| Age | Commit message (Collapse) | Author |
|
The 1.3 update forgot to rename
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_INTEGER_DOT_PRODUCT_FEATURES_KHR to
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_INTEGER_DOT_PRODUCT_FEATURES as this
is the non-aliased name. Without it, the fix wouldn't be applied.
|
|
|
|
|
|
|
|
Only changes are making the autogen print all required structs, and
moving from using the EXT version of tool properties to the core version.
|
|
- Update known-good
- Generate source
|
|
- Update known-good
- Generate source
|
|
If the implementation reports that it supports this feature, it should
fill out VkFormatProperties3KHR properly.
|
|
- Update known-good
- Generate source
|
|
- Update known-good
- Generate source
|
|
|
|
Updated MoltenVK to 1.1.6 release
|
|
- Update known-good
- Generate source
|
|
The structure VkPhysicalDeviceShaderIntegerDotProdcutFeaturesKHR contains only a single
feature boolean flag. However, before the final version was published to the public, it
had many boolean flag values. Pre-release drivers often contain support for unpublished
extensions, such as VK_KHR_shader_integer_dot_product. Because the final version was
much smaller than the pre-release version, several drivers try to write to members that
do not exist in the final version, which cause out of bounds writes to the features struct
that vulkaninfo passes into the driver.
By increasing the size of the features struct allocation manually, we can prevent potential
crashes from OOBs writes from drivers.
|
|
- Update known-good
- Generate source (no change)
|
|
- Update known-good
- Generate source
|
|
- Update known-good
- Generate source
|
|
- Update known-good
- Generate source
|
|
- Update known-good
- Generate source
|
|
Currently some of the generators use a fixed child index to locate
extension enums, which will break if comment blocks are added near the
top of an extension definition. This change makes the enum lookup more
robust by searching for the expected enum name instead of using a
hard-coded offset.
|
|
Bring forward the update_deps script in this repo to match VVL
|
|
f-strings are a python feature which allow more natural formatting of text, making
the code easier to read and work with.
|
|
64 bit flags use the bitvalues field instead of the requires field in the XML
to specify the flagbits name. Thus the autogen needed to account for that.
In addition, if a flag was promoted to core, the autogen didn't use the core
version, this commit also fixes that.
|
|
- Update known-good
- Generate source
|
|
- Update known-good
- Generate source
|
|
When the working group changes the name of a constant:
VK_KHR_MAINTENANCE1_SPEC_VERSION ->
VK_KHR_MAINTENANCE_1_SPEC_VERSION
they will still #define the original name to maintain backward
compatibility with existing code that uses it:
#define VK_KHR_MAINTENANCE_1_SPEC_VERSION 2
#define VK_KHR_MAINTENANCE_1_EXTENSION_NAME "VK_KHR_maintenance1"
#define VK_KHR_MAINTENANCE1_SPEC_VERSION VK_KHR_MAINTENANCE_1_SPEC_VERSION
#define VK_KHR_MAINTENANCE1_EXTENSION_NAME VK_KHR_MAINTENANCE_1_EXTENSION_NAME
In this case, we need to ignore the non-numeric _SPEC_VERSION and
unquoted _EXTENSION_NAME constants because they don't define a separate
extension.
|
|
- Update known-good
- Generate source
|
|
|
|
- Update known-good
- Generate source
|
|
- Update known-good
- Generate source (no change)
|
|
- Update known-good
- Generate source
|
|
KhronosGroup/Vulkan-ValidationLayers switched to this method a while ago
and this syncs the KhronosGroup/Vulkan-Tools version of
helper_file_generator.py to match.
|
|
- Update known-good
- Generate source
|
|
- Update known-good
- Generate source (no change)
Change-Id: I529e0a7175f18f50ae717c4caae27ef066b80839
|
|
- Update known-good
- Re-enable code generation for VK_HUAWEI_subpass_shading
- Generate source
Change-Id: Ibb8d8898c9cdbb655c8162e76b68fa465fe8f912
|
|
- Update known-good
- Generate source
Change-Id: I6f8219e31038b0efe1e141147292d041655dd640
|
|
|
|
- Update known-good
- Generate source
Change-Id: Idd87f69c520cd4ef34a1b9d205f79f7c1c7d9e05
|
|
- Update known-good
- Add support for printing `int64_t` in `scripts/vulkaninfo_generator.py`
- Disable codegen for VK_HUAWEI_subpass_shading
KhronosGroup/Vulkan-Docs#1564
- Generate source
Change-Id: I4417e99ed37da510bfaf94d676e22ac3fb088d5d
|
|
vkconfig prints arrays as it finds them, even if there is only a "None"
element in it. This commit prevents this element from being printed,
removing confusing results where there is a `count = 1` but contains "None".
Change-Id: I1ef8c63458728a9fd3dc531dc9983f309c851034
|
|
- Update known-good
- Generate source (no change)
Change-Id: Ibfdf958ca32779f7635c324af3ed4146186b27c9
|
|
VkSurfaceCapabilities has the width & height way out to the right
while the rest of the structure doesn't. This makes it look inconsistent.
This PR also changes the layout of the flag and flagbit printers to return
earlier if possible.
Change-Id: I1b478838b3c95e33354d3daf13040e891454324e
|
|
- Update known-good
- Generate source
Change-Id: I41d51ecab3257530ad733fe977f92d9c482543e6
|
|
- Update known-good
- Generate source
Change-Id: Ib8666fe063852297f746dbea1b8b34346000b13d
|
|
* Simplify AppQueuefamilyProperties
* Create and store AppQueueFamilyProperties in AppGpu instead of creating them
on the fly
* Make all uses of 2 struct's use the original struct where appropriate
* This means the printers will use props.limits instead of props2.limits
* Remove unused functions
* Remove unused variables
* Add sparseBinding to enabled features (previously there were none)
* make_vulkan_version function introduced
Change-Id: I872dd060dbf709b5b39610775fb88718ff8b19a5
|
|
When setting device format properties, don't set color attachment-
related bits for DS formats and don't set DS attachment bit for
color formats.
Fixes #445
|
|
- Update known-good
- Generate source
Change-Id: Iecb34b6221526dc5c6153c736fce894488989066
|
|
Updated to the latest MoltenVK tag.
|
|
- Update known-good
- Generate source
Change-Id: I8f49422f152ad353658cb753f24564b76591d415
|
|
distutils.dir_util.mkpath will not re-create a directory if it has been
removed with shutil.rmtree because it caches the filesystem under the
hood. This was causing the clone retry to fail on the second iteration
every time.
Switch to using os.makedirs for equivalent functionality without the
cache.
|