Checking if KHR_surface extension is enabled before proceeding in the
terminator_CreateDisplayPlaneSurfaceKHR definition.
Change-Id: I0599efddedc5064859ae1f4e52bdbbf8788d5c12
|
|
invoked
|
|
VkObjects should be cast to uint64_t by using 'reinterpret_cast<uint64_t &>'
Change-Id: I403efdf389dad61c373df40c5fa1fbed7cf96abd
|
|
Correctly update fence tracking for vkQueueBindSparse() call.
|
|
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
These are actually stupidly expensive to construct -- with most of the
cost going to the contained std::locale.
Added up to about 6% of profile on my machine in `smoke -s` with
VK_LAYER_LUNARG_core_validation enabled.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Make sure vkSetEvent() is not called on an event that's in use
by and in-flight cmd buffer. Such use violates queue forward progress.
|
|
|
|
Change-Id: I20b0d64c90e6da940f45b31abf0330b2a415655d
|
|
|
|
|
|
This was completely redundant to the (correct) CB-scoped renderpass
tracking, and only served to break things when different threads
simultaneously recorded command buffers.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
VkDescriptorPoolSize array can contain more than one of same VkDescriptorType
|
|
This reverts commit 541bb5afc50581e7c765fc5c05400540369f0044.
Reverting this as the original issue is still present and being worked on
so restoring original state until correct fix is complete.
|
|
No need to have 3 semaphore states. Killed SemaphoreState struct and
updated all the code to just use the "signaled" bool.
Also minor clean-up of queueMap destroy code.
|
|
Killed some mem_tracker functions that were doing redundant work.
Removed some calls to checkCBCompleted that were superceded by
validateCommandBuffersNotInUse() calls.
Clarified some "my_data" vars as "dev_data."
|
|
Pull physical device memory properties into layer_data and make sure it's
correctly populated at vkCreateDevice() time.
Clarify a few ambiguous "my_data" variables as "dev" or "instance".
Rename physical device properties to be lower case. This is the start of
a slow move to unify code on common standard with lowercase variable names.
|
|
|
|
Previously this function assumed there were as many d/s attachments as
color attachments, which is invalid. Besides not making sense, we'd end
up dereferencing memory beyond that pointed to by
VkSubpassDescription::pDepthStencilAttachment if we had multiple color
attachments.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
|
|
Change-Id: If6b6b24fdf02b960c83148bd002114941eb1adc6
|
|
Previously this used '1'.
Also clean up the enum, and description of device_limits checks.
Found by inspection while reviewing #290.
Change-Id: I710d9b8fad0ca027eb3de67b6c2c70b5628af5a5
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Rename utils lib to VkLayer_utils to use same namespace as
the layers.
The previous Windows CMake code attempted to build both
dynamic and static utils libs, succeeding with only the static.
Change the CMake code to only attempt to build the static lib.
|
|
Use xml.etree for xml parsing for greater portability.
Change-Id: Ic9b3b487e9b363a89d134d048db9fa79fbd66f6f
|
|
Sort python dictionaries that are used in vk-layer-generate.py
which otherwise produce source code with unpredictable order.
Change-Id: I9b26fa61325098998ff510063a649414807d1378
|
|
A unicode header snuck into the front of vk-layer-generate.py.
It does not belong there.
Change-Id: Iad04156d0ed9a0e0c880f88a259b7f5f76406b1d
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
tested: windows/mac
|
|
tested: Mac X
|
|
|
|
|
|
vkSetEvent may occur asynchronously from cmd buffer execution
and the results of vkSetEvent should be immediately visible
across all queues of a device.
This fix will update any event that a queue has seen so that
its VK_PIPELINE_STAGE_HOST_BIT is set at vkSetEvent() time.
|
|
Change-Id: I2ff3b6dab4209b152f487649db011cdfa7c8975b
|
|
|
|
Fix error reported by vkvalidatelayerdoc test.
Change-Id: I7e1bb4e5cb8c113202d479ef95cc0d37334040b1
|
|
As requested by Jon -- instead of appending the expanded layers, insert
them in place of the meta layer.
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
expand_... / unexpand_... scribbled on both the CreateInfo struct and
the list of layer strings, and then unscribbled them on the way back
out. This is a lousy thing to do, and just blows up if the memory isn't
writable (which it needn't be, given the API takes ptrs to const).
Instead, copy the *CreateInfo into a shadow struct on the stack, and be
careful in expand_layer_names never to scribble on the caller's layer
names array.
V2: slight tweak (missed initializer)
Signed-off-by: Chris Forbes <chrisforbes@google.com>
|
|
Change-Id: Ic9b2b2a2ea14bb087d74f6c0b697449b141899a4
|
|
|
|
Given new spec language coming out in early April (clarifying various aspects
of acquiring presentable images), a change was needed to the WSI-swapchain
validation layer. The new validate-layer test will work for "BlitModel"
drivers that advertise a minImageCount of 1. Failing the test now results in
an error instead of a performance warning.
|
|
Change-Id: I9a03c6ea16abe9e362ef81ee87ac217b8b3c6c92
|
|
|
|
Fixes segv in CTS multithreaded command pool test
Change-Id: I80da12fd2bb6856e3c6824f17dd92f1597121a54
|