Preparing for new developer releases of VulkanSceneGraph, vsgXchange, vsgImGui and vsgExample please test :-) #1383
Replies: 3 comments 7 replies
-
|
Testresult on MacOS 15.3 (Arm64) Most examples work. Some examples cannot work, because the Vulkan support on MacOS is not complete (raytracing, vsgextendstate, vsgmeshshader). Crashing examples: vsgcoordinateframe vsgdynamicstate vsgdynamicviews vsgdynamicwindows Some examples run producing strange output (lighting problem): vsgrendertotexture models/teapot.vsgt |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the details. I just tried: vsgoverlay models/teapot.vsgt
vsgrendertotexture models/teapot.vsgtAnd see the same lighting problem as you on this Kubuntu 24.04+Geforce 1650 system. I am pretty sure this down to the teapot.vsgt being recreated with full VSG lighting support so now requires a vsg::Light in the scene to illuminate the teapot. I'll add this setup of these examples. The cases of VK_ERROR_FEATURE_NOT_PRESENT will just need a try/catch block to more elegantly handling the missing feature support. The vsgcoordinateframe example is curious. As paging works in other examples my guess it's the point cloud used to render the starts. Could you comment out the starfield subgraph and see if that makes a difference. If it does then we can focus on resolve the issue with the starfield. |
Beta Was this translation helpful? Give feedback.
-
|
I have added lighting to the vsgoverlay and vsgrendertotexture examples: |
Beta Was this translation helpful? Give feedback.
-
|
Commenting out the starfield prevents vsgcoordinateframe from crashing, but the output looks incomplete, maybe correct yet. vsgoverlay and vsgrendertotexture look good now. Thanks! |
Beta Was this translation helpful? Give feedback.
-
|
Commenting out the starfield will mean the scene won't have stars :-) Good that it doesn't crash now, that means I just focus on reviewing the starfield subgraph. |
Beta Was this translation helpful? Give feedback.
-
|
I've run vsgcoordinateframe with Vulkan debug layer and got some more clues: $ vsgcoordinateframe -d
info: initial viewpoint : 0. universe_view
VUID-VkGraphicsPipelineCreateInfo-topology-08773(ERROR / SPEC): msgNum: 1895430949 - Validation Error: [ VUID-VkGraphicsPipelineCreateInfo-topology-08773 ] Object 0: handle = 0x27d60e0000000019, type = VK_OBJECT_TYPE_SHADER_MODULE; | MessageID = 0x70f9fb25 | vkCreateGraphicsPipelines(): pCreateInfos[0].pStages[0] SPIR-V (Vertex) PointSize is not written to, but Pipeline topology is set to VK_PRIMITIVE_TOPOLOGY_POINT_LIST. The Vulkan spec states: If the pipeline is being created with a Vertex Execution Model and no TessellationEvaluation or Geometry Execution Model, and the topology member of pInputAssembly is VK_PRIMITIVE_TOPOLOGY_POINT_LIST, and either VK_DYNAMIC_STATE_PRIMITIVE_TOPOLOGY dynamic state is not enabled or dynamicPrimitiveTopologyUnrestricted is VK_FALSE, a PointSize decorated variable must be written to if maintenance5 is not enabled (https://vulkan.lunarg.com/doc/view/1.3.290.0/linux/1.3-extensions/vkspec.html#VUID-VkGraphicsPipelineCreateInfo-topology-08773)
Objects: 1
[0] 0x27d60e0000000019, type: 15, name: NULL
VUID-VkGraphicsPipelineCreateInfo-Input-07904(ERROR / SPEC): msgNum: -862178626 - Validation Error: [ VUID-VkGraphicsPipelineCreateInfo-Input-07904 ] Object 0: handle = 0x27d60e0000000019, type = VK_OBJECT_TYPE_SHADER_MODULE; | MessageID = 0xcc9c32be | vkCreateGraphicsPipelines(): pCreateInfos[0].pVertexInputState->pVertexAttributeDescriptions does not have a Location 1 but vertex shader has an input variable at that Location. The Vulkan spec states: If the pipeline is being created with vertex input state and pVertexInputState is not dynamic, then all variables with the Input storage class decorated with Location in the Vertex Execution Model OpEntryPoint must contain a location in VkVertexInputAttributeDescription::location (https://vulkan.lunarg.com/doc/view/1.3.290.0/linux/1.3-extensions/vkspec.html#VUID-VkGraphicsPipelineCreateInfo-Input-07904)
Objects: 1
[0] 0x27d60e0000000019, type: 15, name: NULL
VUID-VkShaderModuleCreateInfo-pCode-08737(ERROR / SPEC): msgNum: -1520283006 - Validation Error: [ VUID-VkShaderModuleCreateInfo-pCode-08737 ] | MessageID = 0xa5625282 | vkCreateShaderModule(): pCreateInfo->pCode (spirv-val produced an error): |
Beta Was this translation helpful? Give feedback.
-
|
I have checked in improvements to the the handling of point sprites in the built-in ShaderSets and vsgcoordianteframe. @rainergericke Could you update to VulkanSceneGraph and vsgExamples master and let me know if these changes resolve the crash in vsgcoordinateframe? |
Beta Was this translation helpful? Give feedback.
-
|
vsgcoordinateframe works now without crash but with stars, great! 
|
Beta Was this translation helpful? Give feedback.
-
|
Hi, I have tested the current master branches on an rpi4 running Ubuntu 24.04 These are the main results, and therefore much more improved then last time when I tested for the 1.1.8 release. vsgheadless models/openstreetmap.vsgt vsgrendertotexturearray models/teapot.vsgt vsgvalidate vsgmeshshader vsgtextureprojection textures/VSGlogo.png vsgraytracing vsgclip models/openstreetmap.vsgt vsgclip models/teapot.vsgt Of course, there is a lot of possible combinations for the cmd line options but I tested the examples without any other then specifying a model. If there is anything particular to test, please let me know. |
Beta Was this translation helpful? Give feedback.
-
This example and command line combination works fine for me so it could be an issue that the required Vulkan features aren't supported. Perhaps just better reporting of the exception would be best. Is there any other information provided?
This is what I see when I run vsgvalidate: info: ---- Test 1 : Sharing Sampler between ImageInfo/InmageView/Image instances ----
info: test 1 passed
info: End of test 1
info: ---- Test 2 : Automatic disabling of mipmapping when using compressed image formats. ----
info: test2 passed
info: End of test 2
info: ---- Test 3 : DescriptorPool reservation in vsg::Context ----
info: Test 3, part 1
info: Test 3, part 2
info: End of test 3
Works fine on my Kubuntu 24.04 + Geforce 1650 system, so the command line usage looks fine. However, I can't find any instances of "Failed to allocate device memory for" in any of the VSG projects so I presume this error must being coming from the Vulkan driver.
I don't see any errors but also don't see the expected effect. The teapot.vsgt and openstreetmap.vsgt use the build in ShaderSets for shaders so I suspect the scene graph setup isn't compatible with the tricks that the vsgclip example uses. The vsgclip example does work with the older shaders used in the lz.vsgt: vsgclip models/lz.vsgt |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hi All,
This week I merged changes to how the VulkanSceneGraph manages color spaces with changes to the default settings for swapchain/color frame buffer and VkFormat of loaded images that resolve color space kludges that have been around since the inception of the VSG, or even the OSG if we want to a long way :-)
I wrote up the changes in the dedicated thread #1379, so head over there for more details about the work.
I'm now doing a call for testing before the next round of developers releases as there changes to defaults they could impact how end users applications look and require developer to tweak their applications and data to undo the implicit kludges that would have been made to accommodate the kludges that used exist in the VSG.
If there are things we need to tweak or document to help users accommodate these changes I'd like to hear about them before release so we can figure out what to change/document so that 1.1.10 can be the best we can make it.
I'd like folks to update to VulkanSceneGraph, vsgXchange, vsgImGui and vsgExamples masters and let us know how you get on with the building/running examples and testing against your own application. Let us know on this thread of how you get on, be it success or failure so we can get a feel for how things are converging to dev release quality.
Thanks,
Robert.
Beta Was this translation helpful? Give feedback.
All reactions