VkPipelineMultisampleStateCreateInfo::rasterizationSamples relation to VkRenderPass of graphics pipeline

[procedural]

I feel like it's not mentioned anywhere explicitly that vkCreateGraphicsPipelines()'s VkGraphicsPipelineCreateInfo's VkPipelineMultisampleStateCreateInfo::rasterizationSamples value dictates what VkSampleCountFlagBits value must be specified in all attachments of VkRenderPass passed to VkGraphicsPipelineCreateInfo, all of its color and depth/stencil attachments

Or, in other words, that all color and depth/stencil attachments of VkGraphicsPipelineCreateInfo::renderPass must be of the same VkSampleCountFlagBits value as VkPipelineMultisampleStateCreateInfo::rasterizationSamples

[procedural]

#543

[procedural]

https://raw.githubusercontent.com/KhronosGroup/Vulkan-Docs/b0b699e446e3eab9bc0d6b192599871eecf25e1e/chapters/commonvalidity/draw_common.adoc

• [[VUID-{refpage}-None-08644]] If a graphics pipeline is bound, the <<dynamic-state-current-value, current value>> of pname:rasterizerDiscardEnable is ename:VK_FALSE, then the <<dynamic-state-current-value, current value>> of pname:rasterizationSamples must: be the same as the current color and/or depth/stencil attachments

https://raw.githubusercontent.com/KhronosGroup/Vulkan-Docs/b0b699e446e3eab9bc0d6b192599871eecf25e1e/scripts/reflow-tests/expect-nested-lists-in-vu-novuid.adoc

• [[VUID-{refpage}-multisampledRenderToSingleSampled-07284]] If rasterization is not disabled in the bound graphics pipeline, then pname:rasterizationSamples for the currently bound graphics pipeline must: be the same as the current subpass color and/or depth/stencil attachments

[Tobski]

Thanks for the feedback!

So I did some digging, this particular condition is a pain because it interacts with several extensions and features, but if I understand your request correctly, a VU does exist, which reads (bold mine):

If the pipeline requires fragment output interface state, and none of the VK_AMD_mixed_attachment_samples extension, the VK_NV_framebuffer_mixed_samples extension, or the multisampledRenderToSingleSampled feature are enabled, rasterizationSamples is not dynamic, and if subpass uses color and/or depth/stencil attachments, then the rasterizationSamples member of pMultisampleState must be the same as the sample count for those subpass attachments

The wording on this is a little confusing, and it seems that at some point we accidentally ifdef'd this out of the core spec, so I have an internal MR to try to bring it back into the core and fix up the wording.

Does that answer your question? Or are you also looking for something that isn't a VU?

[procedural]

Does that answer your question?

Yep, thank you!