Unity's Mono Problem: C# Code Runs Slower Than It Should
Key topics
The Unity game engine's sluggish C# performance has sparked debate, with many attributing it to the engine's reliance on Mono instead of CoreCLR. While some commenters lament Unity's slow progress in making the switch, citing delays and talent drain, others point to the company's tumultuous business decisions and financial constraints as potential roadblocks. As alternatives like Godot gain traction, with its C# support improving, the discussion highlights the trade-offs between sticking with Unity and exploring other options. The conversation is particularly relevant now as developers weigh the pros and cons of investing in a platform with lingering performance issues.
Snapshot generated from the HN discussion
Discussion Activity
Very active discussionFirst comment
1h
Peak period
62
0-6h
Avg / period
17.8
Based on 160 loaded comments
Key moments
- 01Story posted
Dec 28, 2025 at 4:41 PM EST
4d ago
Step 01 - 02First comment
Dec 28, 2025 at 5:47 PM EST
1h after posting
Step 02 - 03Peak activity
62 comments in 0-6h
Hottest window of the conversation
Step 03 - 04Latest activity
Dec 31, 2025 at 11:12 AM EST
1d ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
Might I suggest https://github.com/stride3d/stride, which is already on .net 10 and doesn't have any cross-boundary overhead like Unity.
Unity updates on their plans and progress:
2022 - officially announced plans to switch to CoreCLR - https://unity.com/blog/engine-platform/unity-and-net-whats-n...
2023 - Tech update - https://unity.com/blog/engine-platform/porting-unity-to-core...
Unite 2025 - CoreCLR based player scheduled for Unity 6.7 in 2026 - https://digitalproduction.com/2025/11/26/unitys-2026-roadmap...
As an outsider, it certainly seems like there's reason for skepticism.
They'll get there eventually, but the current roadmap says experimental CoreCLR in late 2026, which then in the best case means production ready in 2027. Unity isn't going anywhere, but at least as a dev who doesn't care about mobile (which is Unity's real market), competing engines have gotten much more attractive in the last couple years and that seems set to continue.
The guy should just have been left alone and shielded from company bullshit.
Whoever from high-management lost this guy is an absolute loser waste of space who didn't do his job.
Fuck companies, and fuck loser managers.
From my perspective, Unity seems very poorly managed in recent years. The editor experience isn't improving while they continue the usual pattern of shipping features in a poor state where they need another couple versions to become properly usable, and of course they make terrible decisions like the runtime fee, a total insanity that caused a huge loss of trust and boosted Godot development enormously.
Of course my perspective is biased by me not being Unity's main target market. I work on PC strategy games, which are on Steam. At our studio, we don't do mobile, advanced graphics features aren't very relevant, and we may have the most complex UI that ever shipped in a Unity game.
People complain about web development but working with native apps can be depressing sometimes.
Godot is the only real open source competitor, their C# support is spotty. If I can't build to Web it's useless for game jams as no one should be downloading and running random binaries.
A real sandbox solution with actual GPU support is needed.
I've seen this issue before, they're making progress but theirs no firm release date.
Plus you then have to extensive testing to see what works in Web builds and what doesn't. I REALLY enjoy vibe coding in Godot, but it's still behind Unity in a lot of ways.
For mathy stuff, 100% c# is going to be better. But if you need to round trip to the engine a lot getting stuff in and out of the dotnet heap can actually hurt performance. You also have to be _really_ careful because there are a lot of cases you generate accidental garbage (biggest one is when you use strings that are getting implicitly converted to StringNames every time you run a function, you can avoid this by pre-generating them as consts but I've run into a fair few people who never ran dotmemory or the like to see the issues).
But C# is what I've used for over a decade. C# has vastly better IDE support. It's a more complete language.
Plus a lot of C# assets/libraries don't really have GDScript counterparts.
It's very difficult to me, I generally stick to high level stuff , C#, JavaScript, Python, Dart, etc.
Not denigrating, genuine question.
Headers.
That's just the start. The C++ build system and package managers are the stuff if nightmares. Modern languages are significantly easier to use.
Don't get me wrong, if you offer a job with a 200k base salary and give me 6 months to learn C++ I'll do it. But I won't enjoy it, and I definitely won't do it as a hobby.
I think it's overly verbose and probably has a lot of unneeded legacy stuff, but in terms of making money so I can afford donuts, it's not bad.
My personal favorite language is probably Dart, but it's basically useless outside of Flutter and I don't have faith in Google to keep Flutter going.
I don't like low level programming. My dream language is literally an LLM that processes a markdown document where I sorta just describe what I want to happen. With options to call defined functions in a normal programing language
If I had money ( VC money) I'd be working on this.
You only dabble in the c++ for the sliver of the project that needs it. 90% of game development is animating stuff and user interface development. GDScript is great for that.
Pointers, manual memory management, and the design by committee standard library are not fun.
If performance is a concern, skip C# and go straight to c++. Now your ripping at max speed with the smallest binary! That's my whole point. GDScript + c++ is my point. Ditch C# it's not worth the squeeze.
Interpreter code is never faster than a dynamic compiler, otherwise what about doing games in Python?
As mentioned on my comment, GDextension experience kind of sucks.
I'm offering real practical advice from experience of having worked on real projects.
I'll make it real clear:
GDScript & c++ > C#
I won't even get into how big of projects I've written in GDScript successfully.
I don't want to have any sort of undefined behavior
I want to have quick code reload button in the editor
I want to still rely of the engine official documentation with examples like it is with GDScript and C#
The purpose of recommending c++ here is:
If GDScript is too slow, reach directly for C++.
I'm specifically recommending GDScript over C# for ease of use and c++ over C# for performance.
GDScript is undoubtedly better integrated in the engine, but I would have expected C# compare more favorably in larger projects than the game jam sized projects I have made.
It compares several C# implementations of raycasts, never directly compares with GDScript, blames the C# performance on GDScript compatibility and has an strike-out'ed section advocating dropping of GDScript to improve C# performance!
Meanwhile, Godot's official documentation[1] actually does explicitly compare C# and GDScript, unlike the the article which just blames GDScript for C#'s numbers, claiming that C# wins in raw compute while having higher overhead calling into the engine
[1]: https://docs.godotengine.org/en/stable/about/faq.html#doc-fa...
I use GDScript because it’s currently the best supported language in Godot. Most of the ecosystem is GDScript. C# feels a bit bolted-on. (See: binding overhead) If the situation were reversed, I’d be using C#. That’s one technical reason to prefer GDScript. But you’re free to choose C# for any number of reasons, I’m just trying to answer the question.
I would have have put it as a matter of preference/right tool with GDScripts tighter engine integration contrasted with C#'s stronger tooling and available ecosystem.
But with how it was phrased, it didn't sound like expressing a preference for GDScript+C++ over C# or C#++, it sounded like C# had some fatal flaw. And that of course makes me curious. Was it a slightly awkward phrasing, or does Godot have some serious footgun I'm unaware of.
It's not a technical skillset issue, it's a case of lacking resources, leadership changing priorities and changing requirements mid-development, causing a lot of work to be redone.
We still have highly skilled developers working in this area, if we can just actually be allowed to do (and keep) we work we have.
I don't beleive such a large regression from .NET framework to CoreCLR.
But also, as far as this article, it's noting a noting a more specific use case that is fairly 'real world'; Reading a file (I/O), doing some form of deserialization (likely with a library unless format is proprietary) and whatever 'generating a map' means.
Again, this all feels pretty realistic for a use case so it's good food for thought.
> Can someone explain what benchmarks were actually used?
This honestly would be useful in the article itself, as well as, per above, some 'deep dives' into where the performance issues were. Was it in file I/O (possibly Interop related?) Was it due to some pattern in the serialization library? Was it the object allocation pattern (When I think of C# code friendly for Mono I think of Cysharp libraries which sometimes do curious things)? Not diving deeper into the profiling doesn't help anyone know where the focus needs to be (unless it's a more general thing, in which case I'd hope for a better deep dive on that aspect.)
> E.x. can you do things like stackalloc temp buffers to avoid allocation, and does the stdlib do those things where it is advantageous?
The C# standard lib (often called the base class library or BCL) has seen a ton of Span<T>/Memory<T>/stackalloc internal usage adoption in .NET 6+, with each release adding more of them. Things like File IO and serialization/deserialization particularly see a lot of notable performance improvements just from upgrading each .NET version. .NET10 is faster than .NET9 with a lot of the same code, and so forth.
Mono still benefits from some of these BCL improvements (as more of the BCL is shared than not these days, and Blazor WASM for the moment is still more Mono than CoreCLR so some investment has continued), but not all of them and not always in the same ways.
I worded my reply poorly, mostly in that I meant 'If Oberon has it's own stdlib, is it doing the modern performant practice' ?
.NET has heavily invested in performance. If I understand your article correctly, you tested .NET 5 which will be much slower at this point than .NET 10 is.
I also think it matters what you mean by “Mono”. Mono, the original stand-alone project has not seen meaningful updates in many years. Mono is also one of the two runtimes in the currently shipping .NET though and I suspect this runtime has received a lot of love that may not have flowed back to the original Mono project.
It'll be interesting to see how the CoreCLR editor performs. With that big of a speed difference the it might be possible for games to run better in the editor than a standalone Mono/IL2CPP build.
[1] https://discussions.unity.com/t/coreclr-and-net-modernizatio...
For what reason? Mono has a pretty good precise GC since many years.
Platform support is also still better with IL2CPP but .NET is catching up.
Some of that sunk cost may be the above mentioned pointer issue and not enough current plans for a smarter FFI interface between C++ and C#.
You tend to want zero per frame allocation as it is and that would probably not change.
As long as your less frequent garbage doesn't overtake the incremental GC, that's not really an issue either. If it's working incrementally as intended stutter shouldn't be an issue.
In a game there's no endless benefit from raw GC throughput like you might see on a server instance that could always push more requests per second.
It shouldn't but it does. Boehm is a conservative GC so when it triggers it needs to scan a lot more memory for pointers than .NET's GC because it has to assume anything in memory could be a pointer.
If your game is allocating so quickly that the incremental GC can't keep up, I would argue that solving this with a "faster" GC is just taking you further into hell.
1. Uses "I"
2. Look how many times it switches verb tenses here:
"One day I was debugging an issue in map generation and it was time-consuming [...]. To make debugging faster, I’ve written a unit test, hoping to cut down on the turn-around time since Unity takes 15+ seconds just to crunch new DLLs [...]. When I ran the test, it finished in 40 seconds."
Setting up release benchmarks is much more complex and we develop the game in Debug mode, so it is very natural to get the first results there, and if promising, validate them in Release.
Also, since our team works in Debug mode, even gains that only speed things up in Debug mode are valuable for us, but I haven't encountered a case where I would see 20%+ perf gain in Debug mode that would not translate to Release mode.
So in C++ terms, it's really just benchmarking "-O2" instead of "-O2 -DNDEBUG". This seems fine.
On a "proper OS", your WebGPU is 1:1 translating all calls to Vulkan, and doing so pretty cheaply. On Windows, your browser will be doing this depending on GPU vendor: Nvidia continues to have not amazing Vulkan performance, even in cases where the performance should be identical to DX12; AMD does not suffer from this bug.
If you care about performance, you will call Vulkan directly and not pay for the overhead. If you care about portability and/or are compiling to a WASM target, you're pretty much restricted to WebGPU and you have to pay that penalty.
Side note: Nothing stops Windows drivers or Mesa on Linux from providing a WebGPU impl, thus browsers would not need their own shim impl on such drivers and there would be no inherent translation overhead. They just don't.
WebGPU is trying to not repeat this mistake, but it isn't a 100% 1:1 translation for Vulkan, so everyone is going to need to agree to how the C API looks, and you know damned well Google is going to fuck this up for everyone and any attempt is going to die.
Chrome is the cancer killing desktop computing.
You can't just ship a WebGPU implementation in the driver because the last-mile of getting the <canvas> on screen is handled by the browser in entirely browser specific ways. You'd require very tight coordination between the driver and browsers, and you still wouldn't be saving much because the overhead you get from WebGPU isn't because of API translation, rather it's the cost to make the API safe to expose in a browser.
Programs can use Vulkan, D3D, OpenGL, OpenCL, etc, to ex: read memory that isn't in your program's space via the GPU/driver/OS not properly handling pointer provenience. Also, IOMMUs are not always setup correctly, and they are also not bug free, ex: Intel's 8 series.
Using hardware to attack hardware is not new, and not a uniquely web issue.
They've tried and failed to make their own games and they just can't do it. That means they don't have the internal drive to push a new design forward. They don't know what it takes to make a game. They just listen to what people ask for in a vacuum and ship that.
A lot of talented people and Unity but I don't expect a big change any time soon.
C-level set goals are abstract and generic, or sometimes plain naive, and this is often coming from generic requests from the board or VCs.
"Hire as many developers as you can, even if there's no work right now", a Softbank request.
"Don't build, just acquire similar products", from a Brazilian capital management that ended up killing that company.
"Kill this team, their product doesn't sell. I don't care if all our other product depends on theirs", from Francisco Partners.
Employees who stay can't really rock the boat, so it self-selects for non-boat-rocking people. Rockstars who stay must adapt or suffer. Eventually you get so many bad people that you do layoffs.
If the CEO is just a parrot repeating what the board says, you get a company full of parrots too. No pirate to guide the ship.
I don't think this is fair. I'd say Unity's inspector window is one of the good parts of Unity because it not just a property window. It's an immediate mode UI that things can hook into to do a lot more than just property editing.
On the other hand, the lack of "love" from Unity's side (at least regarding Inspector) allows a thriving ecosystem for devs to build their own version of what an ideal Inspector drawer should look like, as well as potentially make a living from it. And to boot, who is to say what the ideal Inspector should look like? Do you trust Unity to make an Inspector that fits everyone's wants/needs? I definitely don't - so I'm glad they have a "bare-bones" version that allows us to customize it to our heart's content. Do you want them to be like Apple and "steamroll" everyone and make bad decisions for arbitrary reasons? I definitely don't and I personally HATE a ton of Apple's constant changes and lack of ability to change simple things, such as the inability to disable a lot of animations, which murders my VNC sessions, but I digress.
Regarding non-Inspector things: you already replied to my other rant about unfinished features, so yeah, also in agreement.
I don't agree with you on the Asset Store, for the same reasons - the rate of breakage means that things that are not constantly updated no longer work - and multple versions need to be maintained for parallel engine versions. That combined with the dubious economics of Asset Store (I don't think it makes financial sense to even make these things, let alone maintain them), they mostly end up as abandonware.
And on the Asset Store if you make something indispensable (which is more often than not something the engine should'have OOTB, like competent text rendering), one of the following things will happen:
- Unity will buy you out, and plop your asset in the engine, without doing any integration work, and will stick out like a sore thumb (TextMeshPro). Good for you, bad for consumers, and sucks if you were making a competitor
- They build an in-house solution, that you obviously can't compete with, and will have a huge leg up on you because they have engine access (sucks to be you)
- The engine will never have that feature, because 'you can just buy it', meaning you have to either spends hundreds of dollars/euros on dubious quality assets or hunt for open-source versions with generally even more varying usability. UE4/5 has a lot of these built in, and AAA quality.
> Unity uses the Mono framework to run C# programs and back in 2006 it was one of the only viable multi-platform implementations of .NET. Mono is also open-source, allowing Unity to do some tweaks to better suit game development. [...] An interesting twist happened nearly 10 years later.
Not mentioned is that Mono itself of course improved a lot over the years, and even prior to MS's about-face on open source, it was well known that Unity was hindered by sticking with an old and out-of-date Mono, and they were very successful at deflecting the blame for this by throwing the Mono folks under the bus. Any time complaints about game developers' inability to use newer C# features came up, Mono/Xamarin would invariably receive the ire* because Unity couldn't come to an agreement with them about their license and consulting fees. (Mono was open source under LGPL instead of MIT licensed at the time, and Unity had bought a commercial license that allowed them exempted from even the soft copyleft provisions in the LGPL.) Reportedly, they were trying to charge too much (whatever that means) for the exemptions they sought for the upgrade to the modern versions.
It's now 10+ years later, and now not only Mono (after being relicensed under MIT) but .NET CoreCLR are both available for Unity at no cost for a while, but despite this it still took Unity years before they'd upgraded the C# language support and upgraded to a slightly more modern runtime.
Something else to note: LGPL isn't inherently incompatible with commercial use or even use in closed source software. Some of the platforms Unity wished to be able to deploy to are incompatible, though, with the soft copyleft terms in the LGPL that require that users (or in this case game developers) be allowed to relink against other versions of the covered software (including, for example, newer releases). Perversely, it's because Unity sought and obtained exemptions to the LGPL that both end users and game developers were hamstrung in being able to upgrade Mono themselves! (It wouldn't have helped on, say, locked down platforms like Nintendo's for example, but certainly would have been viable on platforms with fewer restrictions, like PC gaming.)
By now, Unity has gone on to pull a lot of other shenanigans with their own pricing that seems to have sufficiently pissed off the game development community, but it should not be forgotten when they were willing to pass the blame to an open source project for the development/support that the company felt it was entitled to for a price lower than what they were willing to pay, and that they themselves were slow to make progress on even when the price of the exemption literally became $0.
* you can find threads with these sorts of comments from during this period right here on HN, too, but it was everywhere
So much this. According to a 2023 blog article from Unity [0], Unity uses Boehm GC. But Mono itself introduced another, generational GC called SGen [1] more than 10 years ago that became the default at some point. It is just Unity stuck on old mono versions, missing out on all the changes and improvements that went into Mono after their fork, essentially.
[0]: https://unity.com/blog/engine-platform/porting-unity-to-core... [1]: https://www.mono-project.com/docs/advanced/garbage-collector...
[1] - https://news.ycombinator.com/item?id=46415568
There are few things off with the this post that essentially sound as someone more green when it comes to Unity development (no problem, we all start somewhere).
1. The stated approach of separating the simulation and presentation layers isn't all that uncommon, in fact it was the primary way of achieving performance in the past (though, you usually used C++, not C#).
2. Most games don't ship on the mono backend, but instead on il2cpp (it's hard to gauge how feasible that'd be from this post as it lacks details).
3. In modern Unity, if you want to achieve performance, you'd be better off taking the approach of utilizing the burst compiler and HPC#, especially with what appears to be happening in the in sample here as the job system will help tremendously.
4. Profiling the editor is always a fools errand, it's so much slower than even a debug release build for obvious reasons.
Long story short, Unity devs are excited for the mentioned update, but it's for accessing modern language features, not particularly for any performance gains. Also, I've seen a lot of mention around GC through this comment section, and professional Unity projects tend to go out of their way to minimize these at runtime, or even sidestep entirely with unmanaged memory and DOTS.
> approach of separating the simulation and presentation layers isn't all that uncommon
I agree that some level of separation is is not that uncommon, but games usually depend on things from their respective engine, especially on things like datatypes (e.g. Vector3) or math libraries. The reason I mention that our game is unique in this way is that its non-rendering code does not depend on any Unity types or DLLs. And I think that is quite uncommon, especially for a game made in Unity.
> Most games don't ship on the mono backend, but instead on il2cpp
I think this really depends. If we take absolute numbers, roughly 20% of Unity games on Steam use IL2CPP [1]. Of course many simple games won't be using it so the sample is skewed is we want to measure "how many players play games with IL2CPP tech". But there are still many and higher perf of managed code would certainly have an impact.
We don't use IL2CPP because we use many features that are not compatible with it. For example DLC and mods loading at runtime via DLLs, reflection for custom serialization, things like [FieldOffset] for efficient struct packing and for GPU communication, etc.
Also, having managed code makes the game "hackabe". Some modders use IL injection to be able to hook to places where our APIs don't allow. This is good and bad, but so far this allowed modders to progress faster than we expected so it's a net positive.
> In modern Unity, if you want to achieve performance, you'd be better off taking the approach of utilizing the burst compiler and HPC#
Yeah, and I really wish we would not need to do that. Burst and HPC# are messy and add a lot of unnecessary complexity and artificial limitations.
The thing is, if Mono and .NET were both equally "slow", then sure, let's do some HPC# tricks to get high performance, but it is not! Modern .NET is fast, but Unity devs cannot take advantage of it, which is frustrating.
By the way, the final trace with parallel workers was just C#'s workers threads and thread pool.
> Profiling the editor is always a fools errand
Maybe, but we (devs) spend 99% of our time in the editor. And perf gains from editor usually translate to the Release build with very similar percentage gains (I know this is generally not true, but in my experience it is). We have done many significant optimizations before and measurements from the editor were always useful indicator.
What is not very useful is Unity's profiler, especially with "deep profile" enabled. It adds constant cost per method, highly exaggerating cost of small methods. So we have our own tracing system that does not do this.
> I've seen a lot of mention around GC through this comment section, and professional Unity projects tend to go out of their way to minimize these at runtime
Yes, minimizing allocations is key, but there are many cases where they are hard to avoid. Things like strings processing for UI generates a lot of garbage every frame. And there are APIs that simply don't have an allocation-free options. CoreCLR would allow to further cut down on allocations and have better APIs available.
Just the fact that the current GC is non-moving means that the memory consumption goes up over time due to fragmentation. We have had numerous reports of "memory" leaks where players report that after periodic load/quit-to-menu loops, memory consumption goes up over time.
Even if we got fast CoreCLR C# code execution, these issues would prevail, so improved CG would be the next on the list.
[1] https://steamdb.info/stats/releases/?tech=SDK.UnityIL2CPP
FieldOffset is supported by IL2CPP at compile time [0]. You can also install new DLLs and force the player to restart if you want downloadable mod support.
It's true that you can't do reflection for serialization, but there are better, more performant alternatives for that use case, in my experience.
[0] https://docs.unity3d.com/Manual/scripting-restrictions.html
Now I think about it, writing SourceGenerators is actually a great fit for AI agents.
I am not aware of an easy way to load (managed) mods as DLLs to IL2CPP-compiled game. I am thinking about `Assembly.LoadFrom("Mod.dll")`.
Can you elaborate how this is done?
> there are better, more performant alternatives for that use case, in my experience.
We actually use reflection to emit optimal code for generic serializers that avoid boxing and increase performance.
There may be alternatives, we explored things like FlatBuffers and their variants, but nothing came close to our system in terms of ease of use, versioning support, and performance.
If you have some suggestions, I'd be interested to see what options are out there for C#.
> FieldOffset is supported by IL2CPP at compile time
You are right, I miss-remembered this one, you cannot get it via reflection, but it works.
Ah, I was thinking native DLLs (which is what we're using on a project I'm working on). I think you're right that it's impossible for an IL2CPP-built player to interoperate with a managed (Mono) DLL.
>If you have some suggestions [re: serialization], I'd be interested to see what options are out there for C#.
We wrote a custom, garbage-free JSON serializer/deserializer that uses a fluent API style. We also explored a custom codegen solution (similar to FlatBuffers or protobuf) but abandoned it because the expected perf (and ergonomic) benefits would have been minor. The trickiest part with Unity codegen is generating code that creates little to no garbage.
The big advantage is that it reads data directly from game's classes, so there is no boilerplate needed, no prorobufs, no schema. And it supports versioning, adding or removing members mostly without limitations.
I think it's a cool system, maybe I should write a blog post about it :)
Per the separation, I think this was far more common both in older unity games, and also professional settings.
For games shipping on mono on steam, that statistic isn't surprising to me given the amount of indie games on there and Unity's prevalence in that environment. My post in general can be read in a professional setting (ie, career game devs). The IL injection is a totally reasonable consideration, but does (currently) lock you out of platforms where AoT is a requirement. You can also support mods/DLC via addressables, and there has been improvement of modding tools for il2cpp, however you're correct it's not nearly as easy.
Going to completely disagree that Burst and HPC# are unnecessary and messy. This is for a few reasons. The restrictions that HPC# enforce essentially are the same you already have if you want to write performant C# code as you just simply use Unity's allocators for your memory up front and then operate on those. Depending on how you do this, you either can eliminate your per frame allocations, or likely eliminate some of the fragmentation you were referring to. Modern .Net is fast, of course, but it's not burst compiled HPC# fast. There are so many things that the compiler and LLVM can do based on those assumptions. Agreed C# strings are always a pain if you actually need to interpolate things at runtime. We always try to avoid these as much as we can, and intern common ones.
The fragmentation you mention on after large operations is (in my experience) indicative of save/load systems, or possibly level init code that do tons of allocations causing that to froth up. That or tons of reflection stuff, which is also usually nono for runtime perf code. The memory profiler used to have a helpful fragmentation view for that, but Unity removed it unfortunately.
Sure, but the fact that it is competitive with Burst makes it disappointing. If I'm going to go through the trouble of writing code in a different (and not portable!) way then it better be significantly faster. Especially when most code cannot be written as Burst jobs unless you use their (new) ECS.
https://github.com/tbg10101/dotnet-burst-comparison
Making a managed code burst-compatible comes with real constraints that go beyond "write performant C#". In Burstable code, you generally can't interact with managed objects/GC-dependent APIs, so the design is pushed towards unmanaged structs in native collections. And this design spreads. The more logic is to be covered by Burst, the more things has to be broken down to native containers of unmanaged structs.
I agree that designing things in data-oriented way is good, but why to force this additional boundary and special types on devs instead of just letting them write it in C#? Writing burstable code can increase complexity, one has to manage memory/lifetimes, data layout, and job-friendly boundaries, copying data between native and managed collections, etc., not just "writing fast C#".
In a complex simulation game, my experience is that there are definitely things that fit the "raw data, batch processing" model, but not all gameplay/simulation logic does. Things like inheritance, events, graphs, AI (the dumb "game" version, no NN), UI, exceptions, etc. And on top of it all, debugging complications.
Wouldn't you be relieved with announcement: "C# is now as fast as Burst, have fun!"? You'd be able to do the same data-oriented design where necessary, but keep all the other tings handy standing by when needed. It's so close, yet, so far!
> The fragmentation you mention
What you say makes sense. I've actually spent a lot of time debugging this and I did find some "leaks" where references to "dead objects" were keeping them from being GC'd. But after sorting all these out, Unity's memory profiler was showing that "Empty Heap Space" was the culprit, that one kept increasing after every iteration. My running theory is that the heap is just more and more fragmented, and some static objects randomly scattered around it are keeping it from being shrunk. ¯\_(ツ)_/¯
For modding and OTA stuff I just use a scripting language with good interop (I made OneJS partially for this purpose). No more AOT issue and no more waiting for domain reload, etc.
From the article it seems that you're using some form of threading to create things, but you don't really specify which and/or how.
The default C# implementations are usually quite poor performance wise, so if you used for example the default thread pool I can definitively say that I've achieved a 3x speedup over that by using my own thread pool implementation which would yield about the same 30s -> 12s reduction.
Burst threading/scheduling in general is also a lot better than the standard one, in general if I feed it a logic heavy method (so no vectorization) then I can beat it by a bit, but not close to the 3x of the normal thread pool.
But then if your generation is number heavy (vs logic) then having used Burst you could probably drop that calculation time down to 2-3 seconds (in the same as if you used Vector<256> numerics).
Finally you touch on GC, that's definitely a problem. The Mono variant has been upgraded by them over time, but C# remains C# which was never meant for gaming. Even if we had access to the modern one there would still be issues with it. As with all the other C# libraries etc., they never considered gaming a target where what we want is extremely fast access/latency with no hiccups. C# in the business world doesn't really care if it loses 16ms (or 160ms) here and there due to garbage, it's usually not a problem there.
Coding in Unity means having to go over every instance of allocation outside of startup and eliminating them, you mention API's that still need to allocate which I've never run into myself. Again modern isn't going to simply make those go away.
To give some context, things are very complex in our game, we have fully dynamic terrain with terrain physics (land-slides), advanced path-finding of hundreds of vehicles (each entity has its own width and height clearance), trains, conveyors and pipes carrying tens or even hundreds of thousands of individual products, machines, rockets, ships, automated logistics, etc. There is no one thing that could be bursted to get 3x gain. At this point, we'd have to rewrite the entire game in C++.
So what's the reason we use C#? Productivity, ease of debugging and testing, and resilience to bugs (e.g. null dereference won't kill the program). Messing with C++ or even burst would cost us more time and to be honest, the game would possibly not even exist at that point.
Could you share some details about your custom thread pool that got 3x speedup? What was the speedup from? It is highly unlikely that a custom thread pool would have any significant impact on the benchmark in our case. As you can see from Figure 3, threaded tasks run for about 25% of the total time and even with Mono, all tasks are reasonably well balanced between threads. Threads utilization is surely over 90% (there is always slight inefficiency towards the end as threads are finishing up, but that's 100's of ms). An "oracle" thread pool could speed tings up by 10% of 25%, so that is not it.
Vectorization could help too but majority of the code is not easily vectorizable. It's all kinds of workloads, loading data, deserialization, initialization of entities, map generation, precomputation of various things. I highly doubt that automatic vectorization from code generated by IL2CPP would bring more than 20% speedup here. The speedup from burst would mostly come from elimination of inefficient code generated by Mono's JIT, not from vectorization.
For now, we are accepting the Mono tax to be more productive. But I am hoping that Unity will deliver on the CoreCLR dream. In the meantime, my post was meant raise awareness and stir up some discussion, like this one, which is great. I've read lots of interesting thoughts in this comments section.
Yeah, the thing with Burst is that its a lot easier to work with if you start with it than having to replace/upgrade code later, especially if you're not familiar with it. A big issue is usually that you create structs with data and they're referencing other structs etc., all those need to be untangled to really make use of Burst. I myself am also a big C# fan, it is a lot easier than using C. Unity has a lot of issues but there's a reason its so widely adopted and used. (I myself am currently working on a Unity C# tool that I believe will speed up code development significantly).
Your game does sound as if its a VERY ripe target for Burst usage based on the elements that you describe, but the real question should be if you need it. For example if you're already running at 60 fps on whatever your mid target hardware is at whatever max + N% load/size for a game instance then you don't need it. But if you're only hitting 40fps and design-wise want to increase e.g. your map size by 2x then it might be something to look into. Also if you look at e.g. Factorio, they spend a LOT of time optimizing systems, but of course you first need to launch the game (which is and should be the priority).
If you have for example 25 systems (e.g. pathfinding, trains, pipes, etc.) and they're evenly balanced then as you say then you won't increase your game speed by 2x by just converting one of those. BUT if for example your pipes are being processed in 4ms per frame, so you instead adopt other strategies like only processing them every Nth frame or doing M pipes per frame; at that point using Burst to just get that 4ms down to 0.5ms might be a really worthwhile target to make your game play better. The same goes for all your systems where the upgrade will have a cumulative effect.
I highly suggest learning just the basics of Burst in your spare time and trying it out on something basic to get the feel of it. As with all code/libraries it'll unfortunately take some time to figure out how to effectively use it. Roughly speaking: - You don't have to have SOA data, but it helps. At the start just convert methods over 1 to 1. - You have to convert most C# container types to Burst ones, for example in struct Vehicle { Wheel[] wheels } you need to change Wheel[] over to NativeArray<Wheel>, and the Wheel struct itself also need to not use complex types etc. Other types such as NativeSpan are also very useful, instead of storing the wheels just use a ref Span to them instead. - After you have basics going you can try out SOA along with more math/less logic so that the code can be vectorized, once you see that big speedup for certain types of code it's hard to go back.
>Could you share some details about your custom thread pool that got 3x speedup? What was the speedup from? It is highly unlikely that a custom thread pool would have any significant impact on the benchmark in our case. As you can see from Figure 3, threaded tasks run for about 25% of the total time and even with Mono, all tasks are reasonably well balanced between threads. Threads utilization is surely over 90% (there is always slight inefficiency towards the end as threads are finishing up, but that's 100's of ms). An "oracle" thread pool could speed tings up by 10% of 25%, so that is not it.
My thread pool itself is pretty standard, it spins up some heavy threads and uses ManualResetEvent to trigger them. Its advantage lies in pre-registering simple Action (with/without parameters) calls to set methods that'll be called when the thread runs; and with more gaming related options for whether we're waiting on thread completion, interleaving them with other threads etc. A big plus is that it has a self-optimization function, so it'll self-adjust the thread count vs the total time runs take, the total # of amounts of items being processed for the given workload etc. so as to automatically find very good sizes for all those elements to use for the target computer, vs just assuming e.g. 32, 64 or 128 inner elements and launching the max available threads on the PC (as thread pools usually do).
>Vectorization could help too but majority of the code is not easily vectorizable. It's all kinds of workloads, loading data, deserialization, initialization of entities, map generation, precomputation of various things. I highly doubt that automatic vectorization from code generated by IL2CPP would bring more than 20% speedup here. The speedup from burst would mostly come from elimination of inefficient code generated by Mono's JIT, not from vectorization.
Yeah, if its startup/generating code that's mostly bypassed by loading a game then its not worth switching over. Do note that code compiled by Burst will in general be more optimized than Mono just due to better tooling, but in general its not worth moving over just for that due to the amount of work you need to do so. The real wins come in if some generating element that's done often is taking too long, or during gameplay where you can replace elements in the game that take e.g. N milliseconds to calculate every frame and drop those down to 1/10th - 1/100th of the time it used to take.
Good luck!
Unity's whole shtick is that they make something horrible, then improve upon it marginally. The ground reality is that these performance enhancement schemes still fall very much short of just doing the basic sensible thing - using CoreCLR for most code, and writing C++ for the truly perf critical parts.
IL2Cpp is a horror-kludge of generated code, that generates low-quality C++ code from .NET IL, relying on the opitmizing compiler to extract decent performance out of it.
You can check it out: https://unity.com/blog/engine-platform/il2cpp-internals-a-to...
The resulting code gives up every possible convenience of C# (compile speed, convenience, debuggability), while falling well short of even modern .NET on performance.
The Burst compiler/HPC# plays on every meme perpetuated by modern gamedev culture (structure-of-arrays, ECS), but performance wise, generally still falls short of competently, but naively written C++ or even sometimes .NET C#. (Though tbf, most naive CoreCLR C# code is like 70-80% the speed of hyper-optimized Burst)
These technologies needless to say, are entirely proprietary, and require you to architect your code entirely their paradigms, use proprietary non-free libraries that make it unusable outside unity, and other nasty side effects.
This whole snakeoil salesmanship is enabled by these cooked Unity benchmarks that always compare performance to the (very slow) baseline Mono, not modern C# or C++ compilers.
These are well-established facts, benchmarked time and time again, but Unity marketing somehow still manages to spread the narrative of their special sauce compilers somehow being technically superior.
But it seems the truth has been catching up to them, and even they realized they have to embrace CoreCLR - which is coming soonTM in Unity. I think it's going to be a fun conversation when people realize that their regular Unity code using CoreCLR runs just as fast or faster than the kludgey stuff they spent 3 times as much time writing, that Unity has been pushing for more than a decade as the future of the engine.
Efforts like Managed DirectX and XNA were driven by highly motivated individuals, and were quickly killed as soon as those individuals changed role.
One could blame them for leaving the project, or see that without them managemenent did not care enough to keep them going.
While at the same time, since Unity relies on such alternative approaches, it also creates a false perception on how good .NET and C# are in reality, for those devs that never learned C# outside Unity.
In a similar way it is like those devs that have learnt Java in Android, and get sold on the Kotlin vs Java marketing from Google, by taking Android Java as their perception of what it is all about.
Going back to game development and .NET, at least Capcom has the resources to have their own fork of modern .NET, e.g. Devil May Cry for the Playstation was done with it.
"RE:2023 C# 8.0 / .NET Support for Game Code, and the Future"
https://www.youtube.com/watch?v=tDUY90yIC7U
XNA was very influential for me as well - when I was in high school, I tried to get into 3D game dev, and I started with Frank. D Luna's otherwise excellent book on DirectX gamedev - man that thing was a tome. However, having to learn DirectX, C++, linear algebra, shaders, WIN32 API, COM etc. at the same time (which to be fair were explained very thoroughly by the book), was just too much for me back then, not to mention the absolute pain of trying to get models and assets in the game.
Later on I discovered XNA, and it was a breath of fresh air for me - a much easier language, good IDE support and a decent way of importing assets, and an much nicer API made it so much easier to get started.
And the truly great thing about it was that it didn't dumb things down or hide stuff from the developer - it merely provided sane defaults, and utility functions so that you didn't have to engage with all that complexity at once.
I think Unity was also great as well, at least in the beginning (the first decade of existence), but it's chief issue is that Unity's 'dialect' of C# was very different from how you programmed in regular C# (or mostly any other engine) - my feeling is that Unity should've spun their own language/runtime rather than trying to make C# into what it wasn't designed to be.
https://en.wikipedia.org/wiki/Boo_(programming_language)
I would argue that C# has always been a good alternative for games, starting with Arena Wars, the problem was Microsoft not being serious about AOT or low level programming, because that was left for C++/CLI.
https://en.wikipedia.org/wiki/Arena_Wars
Here is the person responsible for pushing XNA, even though management wasn't into it.
"The billion dollar decision that launched XNA"
https://youtu.be/wJY8RhPHmUQ?si=_3pic4pEiOlqQzvm
When she left Microsoft, XNA was promptly replaced by DirectXTK, because C++ is the only true way for DirectX team,
https://walbourn.github.io/directxtk/
But then I remember they have to deal with IL2CPP, because lots of mobile/console platforms do not allow JIT as policy.
.NET does now have 'full AOT' as a thing at least, and Streets of Rage 4 I believe used CoreRT for at least a one of the platforms it was released for.
What's more hopeful about that, is that you can see backers of many different architectures contributing towards the ecosystem.
It’s not saying much that everything has tradeoffs. During the “decade” you are talking about, CoreCLR didn’t have a solution for writing anything for iOS, and today, it isn’t a solution for writing games for iOS. What you are calling kludges was ultimately a very creative solution. Usually the “right” solution, the nonexistent one that you are advocating with, ends with Apple saying no.
That is why Unity is a valuable piece of software and a big company: not because of C# runtimes, but because they get Apple and Nintendo to say yes in a world where they usually say no.
During the 'decade' where CoreCLR was not a solution, Mono (Xamarin) still was - in fact their entire commercial appeal (before they were bought out by Microsoft) was that they provided an AOT compiled .NET for mobile devices.
Unity got stuck on an ancient version compared to the more modern Mono implementations (I think this is the case to this day), and Unity's version was much, much slower.
Afair, most of the time, the MS version had them (Xamarin) beat, but the difference wasn't huge, between the two, especially compared to Unity's mono. It was an AOT runtime, not sure about Nintendo, but their entire business model hinged on being able to ship to Apple's app store.
I hate to dig up the past, but Unity's long-standing issue was their ancient GC (which was not incremental then), combined with a simple compiler bug, that made every foreach loop allocate an iterator on the heap. The combination of the two meant that basically every non-trial Unity game that used foreach extensively, stuttered. This simple compiler fix took them years to upstream, with people hacking around the issue by replacing the shipped compiler with a patched one.
And I still stand by my point - if Unity went with an upstream Mono, and made it convenient and easy to link with C++ code, it'd have had the same or better performance out of the box as they had with their exotic stuff.
And I also back up the fact that Unity's marketing was deceptive - HPC#/Burst/DOTS/ECS was marketed as some gateway to previously unheard of performance, but when some people went around benchmarking it, basic sensible C++ had it beat (I can't find the benchmarks rn, but multithreaded DOTS was slower than single threaded c++ in a simulation).
What I said about Burst holds up as well.
These are not tradeoffs, but bad technical deicisions, whose legitimacy can be only defended when you ignore the most sensible alternatives.
I think the biggest hurdle a unity contender has to overcome, is how to provide both 'similar enough' primitives as well, as a way to easily consistently handle the graphics/sound pipeline (i.e. simple way to handle different platform bindings) while also making sure all of that infrastructure is AOT friendly.
see, i'm saying, i like to construct games, and so do a ton of other people, and they ignore all the stuff you are talking about all the time. it's not just me. it truly doesn't matter. this is because iOS and the Switch are already such buggy systems, that if you get worked up about any particular performance bug or IT flaw, you will never come up with good game ideas, let alone ship.
C++ code is much faster than C#, but modern C# has become a lot better with all the time that's been invested into it. But you can't just take a random bit of C code and think that it's going to be better than an optimized bit of C#, those days are long past.
Secondly, the whole point of Burst is that it enables vectorization, which means that if you've converted code to it and it's used properly that its going to support instructions up to 256 wide (from what I remember it doesn't use AVX512). That means that it's going to be significantly faster than standard C# (and C).
If the author is generating for example maps and it takes 80 seconds with Mono, then getting to between 10-30 seconds with Burst is easy to achieve just due to its thread usage. Once you then add in focused optimizations that make use of vectorization you can get that down to probably 4 odd seconds (the actual numbers really depend on what you're doing, if its a numerical calculation you can easily get to 80x improvement, but if there's a lot of logic being applied then you'll be stuck at e.g. 8x.
For the last point, new modern C# can't just magically apply vectorization everywhere, because developers intersperse far too much logic. It has a lot of libraries etc. that have become a lot more performant, but again you can't compare that directly to Burst. To compare to Burst you have to do a comparison with Numerics, etc.
CoreCLR NativeAOT is already shipping real games on Nintendo, PS5, and Xbox.
JIT isn't allowed on iPhones either, and this is what NativeAOT solves. Also, .NET is moving WASM support to CoreCLR (rather than mono) in an upcoming version as well.
"RE:2023 C# 8.0 / .NET Support for Game Code, and the Future"
https://www.youtube.com/watch?v=tDUY90yIC7U
As always, it is a matter of having the skill to deliver, instead of GC phobia.
I would have estimated a year, or two tops, for that project.
Hard task, no doubt. Unity needs to throw everything at this problem. C# in general has gotten insanely fast by default. It's very much worth taking the time to upgrade/update.
Whilst we don't compare in size and api surface, it took us a few months to get off 472 and onto dotnet6. But once we were on dotnet6, moving to the LTS after that was relatively painless; usually a few hours of work.
17 more comments available on Hacker News