Accelerate the Merge vertex post processing step

[motazmuhammad]

The function int JoinVerticesProcess::ProcessMesh( aiMesh* pMesh, unsigned int meshIndex)
can be accelerated by using an std::unordered_map instead of binary search to create the unique vertices.
To do that the following was done.

1. Define a hash for a vertex
2. Define an equal_to for a vertex
3. loop over all the vertices if the vertex is already taken use the map to get its index, else add it.

Effectively, this should reduce the complexity from O(nlog(n)) to O(n). Therefore, accelerating the postprocessing step

[kimkulling]

Looks fine, I will merge it,

[kimkulling]

Nice approach, hopefully, the memory consumption will not explode for big models. But this will be a different story.

[kimkulling]

Merged, thanks a lot for your contribution.

[motazmuhammad]

Thank you very much for accepting my contribution.

…

On Sat, May 14, 2022 at 10:54 AM Kim Kulling ***@***.***> wrote: Merged, thanks a lot for your contribution. — Reply to this email directly, view it on GitHub <https://urldefense.com/v3/__https://github.com/assimp/assimp/pull/4527*issuecomment-1126682587__;Iw!!Nmw4Hv0!wEcF-fwE3l-hbUspBe7zKxHf8gFIBFIL4Ge7Hk08GjHTU1vRkfYexdvepWjEsPTCyWTSLgoVzJ4tdDKn6b5I-Uk8UEpghU8tng$>, or unsubscribe <https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AEXGJM5L5EIWGMGIBOSYPLTVJ5Z5PANCNFSM5V4HYXMQ__;!!Nmw4Hv0!wEcF-fwE3l-hbUspBe7zKxHf8gFIBFIL4Ge7Hk08GjHTU1vRkfYexdvepWjEsPTCyWTSLgoVzJ4tdDKn6b5I-Uk8UEoMIE1GZQ$> . You are receiving this because you authored the thread.Message ID: ***@***.***>

-- This message and its contents, including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email.

[CarpetHead]

We are having an issue in assimp v5.2.4 where the merge vertices postprocessing step breaks rigged models. Could this PR be the cause? We are now seeing duplicated vertex weights. See attached image for 5.2.4 and 5.2.3 vertex weights per bone

[nezticle]

We are having an issue in assimp v5.2.4 where the merge vertices postprocessing step breaks rigged models. Could this PR be the cause? We are now seeing duplicated vertex weights. See attached image for 5.2.4 and 5.2.3 vertex weights per bone

We have experience this same thing in Qt Quick 3D with 5.2.4. It breaks Morph animations because all of the weights are identical. We're reverting to 5.2.3 for the time being.

[JG-Adams]

I have not experienced this issue. And I'm not sure how to recreate it.
But I tried to understand the change made here. It look like inside JoinVerticesProcess::ProcessMesh the mechanism to avoid doubles was replaced with something outside of it? Where does it function in?
unordered map by itself shouldn't take more than one of any kind.
This need some explanation, because it don't make a whole lot of sense to me.
We should review the old code. Because I notice the comment says it does something important that otherwise would mess up the mesh.

Edit: then again. It shouldn't have caused any problem should it?

[inhosens]

This patch set does not compare color, texcoords and anim meshes. I am not sure it can accelerate the routine when the whole comparisons are added.

[CarpetHead]

btw it looks like this PR fixed the issue https://github.com/assimp/assimp/pull/4707/files

[JG-Adams]

btw it looks like this PR fixed the issue https://github.com/assimp/assimp/pull/4707/files

That look like it does. :)