In regards to disasmo: I’ve recently also used / tried to use disasmo, however, after discovering later that visual studio natively supports showing the actual asm of the code while debugging, I realized that the actual asm of visual studio and the asm of disasmo differed wildly. Now, because of that I only used disasmo very briefly and so It could be that I simply used that tool incorrectly, however that was something I wanted to point out as visual studio did give me asm that looked a lot more like the asm I was expecting.
The asm Visual Studio gives you during debugging is the Debug codegen with all optimizations disabled, and it also has redundant code for better debugging experience so that you can break at any line of your source code.
While disasmo gives you optimized code which is the one you will actually run in production environment.
That might be the case.
However i noticed I failed to specify that I was specifically trying to inspect the (auto) vectorized instructions and manually vectorized instructions at the time, and disasmo just wouldn’t emit the expected code in neither case, where visual studio actually did; and the asm between the two was not comparable at all.
It’s been a few weeks now since then, I don’t super remember everything. And like I said, I might’ve just used disamo wrong. Simply wanted to point out some struggles I had with it.
4 posts were split to a new topic: Advice regarding C# library and Tensors
Don’t hijack this thread for off-topic C# questions.
Thanks for the answer. Seems like we’re sticking with using NativeArray as our “span type” then, as it has the most utility in best interaction with burst, and can be converted to span.
So are there any news you can share with us? There is such a silence here
Since the switch to the new forums, there is seemingly less activity in this thread.
Mate, this thread was put up after the move to the new forums.
To be fair I think what they can show would be kept for surprise unite 2025 (April right?)
But if unity 7 still fail to be declared as alpha roadmap in this year it mean the coreclr was totally hopeless in my opinion
Why? They’ve said months ago that they already had it working in the editor natively, but that it would take some time to ship all of Unity’s packages over. The biggest thing they’re stuck on right now besides that, from what I understand, is replacing asmdef with csproj.
I can see a world where CoreCLR is released supporting asmdef in the worst case scenario, but not one where it is outright discarded. Unity has always been slow with publicly releasing Editor-integrations.
But, fair’s fair: Josh Peterson’s estimate of 2 years was, let’s say… optimistic
Exactly this reason that they already have closed alpha internally. If they still cannot announce public alpha I can only think that roadblock issue are enormously hopeless to be cleared anytime soon
Quick question. Will CoreCLR support android/Ios/WebGL/WebGPU build targets?
CoreCLR will replace Mono. Platforms that cannot use JIT will continue to use IL2CPP (iOS, WebGL).
Will IL2CPP be updated to support the new .net 6/8 compatibility and language increases? Or would you effectively have to create an entire other game for IL2CPP?
IL2CPP will be updated to support all new APIs and language features.
Wait people still actually develop and ship desktop builds using mono instead of il2cpp by choice? (As opposed to the case of a long lived codebase that was first built for mono and never moved). Really? Why?
(post deleted by author)
Some games are still a bit faster with Mono in some rare cases. For example, I think some heavy uses of generics and interfaces, especially covariant/contravariant, can still be faster on Mono/JIT. Mono can be faster to build too, which can be nice when testing on devices.
Other reason may be that IL is friendlier to some tools for modding and patching than assembly. You can use things like Cecil with Mono builds.
Thanks for explaining. I knew about the patching angle, but assumed it applied to very few games (and in fact that most games want their generated code to be harder rather than easier to inspect.) Also, iteration time for debug completely agree (we use it too), but I was referring to final builds.
Had no idea about the generics/interfaces performance advantage for Mono. Thanks.