Slow Performance on Android TV

EExNull-Tyelor · 25 de Abr de 2023

Harald Thanks for posting the additional info. In general, the skeleton metrics values are all on the high side, which however might be justified depending on your scene setup, having only three skeletons visible. Obviously it would always help to reduce the vertex, bone, constraint, etc. counts, and especially avoiding using clipping polygons.

If you want a bit of extra context, check out Election Year Knockout. We are making a sequel to our game, and so the Spine characters are essentially the entire game, hence the high metric count.

Harald FYI: There is also a section on performance recommendations on the spine-unity documentation page here:

Thanks! I went and read through there, but didn't see any "a-ha!"s or gotchas that I think I might've missed there unfortunately...

Harald You could have a quick try whether disabling clipping at the opponent SkeletonAnimation component improves the situation by changing the Inspector to Debug mode and at the SkeletonAnimation component disabling Use Clipping.

Harald Please also be sure to measure FPS and general timings with Deep profile setting disabled and with release builds, otherwise profiler augmentations might distort the actual timings.

I went and gave these a quick test, making sure I wasn't using a development build/profiling. With only the Opponent spine rendering, and nothing else, we get around 28 FPS on the Android TV regardless of whether or not the clipping is enabled or disabled (we are targeting 60 FPS in game and 30 FPS in menu). The moment another SkeletonAnimations is added (either the player or crowd) we lose about 10 FPS.

Something else I'd like to point out is that the exact same Player and Crowd Skeletons, along with different Opponent Skeletons are currently used in Cocos2d-x and run at around 60 FPS on the same Android TV. I don't think any of us anticipated such a massive difference in performance for the Spine Runtimes between C++/Cocos2d-x and C#/Unity, and it's quite disheartening to be honest since most of our players on Election Year Knockout are on Android TV...

Mario · 26 de Abr de 2023

ExNull-Tyelor I wouldn't necessarily put the blame on the runtime itself, it sounds like the Unity renderer might be at fault here, especially if the Cocos2d-x version runs smoothly. Our Cocos2d-x renderer is actually a custom thing that directly calls into OpenGL.

I'm not super familiar with Unity's renderer options, but my guess is whatever you selected might be too much for Android TV devices. These devices generally use SoCs with GPUs that are terrible when it comes to alpha blended overdraw and complex fragment shaders. If you're using a Unity renderer that uses a complex fragment shader internally, that might be what is causing the performance issues. What Unity version and renderer are you using?

EExNull-Tyelor · 26 de Abr de 2023

Mario I wouldn't necessarily put the blame on the runtime itself, it sounds like the Unity renderer might be at fault here, especially if the Cocos2d-x version runs smoothly. Our Cocos2d-x renderer is actually a custom thing that directly calls into OpenGL.

If that were indeed the case I'd expect the bottleneck to be in the 'GPU Usage' and not the 'CPU Usage' in the Unity Profiler, however I see the exact opposite. Specifically in my first post you can see the calls on the CPU that are taking the most time are all stemming from SkeletonAnimation.LateUpdate with over 55 ms used to LateUpdate the Skeleton's mesh.

Mario I'm not super familiar with Unity's renderer options, but my guess is whatever you selected might be too much for Android TV devices. These devices generally use SoCs with GPUs that are terrible when it comes to alpha blended overdraw and complex fragment shaders. If you're using a Unity renderer that uses a complex fragment shader internally, that might be what is causing the performance issues. What Unity version and renderer are you using?

We're not using HDRP/URP or anything like that, just the default renderer that Unity is packaged with for a 2D Mobile Project template. Specifically we are using Unity 2021.3.20f1 at the moment, one of currently supported and suggested versions for Unity. We have the Adaptive Performance package as well, but haven't done anything with that API yet. As far as I'm aware this should be one of the most lightweight setups for Unity rendering and Spine.

We also aren't doing any complex blending or anything like that on our Crowd nor our Player, and are just using the default Spine Shaders/Materials for those as well. On the opponent we are doing a slightly modified version of the Spine shader for outlines, but even with only the Player and Crowd instantiated we get about _ FPS (on a release build without profiling). The Crowd and Player do both have their slots colored programatically on their Awake calls, to allow for player customization and matching the crowd to the background colors, but this is only done once on Awake. We also did this same exact thing, on the same exact skeletons, in EYKO using Cocos2d-x, so it seems very unlikely that this could be the issue (especially since the bottleneck definitely appears to be in the CPU usage).

Edit: After profiling without Deep Profiling (as I had done for my very first post), I do indeed see the issue is likely GPU bound, though I'm struggling to see how Unity's renderer is so much less efficient than Cocos2d-x's OpenGL renderer. I know that Unity's renderer is heavy but I still assumed far better performance than this on TV. Removing the use of an outline (or custom rim-light shader) does actually get us about 10 extra FPS, from an unstable 20. However we are still not getting 60 FPS. But after removing that we are seeing around 30 FPS (though extremely unstable) with all three skeletons. The profiler now is indicating that it is indeed likely GPU bound since the biggest CPU call is now Gfx.WaitForPresentOnGfxThread which indicates that the CPU is waiting a while for the GPU to finish rendering. Why this might be the case I'm not sure, as without any Skeletons it easily manages to hit a stable 60 FPS as well with these scenes.

With the nature of our game we can't really afford to change the way the artists create these skeletons, and I don't really see much way to simplify our renderer or shaders much more than they already are to get better performance, unless there is a simplified Spine shader suitable for our use case on Mobile/TV devices.

Edit 2: I've opened a forum post here on the Unity Forums to hopefully get some opinions, suggestions, and recommendations from Unity-specific experts/developers.

Harald · 27 de Abr de 2023

ExNull-Tyelor We're not using HDRP/URP or anything like that, just the default renderer that Unity is packaged with for a 2D Mobile Project template. Specifically we are using Unity 2021.3.20f1 at the moment, one of currently supported and suggested versions for Unity. We have the Adaptive Performance package as well, but haven't done anything with that API yet. As far as I'm aware this should be one of the most lightweight setups for Unity rendering and Spine.

Please note that the standard render pipeline is not the most-lightweight pipeline. Universal Render Pipeline (URP), which is the successor of the Lightweight Render Pipeline (LWRP), is usually the first recommendation for mobile devices. It's the lightweight counterpart to High Definition Render Pipeline (HDRP)

Could you please share a screenshot of your Project Settings - Player settings? You could have a try whether using the IL2CPP scripting backend improves the situation.

EExNull-Tyelor · 27 de Abr de 2023

Harald Could you please share a screenshot of your Project Settings - Player settings? You could have a try whether using the IL2CPP scripting backend improves the situation.

Certainly! (Sorry the image is so long, the spoiler tag doesn't shorten the empty space the image takes up on these forums, or I'd have hidden it in a spoiler.)

After setting our scripting backend to IL2CPP (which I thought I already did, but must've done only for iOS), as well as disabling the outline shader effect, we get... an unstable 30 FPS still with all three skeletons (though it might be more stable than without IL2CPP). Most of the time here is still being spent waiting for the GPU in the script as far as I can tell (Gfx.WaitForPresentOnGfxThread).

Some other testing has shown that using URP, with the default Spine URP Example 2D Scene from Spine, I only get about 15 FPS on the Android TV (while profiling), with most of the time being spent waiting for the GPU in the script (Gfx.WaitForPresentOnGfxThread).

After disabling all the point and directional lights in that scene I get around 57-59 FPS (while profiling) but with spikes down to 30 FPS, seemingly on a cyclical thing. Again these spikes seem to also be due to time being spent waiting for the GPU (Gfx.WaitForPresentOnGfxThread).

~~I'll report back here once I test all three skeletons rendering with URP next.~~

Edit: Welp. Trying URP with the three skeletons, Player, Opponent, and Crowd, is not boding well so far.

Using URP with the default Spine Skeleton shaders only gives us about 15 FPS now, whereas the default renderer with default shaders got us around 30 FPS (albeit unstably). Using URP with the Universal Render Pipeline/2D/Spine/Sprite shader gives us a whopping 6 FPS Then finally using URP with the Universal Render Pipeline/2D/Spine/SkeletonLit shader gives us 10 FPS. GPU Instancing doesn't seem to make a difference here either.

In all three of these tests there were no normal maps, emission maps, fragment shaders (outside of any you've used for your shaders in Spine's Unity Package), no light sources, etc. It was quite literally three skeletons, a camera, a global light 2d, and a single canvas with a single TextMeshPro on it to display the FPS averaged over the last 60 frames. About as simple of a scene as we could make for this test.

Mario · 27 de Abr de 2023

After disabling all the point and directional lights in that scene

How many point and directional lights do you have? I assume you had no such thing in Cocos2d-x. The way those are implemented in Unity's renderers means your fragment shader is likely super complex, which would explain the high GPU side load on those low-performance Android TV chipsets.

EExNull-Tyelor · 27 de Abr de 2023

Mario How many point and directional lights do you have? I assume you had no such thing in Cocos2d-x. The way those are implemented in Unity's renderers means your fragment shader is likely super complex, which would explain the high GPU side load on those low-performance Android TV chipsets.

We have no lighting in either our Cocos2d-x game or our Unity game. I was referring to Spine's URP 2D Example Scene which is preloaded with several point and directional lights. (Specifically the one that has two StretchymanURPs and the RaptorProURP skeleton in it.)

Edit: I've updated the post above with more testing using the URP with our three skeletons (the Player, Opponent, and Crowd). The results were... less than promising

Edit 2: I'd also like to add that I just attempted to disable all post-processing, shadows, and lighting in the URP Asset data, and only get about 20 FPS with that setup (though its still unstable with spikes down to 15 FPS).

Harald · 28 de Abr de 2023

@ExNull-Tyelor Very sorry to hear that the situation did not improve with using the URP pipeline! To be sure, could you perhaps share the settings of the used Universal Render Pipeline Asset that is assigned at Project Settings - Graphics - Scriptable Render Pipeline Settings. Also please check whether under Project Settings - Quality there is no Render Pipeline Asset assigned which overrides the one in the Graphics section. If you haven't already, please be sure to set everything under the Lighting section to the minimal settings, to e.g. not use multiple secondary per-pixel lights.

Additionally, could you please share your Material settings you are using at your skeletons? Does the situation change if you are using an unlit URP shader (either Universal Render Pipeline/Spine/Skeleton)?

EExNull-Tyelor · 28 de Abr de 2023

Harald To be sure, could you perhaps share the settings of the used Universal Render Pipeline Asset that is assigned at Project Settings - Graphics - Scriptable Render Pipeline Settings.

Sure! Here are the settings I used for my URP Asset during testing:

Then the settings in the Renderer Data Asset are:

Harald Also please check whether under Project Settings - Quality there is no Render Pipeline Asset assigned which overrides the one in the Graphics section.

I assigned the URP Asset to each Render Pipeline Asset slot for each Quality Level in Project Settings - Quality and not in the Project Settings - Graphics since I assumed this would give us greater control to have more complex Render Pipelines for Desktop builds vs Mobile builds. Should I assign it through the Graphics settings instead?

Harald Additionally, could you please share your Material settings you are using at your skeletons? Does the situation change if you are using an unlit URP shader (either Universal Render Pipeline/Spine/Skeleton)?

That was one of the shaders I was attempting to use for this experiment actually. The settings I'm using are:

I'm using the same general settings for the other two skeletons materials as well (one for the Opponent and two for the Player).

EExNull-Tyelor · 1 de May de 2023

Hate to be a bother, but any extra help or information you might be able to offer @Harald ?

A small test showed a rotating capsule was able to hit 60 FPS on this Android TV while using URP. Though the RenderFrame was taking nearly 10 ms

Misaki · 1 de May de 2023

ExNull-Tyelor Sorry for the wait! Yesterday was a public holiday and Harald will be back today, so please wait a little longer for him.

Harald · 2 de May de 2023

Sorry for the late reply!

ExNull-Tyelor I assigned the URP Asset to each Render Pipeline Asset slot for each Quality Level in Project Settings - Quality and not in the Project Settings - Graphics since I assumed this would give us greater control to have more complex Render Pipelines for Desktop builds vs Mobile builds. Should I assign it through the Graphics settings instead?

No, that's fine as well, you just need to be sure which render pipeline asset will be used, to tweak the actually used asset.

While I don't think it will improve the situation, you could have a try changing Light Blend Styles entry Rim to Mask Texture Channel from R to None. I doubt that any unnecessary pass will be triggered if no lights are active at all, but just to be sure no unnecessary overhead is generated.

ExNull-Tyelor That was one of the shaders I was attempting to use for this experiment actually. The settings I'm using are:

Here you are using a lit 2D shader (truncated, but likely Universal Render Pipeline/2D/Spine/Skeleton Lit). Please use an unlit non-2D shader if you don't need any lighting applied. Like Universal Render Pipeline/Spine/Skeleton.

EExNull-Tyelor · 2 de May de 2023

Misaki ExNull-Tyelor Sorry for the wait! Yesterday was a public holiday and Harald will be back today, so please wait a little longer for him.
Harald Sorry for the late reply!

Not a problem at all, and I apologize for trying to bother you all on your holiday!

Harald Here you are using a lit 2D shader (truncated, but likely Universal Render Pipeline/2D/Spine/Skeleton Lit). Please use an unlit non-2D shader if you don't need any lighting applied. Like Universal Render Pipeline/Spine/Skeleton.

Interestingly, changing the Skeleton's shaders to the one you suggested, Universal Render Pipeline/Spine/Skeleton, causes the Skeletons to stop rendering entirely.

Edit: Actually both the Universal Render Pipeline/Spine/Skeleton and Universal Render Pipeline/Spine/Skeleton Lit cause the Skeleton to stop rendering entirely, as does Universal Render Pipeline/Spine/Sprite. Though Universal Render Pipeline/Spine/Outline/Skeleton-OutlineOnly does render only the outline of the Skeleton. Only the Spine shaders under Universal Render Pipeline/2D/Spine/ are working (Skeleton Lit and Sprite) are actually rendering somewhat properly, but the Sprite shader creates outlines between all the symbols that aren't in the actual atlas file.

Harald While I don't think it will improve the situation, you could have a try changing Light Blend Styles entry Rim to Mask Texture Channel from R to None. I doubt that any unnecessary pass will be triggered if no lights are active at all, but just to be sure no unnecessary overhead is generated.

Doing this with the Lit Skeleton shaders doesn't seem to give any noticeable improvement in performance. Though I can't seem to use the un-lit shaders like you suggested above to see if there is a difference there...

Harald · 3 de May de 2023

Not a problem at all, and I apologize for trying to bother you all on your holiday!

No need to apologize, how should you know!

ExNull-Tyelor Interestingly, changing the Skeleton's shaders to the one you suggested, Universal Render Pipeline/Spine/Skeleton, causes the Skeletons to stop rendering entirely.

Terribly sorry, my mistake, for some reason I incorrectly assumed that the unlit Universal Render Pipeline/Spine/Skeleton shader would render normally with 2D renderer, however it is not finding the respective shader pass and does not render anything at all. We have just pushed a commit to the 4.1 branch which adds an unlit URP 2D shader, available under Universal Render Pipeline/2D/Spine/Skeleton.

A new Spine URP Shaders 4.1 UPM package is available for download here as usual:
https://esotericsoftware.com/spine-unity-download
Please let us know if this improves the situation.

EExNull-Tyelor · 3 de May de 2023

Harald erribly sorry, my mistake, for some reason I incorrectly assumed that the unlit Universal Render Pipeline/Spine/Skeleton shader would render normally with 2D renderer, however it is not finding the respective shader pass and does not render anything at all. We have just pushed a commit to the 4.1 branch which adds an unlit URP 2D shader, available under Universal Render Pipeline/2D/Spine/Skeleton.

A new Spine URP Shaders 4.1 UPM package is available for download here as usual:
https://esotericsoftware.com/spine-unity-download
Please let us know if this improves the situation.

Hey no problem at all! I'm happy to report that after updating the package and then switching the shaders to Universal Render Pipeline/2D/Spine/Skeleton it basically doubles the frame rate from 15 FPS to 30-34 FPS on our Android TV device while profiling a Development Build!

Still not the 60 FPS we had hoped for, but we can probably get away with having a lower FPS for TVs

Thank you and the rest of the Esoteric Software team so much for the help you have provided us! If you ever have any other help, suggestions, improvements or fixes that you think might help us out, we'd love to hear them as well.

For a little bit more info the frame debugger is showing that two of our meshes can't be SRP batched, but I'm not entirely sure why. Could it have something to do with this CBUFFER thing, since I assume the Android TV wouldn't support Vulkan.

The Gfx.WaitForPresentOnGfxThread seems to be quicker on a cyclical pattern as well, which seems odd to me. You can see this on the CPU Usage with these Gfx.PresentFrame spikes as well, taking between 13ms and 28ms to complete, which causes Gfx.WaitForPresentOnGfxThread to fluctuate between 5ms and 22ms.

Anywho, thank you again so much for the help you've already given us and for this un-lit 2d skeleton patch

Harald · 4 de May de 2023

ExNull-Tyelor Hey no problem at all! I'm happy to report that after updating the package and then switching the shaders to Universal Render Pipeline/2D/Spine/Skeleton it basically doubles the frame rate from 15 FPS to 30-34 FPS on our Android TV device while profiling a Development Build!

Very glad to hear that it improved the situation, thanks for the info and for your kind words! Please also be sure to judge the final timings with Development Build disabled.

ExNull-Tyelor For a little bit more info the frame debugger is showing that two of our meshes can't be SRP batched, but I'm not entirely sure why. Could it have something to do with this CBUFFER thing, since I assume the Android TV wouldn't support Vulkan.

This is indeed strange. Which Graphics API are you using in your Player settings? You could explicitly set a graphics API like OpenGLES3 instead of "Auto" to be sure which one is used. We will do some more investigations on our end, so far we received the expected SRP batching of skeletons with the new unlit shader on our end when using e.g. OpenGLES3.

Apart from that, I see 29 draw calls in your profiler screenshot, but you mentioned you only have 3 skeletons active. I assume that each of your skeletons is not using a single atlas page texture but multiple. How many are you using per skeleton? While I'm, not sure how you setup the skins of your characters, you could consider either packing combined skins to a single atlas page, or grouping attachment images differently to atlases (or changing draw order, if possible), so that you don't have as many necessary texture switches, like avoiding ABABABA and instead grouping to e.g. AABBBAA which will then result in draw calls ABA.

ExNull-Tyelor The Gfx.WaitForPresentOnGfxThread seems to be quicker on a cyclical pattern as well, which seems odd to me. You can see this on the CPU Usage with these Gfx.PresentFrame spikes as well, taking between 13ms and 28ms to complete, which causes Gfx.WaitForPresentOnGfxThread to fluctuate between 5ms and 22ms.

A quick guess is that this could be due to unlucky timing when waiting for VSync, taking a bit longer and then having to wait for the next VSync timepoint.

EExNull-Tyelor · 4 de May de 2023

Harald This is indeed strange. Which Graphics API are you using in your Player settings? You could explicitly set a graphics API like OpenGLES3 instead of "Auto" to be sure which one is used. We will do some more investigations on our end, so far we received the expected SRP batching of skeletons with the new unlit shader on our end when using e.g. OpenGLES3.

I had it set to use Vulkan only, since on Android with Unity only Vulkan supports Dynamic Resolution. Setting it back to only OpenGLES3 causes the spikes above 60 FPS to become more narrow, gaining us about 2 FPS interestingly enough.

Harald Apart from that, I see 29 draw calls in your profiler screenshot, but you mentioned you only have 3 skeletons active. I assume that each of your skeletons is not using a single atlas page texture but multiple. How many are you using per skeleton? While I'm, not sure how you setup the skins of your characters, you could consider either packing combined skins to a single atlas page, or grouping attachment images differently to atlases (or changing draw order, if possible), so that you don't have as many necessary texture switches, like avoiding ABABABA and instead grouping to e.g. AABBBAA which will then result in draw calls ABA.

Is there some more information about this somewhere that I can give to the artists? I'm certain that this is what's happening, as only one of the skeletons has more than one atlas and uses skins (the player spine uses two atlases with several skin variations). In the inspector because of this it creates 29 materials for this one skeleton. But I don't do the art or work with Spine directly myself as the engineer on the team, so I'm not exactly sure how to explain this to the artists or help them solve this issue. Some additional context is that we are also dynamically coloring many of the region and mesh attachments for the player's skeleton as well. Would this, in addition to the second atlas, be causing the ABABABA[...] rendering loop, or is it solely on how the spine itself was created by the artists/how the atlas was exported.

Rendering only the other two skeletons, opponent and crowd, brings us up to 50 FPS as well.

Harald A quick guess is that this could be due to unlucky timing when waiting for VSync, taking a bit longer and then having to wait for the next VSync timepoint.

Unfortunately that couldn't be it, as not only do I have VSync disabled, but both Every VBlank and Every Second VBlank are both ignored on Android, iOS, and tvOS according to the Project Settings window.

Another curiosity is that the crowd skeleton is the only one I can see in the Frame Debugger. No frame that I click on seems to contain the opponent skeleton anywhere that I can see (and it uses an entirely different atlas than the crowd does). This is a screenshot of the frame debugger without the player skeleton rendering:

Harald Very glad to hear that it improved the situation, thanks for the info and for your kind words! Please also be sure to judge the final timings with Development Build disabled.

In testing this I found with all three skeletons the Release Build had no noticeable improvement in FPS over the Development Build oddly enough. Perhaps because this scene and project are so simple?

Harald · 5 de May de 2023

ExNull-Tyelor I had it set to use Vulkan only, since on Android with Unity only Vulkan supports Dynamic Resolution. Setting it back to only OpenGLES3 causes the spikes above 60 FPS to become more narrow, gaining us about 2 FPS interestingly enough.

Thanks for the info! We will try if we can reproduce this issue using Vulkan as rendering API then.

ExNull-Tyelor Is there some more information about this somewhere that I can give to the artists? I'm certain that this is what's happening, as only one of the skeletons has more than one atlas and uses skins (the player spine uses two atlases with several skin variations).

You can find a short explanation here:
https://esotericsoftware.com/spine-unity#Material-Switching-and-Draw-Calls
Improving the draw order or grouping attachments differently to atlas page textures could certainly reduce the number of draw calls when you have only few atlas textures (i.e. few Materials displayed at the bottom in the skeleton's Inspector). What does the current list of materials at your skeleton look like in the Inspector, do you only see two materials alternating many times, or do you see many different materials (e.g. because you're using blend modes at certain slots and slot material overrides)? In any case, grouping attachments (ordering them so that they are in sequence in the draw order) that require the same material would "fuse" two or more draw calls to a single one.

You might also try if runtime repacking fits all your used attachments into a single atlas texture, see sub-section "Runtime Repacking" here:
https://esotericsoftware.com/spine-unity#Combining-Skins

ExNull-Tyelor Some additional context is that we are also dynamically coloring many of the region and mesh attachments for the player's skeleton as well.

If you only use slot colors and not slot material overrides (the CustomMaterialOverride property), this should not matter, since it's only writing different vertex colors then. It can be done in a single draw call.

BTW: Are you using Additive blend mode at some slots perhaps? If so, and if you're not already using the default PMA workflow, you might want to switch from straight alpha to PMA, then additive slots can be rendered in a single draw-call.

ExNull-Tyelor Unfortunately that couldn't be it, as not only do I have VSync disabled, but both Every VBlank and Every Second VBlank are both ignored on Android, iOS, and tvOS according to the Project Settings window.

Ok, thanks for the info.

ExNull-Tyelor Rendering only the other two skeletons, opponent and crowd, brings us up to 50 FPS as well.

Hm, this unfortunately still does not sound stellar.

ExNull-Tyelor Another curiosity is that the crowd skeleton is the only one I can see in the Frame Debugger. No frame that I click on seems to contain the opponent skeleton anywhere that I can see (and it uses an entirely different atlas than the crowd does). This is a screenshot of the frame debugger without the player skeleton rendering:

That's strange indeed. You could do a test in the Unity Editor and watch the Game view while going back and forth in the draw-call-timeline at the top of the screenshot, then you should see when the crowd appears in the list of draw calls. If the list is completely different, that won't help of course..

ExNull-Tyelor In testing this I found with all three skeletons the Release Build had no noticeable improvement in FPS over the Development Build oddly enough. Perhaps because this scene and project are so simple?

Ok, that at least saves compile time for diffent builds when testing then . A quick guess it that it could be because it's GPU limited and any optimizations would mainly affect performance on the CPU side. Anyway, that's just speculations of course.

EExNull-Tyelor · 5 de May de 2023

Harald You can find a short explanation here:
https://esotericsoftware.com/spine-unity#Material-Switching-and-Draw-Calls

Thanks, I'll give this a read through and show it to the artists as well!

Harald What does the current list of materials at your skeleton look like in the Inspector, do you only see two materials alternating many times, or do you see many different materials (e.g. because you're using blend modes at certain slots and slot material overrides)? In any case, grouping attachments (ordering them so that they are in sequence in the draw order) that require the same material would "fuse" two or more draw calls to a single one.

I see the same two materials alternating back and fourth 29 times, which is why I had assumed that this could be part of the issue and that "fusing" these 29 draw calls into 2 or 3 I'd think would get us closer to the 50 FPS with all three skeletons rendering. We also aren't doing any type of blending or anything with these spines, just the "Normal" blending mode.

Harald You might also try if runtime repacking fits all your used attachments into a single atlas texture, see sub-section "Runtime Repacking" here:
https://esotericsoftware.com/spine-unity#Combining-Skins

I can try that out when we start using the skins we've implemented in this test scene, but right now I have only applied a single skin through the SkeletonAnimation inspector window. (I might be doing something wrong too, because a very quick test I threw together to test the Runtime Repacking feature gave me a malformed atlas and several errors in the Console window with the following format:
Graphics.CopyTexture called with region not fitting in source element: trying to copy from region (x:2218, y:1147, width:194, height:136) of mip 0 (size: 2048x2048)

I did set those textures to Read/Write as well, and it didn't improve the situation there.

The script I used for this test was mostly CnP from your Spine Examples for Mix and Match, but here it is anyways in case you spot something I'm doing entirely wrong. My biggest guess as to what I'm doing wrong is that I'm using the wrong material, since the skeleton uses two materials (alternating back and forth 29 times as I mentioned earlier).

using UnityEngine;
using Spine.Unity;
using Spine.Unity.AttachmentTools;

public class DummyRepacker : MonoBehaviour
{
    public SkeletonAnimation Skeleton;

    public Texture2D runtimeAtlas;
    public Material runtimeMaterial;
    
    void Start()
    {
        // Create a repacked skin.
        Spine.Skin repackedSkin = new Spine.Skin("repacked skin");
        Spine.Skin skin = Skeleton.skeleton.Data.FindSkin("male");
        repackedSkin.AddSkin(skin);
        skin = Skeleton.skeleton.Data.FindSkin("male_head_1");
        repackedSkin.AddSkin(skin);

        repackedSkin = repackedSkin.GetRepackedSkin("repacked skin", Skeleton.SkeletonDataAsset.atlasAssets[0].PrimaryMaterial, out runtimeMaterial, out runtimeAtlas);
        Skeleton.skeleton.SetSkin(repackedSkin);

        // Use the repacked skin.
        Skeleton.Skeleton.Skin = repackedSkin;
        Skeleton.Skeleton.SetSlotsToSetupPose();
        Skeleton.AnimationState.Apply(Skeleton.Skeleton); // skeletonMecanim.Update() for SkeletonMecanim

        // You can optionally clear the cache after multiple repack operations.
        AtlasUtilities.ClearCache();
        Resources.UnloadUnusedAssets();
    }
}

Harald That's strange indeed. You could do a test in the Unity Editor and watch the Game view while going back and forth in the draw-call-timeline at the top of the screenshot, then you should see when the crowd appears in the list of draw calls. If the list is completely different, that won't help of course..

Sorry about that, it was just my misunderstanding of how the Frame Debugger is displayed. With both the Crowd and the Scam Caller they get properly rendered with SRP Batch in two draw calls, and if I disable one or the other I see a single draw call in SRP Batch now. So it's only the player spine that isn't getting properly batched, even with OpenGLES3 enabled.

Harald Ok, that at least saves compile time for diffent builds when testing then . A quick guess it that it could be because it's GPU limited and any optimizations would mainly affect performance on the CPU side. Anyway, that's just speculations of course.

Actually that makes a lot of sense, I hadn't thought of that. Because the CPU isn't being hit almost at all with these test I'm running, everything is GPU bound. I'm sure that the Unity Profiler doesn't use hardly any GPU bandwidth at all, and likely just eats up CPU Usage instead.

Harald · 8 de May de 2023

ExNull-Tyelor I see the same two materials alternating back and fourth 29 times, which is why I had assumed that this could be part of the issue and that "fusing" these 29 draw calls into 2 or 3 I'd think would get us closer to the 50 FPS with all three skeletons rendering. We also aren't doing any type of blending or anything with these spines, just the "Normal" blending mode.

Thanks for the additional info, that's good news regarding potential improvement, as 29 draw calls for a single skeleton sounds close to worst-case order. Then we would definitely recommend changing draw order or grouping of attachment images to atlas pages.

ExNull-Tyelor I can try that out when we start using the skins we've implemented in this test scene, but right now I have only applied a single skin through the SkeletonAnimation inspector window. (I might be doing something wrong too, because a very quick test I threw together to test the Runtime Repacking feature gave me a malformed atlas and several errors in the Console window with the following format:
Graphics.CopyTexture called with region not fitting in source element: trying to copy from region (x:2218, y:1147, width:194, height:136) of mip 0 (size: 2048x2048)

I did set those textures to Read/Write as well, and it didn't improve the situation there.

Did you check whether (apart from "Read/Write Enabled") the other requirements are met? You can find them in the note box listing "Important Note: If repacking fails or creates unexpected results, it is most likely due to any of the following causes:" in the Runtime Repacking section.

If all of these are met, could you perhaps send us a minimal Unity project that still shows this issue? You can send it as a zip package to contact@esotericsoftware.com, briefly mentioning this forum thread URL so that we know the context.

ExNull-Tyelor Sorry about that, it was just my misunderstanding of how the Frame Debugger is displayed. With both the Crowd and the Scam Caller they get properly rendered with SRP Batch in two draw calls, and if I disable one or the other I see a single draw call in SRP Batch now. So it's only the player spine that isn't getting properly batched, even with OpenGLES3 enabled.

No need to apologize. That's good to hear at least. Then merging at least some of the player-skeleton's draw calls by avoiding unnecessary atlas page texture switches should improve the situation.