aboutsummaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2015-10-30bug 14567: What is VK_QUEUE_EXTENDED_BIT? (WIP)Chia-I Wu
Removed. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14567
2015-10-30bug 14876: Can timestamps be written inside a renderpass? (WIP)Chia-I Wu
Timestamps are written to a query pool. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14876
2015-10-30bug 14362: VK_NULL_HANDLE does not work in C (WIP)Chia-I Wu
Drop type safety on 32-bit. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14362
2015-10-30v185: bump API minor versionChia-I Wu
2015-10-30bug 14864: Remove redundant VkPipelineShaderStageCreateInfo.stage field? (WIP)Chia-I Wu
Removed. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14864
2015-10-30MR 461: Documented updated vkQueueSubmit and remove wait + signal semaphore ↵Chia-I Wu
operations on queues. (WIP) https://gitlab.khronos.org/vulkan/vulkan/merge_requests/461
2015-10-30support and use VkSubmitInfo semaphore signaling and waitingChia-I Wu
2015-10-30bug 14868: Replace VkTimestampType with VkPipelineStageFlagBits (WIP)Chia-I Wu
https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14868
2015-10-30v184: bug 14719: vkCmdCopyBufferToImage alignment requirements (WIP)Chia-I Wu
Update vulkaninfo to print new limits. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14719
2015-10-30bug 14822: timestampFrequency underspecified (WIP)Chia-I Wu
Replaced by timestampPeriod, number of nanoseconds per tick. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14822
2015-10-30bug 14725: Remove or replace VK_SPARSE_MEMORY_BIND_REPLICATE_64KIB_BLOCK_BIT ↵Chia-I Wu
(WIP) Renamed. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14725
2015-10-30bug 14462: endianness of code for shader modules (WIP)Chia-I Wu
Change the type of pCode to uint32_t*. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14462
2015-10-30bug 14855: vkGetPipelineCacheSize is unique (consistency issue) (WIP)Chia-I Wu
Update vkGetPipelineCacheData() to replace vkGetPipelineCacheSize(). https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14855
2015-10-30bug 14861: rasterization review issues (WIP)Chia-I Wu
Update vulkaninfo to print features->strictLines. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14861
2015-10-30bug 14365: Rename slopeScaledDepthBias to depthBiasSlopeScaled (WIP)Chia-I Wu
s/slopeScaledDepthBias/depthBiasSlopeFactor/g s/depthBias\b/depthBiasConstantFactor/g https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14365
2015-10-30bug 14844: Query Pool issues (WIP)Chia-I Wu
We did not and still do not honor the stride nor flags. https://cvs.khronos.org/bugzilla/show_bug.cgi?id=14844
2015-10-30vulkan.py: use generated "core" APIChia-I Wu
The parsed vkCmdSetBlendConstants prototype makes a difference to vktrace. But vktrace should handle it just fine.
2015-10-30vulkan.py: fix vulkan.h parsingChia-I Wu
Ignore PFN_vkVoidFunction typedef. Fix layer dispatch table output.
2015-10-29layers: Fix MemTracker uninitialized variablesMark Lobodzinski
Caused issues in SDK tests on Windows
2015-10-29demos: Properly destroy command pool in cube demoMark Lobodzinski
2015-10-29layers: Add DestroyImage to image layer's device GPA tableMark Lobodzinski
2015-10-29loader: Update Linux and Windows ICD/layer discover documentationjon
2015-10-29loader: Update linux documentation on where ICD lives.jon
2015-10-29loader: Move utility function so it is found on both Linux and Windowsjon
2015-10-29loader: set DEFAULT_VK_{DRIVER,LAYER}S_PATH to "" on LinuxDaniel Dadap
When loader_get_fullpath() gets an empty string as the 'dir' parameter, it immediately writes out the original, unexpanded filename. Setting the default drivers and layers paths to an empty string on Linux will allow the unmodified filenames to be passed to dlopen(3) without having to maintain a separate code path for Windows. Now that there is no longer a default path for drivers or layers on Linux, ${CMAKE_INSTALL_LIBDIR} no longer needs to be baked into the loader binary. Note that any ICDs or layers that are currently getting installed to the previous default paths (e.g. /usr/lib/vulkan/icd or /usr/lib/vulkanlayer) will not be discoverable when specified as plain filenames, unless these directories happen to be configured in the loader's search path on the system in question.
2015-10-29loader: add helper for combining pathsDaniel Dadap
Add a helper function to combine an arbitrary number of path elements, separating them with the platform-specific directory separator, and use it in places where path combining is currently achieved through direct string manipulation. As an additional cleanup, update loader_get_fullpath() to duplicate the search path string internally, rather than relying on the caller to do so. Since loader_get_fullpath() and loader_expand_path() no longer modify input strings, and loader_expand_path() no longer requires that the caller append the DIRECTORY_SYMBOL ahead of time, acrobatics to duplicate strings to prevent modification and/or to add directory separators are no longer necessary. This change will result a behavior change on Windows: the existing logic for loading an ICD that is specified as a bare filename, and not as an absolute or relative path, is to search for a driver with the given ICD name in the DEFAULT_VK_DRIVERS_PATH search path on Linux, and to use the given ICD name as-is on Windows. This could lead to DLL injection attack on Windows. Merge the Linux/Windows behavior to limit the search for ICD filenames to the default search path on Windows as well, which is also more consistent with the behavior described in the loader spec. Conflicts: loader/loader.c
2015-10-28demos: fclose in cube and triMike Stroyan
Balance every fopen with fclose.
2015-10-28demos: windows fixes for cube and tri resizeMike Stroyan
Defer resize until after first prepare. Record new window size from WM_SIZE. Leave demo->swapchain.handle valid to pass to CreateSwapchainKHR.
2015-10-28Layers: MemTracker check vkCmdCopyQueryPoolResultsMike Stroyan
MemTracker check of buffer memory and usage flags in vkCmdCopyQueryPoolResults
2015-10-28demos: add occlusion features to vulkaninfoMike Stroyan
Add reporting of VkPhysicalDeviceFeatures occlusionQueryNonConservative to vulkaninfo.
2015-10-28use conditional compilation based on _WIN32 instead of WIN32 and __linux__Johannes van Waveren
2015-10-28demos/cube: Provide a non-null VertexInputStateCreateInfoJason Ekstrand
In the spec for VkGraphicsPipelineCreateInfo, some structs are explicitly listed as being able to be NULL if they are not used. VertexInputStateCreateInfo is not one of them.
2015-10-28demos/cube: Explicitly specify std140 layout on UBOsJason Ekstrand
Without this, glslang won't actually provide the offsets and strides required by SPIR-V.
2015-10-28layers: Validate buffer view usage flagsMark Lobodzinski
2015-10-28layers: Validate number of layers in subresource.Jeremy Hayes
Reviewed by Tobin.
2015-10-28tests: Added CopyDescriptorUpdateErrors test to verify 3 errors during ↵Tobin Ehlis
descriptorset copy updates
2015-10-28layers: Add DrawState validation for copy descriptor updatesTobin Ehlis
2015-10-28layers: Simplify descriptor validation in DrawStateTobin Ehlis
Many descriptor validation helper functions were returning skipCall based on the potential for a bad update structure type. However, at the beginning of update validation, we validate that the update type is ok. Therefore, the other helper functions don't need to continually repeat this check, and can be simplified to return relevant values directly.
2015-10-28layers: Improve DrawState write descriptor updateTobin Ehlis
Validate that stageFlags are the same for a single write update. Validate that all sampler updates from a single write update are either immutable or non-immutable. Refactor update contents check to have a loop per switch case instead of switching for each loop iteration. Added two new validation errors: DRAWSTATE_DESCRIPTOR_STAGEFLAGS_MISMATCH DRAWSTATE_INCONSISTENT_IMMUTABLE_SAMPLER_UPDATE
2015-10-28layers: Remove DrawState g_lastBoundDescriptorSet variable - it's no longer usedTobin Ehlis
2015-10-27demos: Clean up tri depthCody Northrop
We animated tri to make it clear it was rendering correctly based on ISV feedback. It looks better, but still questionable. This change makes the depth cutoff a straight line by giving the top two vertices the same depth.
2015-10-27draw_state: Fix enum in validate_draw_state_flagsCody Northrop
DRAWSTATE_STENCIL_NOT_BOUND -> DRAWSTATE_SCISSOR_NOT_BOUND
2015-10-27windows: Fix vector resizing in draw_stateCody Northrop
2015-10-27demos: print new device limits in vulkaninfoTony Barbour
2015-10-27misc: Remove dynamic array definititionsCourtney Goeltzenleuchter
Windows doesn't like dynamicly sized arrays. Replace with new'd array of objects or object pointers instead.
2015-10-27misc: Add intrin.h to support ffs utilityCourtney Goeltzenleuchter
2015-10-27misc: Windows does not like named structure initializationCourtney Goeltzenleuchter
2015-10-27Bug 14621: Change vkGet* functions to not return VkResultCourtney Goeltzenleuchter
Missed one function in prior patch.
2015-10-27tri/cube: Adjust demo_resize() to API changes.Ian Elliott
2015-10-27tri/cube: Add WM_SIZE to Win32 event handler to resize a window.Ian Elliott