Batch skinned meshes on platforms where storage buffers are available. #16599
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This commit makes skinned meshes batchable on platforms other than WebGL 2. On such platforms, it replaces the two uniform buffers used for joint matrices with a pair of storage buffers containing all matrices for all skinned meshes concatenated together. The indices into the buffer are stored in the mesh uniform and mesh input uniform. The GPU mesh preprocessing step copies the indices in if that step is enabled. On the `many_foxes` demo, I observed a frame time decrease from 15.470ms to 11.935ms. This is the result of reducing the `submit_graph_commands` time from an average of 5.45ms to 0.489ms, an 11x speedup in that portion of rendering.
pcwalton
added
C-Performance
JMS55
reviewed
Dec 3, 2024
|
I'm not sure if this is working? I ran many_foxes through renderdoc, and I still see hundreds of vkCmdDraws, each with two vkCmdBindDescriptorSets before it. One for StandardMaterial, one for skinned_mesh_bind_group. I would've expected only one vkCmdBindDescriptorSets for skinned_mesh_bind_group, and then since the meshes are the same we don't have to rebind StandardMaterial either, and then you would be able to collapse all the draws down to one draw with instance_count=N. Am I missing why this doesn't seem to reduce commands? |
|
Hmm, it works for me on my G14 Windows 11, with
|
|
Windows, Intel 155h. |
|
It works on my M2 mac mini too. Hmm. |
kristoff3r
approved these changes
Dec 6, 2024
pcwalton
added
S-Ready-For-Final-Review
|
Running into a panic with this on the backtrace |
|
Following up from Discord, I tested this branch on an Intel i5-1240P with the following Renderdoc settings: I believe this is sufficient evidence that the feature works on Intel iGPUs, but I'm still learning this tooling so take with a grain of salt. |
|
Looks like motion vectors are also broken with this (tested by adding the backtrace |
|
I fixed the motion vectors issue, but was unable to reproduce the |
I can't reproduce the Motion vectors no longer crash but they also don't appear to be working for skinned meshes now. Motion vector prepass texture once everything is rendered (left main, right this PR):
|
|
This seems fine to merge, but I'm doing a manual example run here to check for regressions. Ping me once that's done / tomorrow if I forget. |
|
Ok, the skinned mesh motion vector thing should be fixed now. |
Elabajaba
approved these changes
Dec 10, 2024
|
I believe the CI failure is #15981. |
|
Example run is green. |
BD103
pushed a commit
to BD103/bevy
that referenced
this pull request
Dec 10, 2024
bevyengine#16599) This commit makes skinned meshes batchable on platforms other than WebGL 2. On supported platforms, it replaces the two uniform buffers used for joint matrices with a pair of storage buffers containing all matrices for all skinned meshes packed together. The indices into the buffer are stored in the mesh uniform and mesh input uniform. The GPU mesh preprocessing step copies the indices in if that step is enabled. On the `many_foxes` demo, I observed a frame time decrease from 15.470ms to 11.935ms. This is the result of reducing the `submit_graph_commands` time from an average of 5.45ms to 0.489ms, an 11x speedup in that portion of rendering.  This is what the profile looks like for `many_foxes` after these changes.  --------- Co-authored-by: François Mockers <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This commit makes skinned meshes batchable on platforms other than WebGL 2. On supported platforms, it replaces the two uniform buffers used for joint matrices with a pair of storage buffers containing all matrices for all skinned meshes packed together. The indices into the buffer are stored in the mesh uniform and mesh input uniform. The GPU mesh preprocessing step copies the indices in if that step is enabled.
On the
many_foxesdemo, I observed a frame time decrease from 15.470ms to 11.935ms. This is the result of reducing thesubmit_graph_commandstime from an average of 5.45ms to 0.489ms, an 11x speedup in that portion of rendering.This is what the profile looks like for
many_foxesafter these changes.