In the example Inputattachments,subpass dependencies didn't take care of the layout transition of depth image.

[CretaceousConstructor]

I ran this example with Vulkan configurator's Synchronization utility on,and got errors:

SYNC-HAZARD-WRITE_AFTER_WRITE(ERROR / SPEC): msgNum: -33950239 - Validation Error: [ SYNC-HAZARD-WRITE_AFTER_WRITE ] Object 0: handle = 0x2bb0785d7b0, type = VK_OBJECT_TYPE_RENDER_PASS; | MessageID = 0xfdf9f5e1 | vkCmdBeginRenderPass: Hazard WRITE_AFTER_WRITE vs. layout transition in subpass 0 for attachment 2 aspect depth during load with loadOp VK_ATTACHMENT_LOAD_OP_CLEAR.

    Objects: 1

        [0] 0x2bb0785d7b0, type: 18, name: NULL

(more errors in the attached file) inputattachmentsWithExplicit.txt

the layout transiton of depth image from VK_IMAGE_LAYOUT_UNDEFINED(VK_SUBPASS_EXTERNAL) to VK_IMAGE_LAYOUT_DEPTH_ATTACHMENT_OPTIMAL (supass 0) has to happen before the depth image is used as output-depth values recipient(or more concisely,as render target ).my solution is here:

dependencies[0].srcSubpass = VK_SUBPASS_EXTERNAL;
dependencies[0].dstSubpass = 0;
dependencies[0].srcStageMask =VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
dependencies[0].dstStageMask = 
   VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | 
   VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT;
dependencies[0].srcAccessMask = VK_ACCESS_MEMORY_READ_BIT;
dependencies[0].dstAccessMask   =   
    VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | 
    VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | 
    VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
dependencies[0].dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT;

This make sure that the layout transition will happen before VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT stage when depth values need to be writen to depth image.But I'm not sure if this is the optimal method.

[kyamant]

This has been a problem with many examples (even triangle).

I applied @CretaceousConstructor 's solution to void VulkanExampleBase::setupRenderPass() and the validation error went away.

[SaschaWillems]

I'm aware of the synchronization issues. I'm in the process of completely reworking synchronization in my samples and will tackle issues as this in that process.

[kyamant]

Seems like this went away with 1.2.182.0

[CretaceousConstructor]

Seems like this went away with 1.2.182.0

Is that because some new implicit transitions have been added ?

[kyamant]

Seems like this went away with 1.2.182.0

Is that because some new implicit transitions have been added ?

Not sure about the underlying mechanics; just an observation.

[kyamant]

Curiously enough, the validation messages came back with 1.2.189.2 ?