Remove support for VkSemaphore

[SS-JIA]

Summary:

Changes

Revert some changes to the D78360038 / #12527 stack which enabled support for submitting command buffers with an associated semaphore.

Context

The original intent was to allow command buffers to be correctly ordered when submitting multiple command buffers for model inference. Previously it was thought that the Vulkan API would not be aware of the dependency between two separate command buffer submissions, so a semaphore would be needed to ensure correct execution order between them.

However, I noticed the following validation layer error on Mac:

Validation 0 vkQueueSubmit(): pSubmits[0].pSignalSemaphores[0] (VkSemaphore 0x10f000000010f) is being signaled by VkQueue 0x1181082a8, but it was previously signaled by VkQueue 0x1181082a8 and has not since been waited on.
The Vulkan spec states: Each binary semaphore element of the pSignalSemaphores member of any element of pSubmits must be unsignaled when the semaphore signal operation it defines is executed on the device (https://vulkan.lunarg.com/doc/view/1.4.313.0/mac/antora/spec/latest/chapters/cmdbuffers.html#VUID-vkQueueSubmit-pSignalSemaphores-00067)

The reason this happens is because we store the VkSemaphore together with the VkCommandBuffer, and use the same VkSemaphore in the submit info every time the command buffer is submitted. However, there's no mechanism to reset the VkSemaphore to "unsignaled" once it's been signaled.

Therefore, as-is the VkSemaphores do not serve any purpose after the first inference.

The correct approach if we were to use semaphores is to create a new one with every submission, and not have it be attached to a specific command buffer

However, after some deeper research I found that VkSemaphore is not actually needed to ensure correct execution order between command buffers; the command pipeline barriers that we already insert should be sufficient. My primary source is this stackoverflow question which references the Vulkan API spec in its answer.

Therefore, remove the VkSemaphore machinery since it's not required.

Differential Revision: D79468286

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13070

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure, 1 Unrelated Failure

As of commit 1a4b217 with merge base 1d80837 ():

NEW FAILURE - The following job has failed:

• Build Presets / apple (ios-simulator) / build (gh)
  RuntimeError: Command bash /Users/runner/work/_temp/exec_script failed with exit code 65

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• pull / test-setup-linux-gcc / linux-job (gh) (detected as infra flaky with no log or failing log classifier)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[facebook-github-bot]

This pull request was exported from Phabricator. Differential Revision: D79468286

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.