Node to Code from Blueprint Editor

AngelWolf · June 3, 2015, 9:41pm

Hi all. I’ve been tossing around ideas about automating the conversion of BP into C++. While others seem to be taking the approach of converting an entire blueprint, perhaps all we really need is a way to convert a series of nodes into a single line or chunk of C++ code. What I propose is a feature that allows you to select a node or series of nodes, right click, and click on a new menu item called Node(s) to Code. The nodes and connections between them are then analyzed and converted into a line or lines of formatted C++ code on the clipboard, where it can be pasted in to visual studio and adjusted. This would save lots of time typing out these lines by hand.

Shadowriver · June 3, 2015, 11:09pm

But you would need to pick nodes that have single common orgin in order for them to create code sequence

KVogler · June 4, 2015, 7:13pm

Unless I got the concept wrong, but isnt that whats happening when you “compile” a BP ?

Martin_Bean · June 4, 2015, 7:31pm

As far as I know: no

compiling a blueprint results in bytecode. This bytecode is executed. You can read more about bytecode here: Bytecode - Wikipedia

There are all kinds of optimization techniques one can implement to make the execution very fast. Java and .Net are quite good at this. I dont know what optimizations Epic has implemented. I dont know how it was measured, but a while a go I read somewhere that a blueprint is around 10-20 percent slower. But I guess it also depends on what you are doing in the blueprint.

JaredTherriault · June 5, 2015, 2:42pm

Blueprint is slower than that. C++ is up to 1000% faster than Blueprint. The biggest slowdown though, is the transition between the VM layer (bytecode) and the native (C++) calls. So the fewer nodes you have, the faster your code. Collapsing nodes to functions of course just moves those nodes to a new graph - to see gains you’d have to rewrite those nodes into a single C++ function. For the most part though, speed differences are imperceptible. Only when you’re stringing together many complex nodes will you see marked slowdown, at least in my experience.

Martin_Bean · June 5, 2015, 2:59pm

You are correct, I looked it up… it wasnt 10 percent slower but 10 times slower. Bit of a difference

'[Epic:

Billy Bramer;19464’]
Another thing I wanted to touch on specifically was also blueprint performance vs. C++, because that came up in the stream a few times and I don’t think we did a fantastic job answering that particular part (Cameron gets over-excited, what can I say? :p) The answer here ends up a little bit complicated. The simple version of the answer is that native C++ outperforms blueprints. Cameron was correct in that if you see a blueprint-exposed function you can call on an actor, etc., it’s really just calling back to the C++ code written for it, however, there is an overhead cost associated with executing blueprints that isn’t present with purely native code. As engine programmer James Golding mentions in this thread: Blueprint Performance Benchmark? - Blueprint - Epic Developer Community Forums , if we consider blueprints as similar to UE3’s old UnrealScript, it would be a rough rule-of-thumb of around a 10x difference, but that could vary considerably depending on the scenario. Ultimately if you’re only using a few nodes that are calling back into C++ functionality that is handling most of the heavy lifting of an operation, that’d likely end up faster than that estimate. The estimate also initially sounds large, but keep in mind that there is a very big difference between “slower than C++” and “too slow to be used.” Hundreds of UE3 games (including all of Epic’s titles) were written with extensive amounts of UnrealScript handling the gameplay logic. The vast majority of cases you’d want to use blueprints in will be fine, even if they aren’t exactly the speed of C++. Fortnite has numerous uses of blueprints all over the place. The parts of the game that are super performance critical tend to already be in C++, so you shouldn’t feel worried about putting gameplay logic in blueprints. The areas where you do want to be careful are the same as if you were writing in C++. Just like you wouldn’t want to run really performance-intensive code inside every C++ actor’s tick functions, you also shouldn’t put tons of performance crazy operations inside the Tick event of a blueprint. We tend to make a lot of our blueprint logic event-based (triggered when some event happens, like taking damage, etc.) instead of trying to do everything every frame in a tick function.

source:

Conkerballs · August 15, 2015, 4:32pm

I want to apologize for this Necro however I believe this would be an amazing feature. I have the impossible in mind like you right click a BP file and hit ‘Generate C++ Script’ takes some time to convert it. Open it in VS and learn what you just did in Blueprints.

This would be an amazing feature for the visual scriptwriters to have that same power and optimization but also, the best learning tool out there for C++ people would go absolute nuts for this. Hell I would pay a fair chunk of change for a plugin that had this functionality.

I understand that somethings aren’t possible tho however this shouldn’t be buried would be nice to get an Official Yes this is possible and we will aim to achieve it or simply sorry this is not possible.

Apologies for necro.

KVogler · August 16, 2015, 1:35am

I would love to do it, but Im still in the lower regions of the C++ learning curve…
If someone is willing to help me with some C++ related “issues” (mostly syntax and UE4-api related), Ill give it a shot
Writing parsers is my hobby

Hevedy · August 16, 2015, 12:08pm

There is code in the 4.8 and 4.9 about new system for Blueprint and probably for the 4.10 there should be the new compiler the Blueprint will be converted and compiled in C++ for the release versions of the projects and that boost the performance and speed of the Blueprints to the same or more or less the same of your C++ code.

Conkerballs · August 16, 2015, 1:04pm

That would be amazing…

I would literally be so happy with this, it means there is even less bad implications from writing all code in Blueprint. I don’t write in BP because i’m to lazy to learn C++ or it’s to hard. Its just very fun and I am a visual learner. But I would also like the feature to scan through the created C++ files so if you wanted to learn C++ from this you could say “Ohh that’s how to write a function in C++” etc… However I would be happy with just compiling to C++ for that performance boost.

Its a shame there is no mention of this yet… Would be something to get quite excited about.

Nate · August 17, 2015, 9:57am

I did a simple map spawning 800 flying missiles using AlFlakky’s Homing missile blueprint and converted C++ version to find out how much speed difference between Blueprint vs C++, here’s my finding:

In 4.8.3

Blueprint: 36 fps
C++: 42fps

In 4.9.x (latest github 4.9 branch)

Blueprint: 38 fps
C++: 41fps

In 4.10.x (latest github master branch)

Blueprint: 42 fps
C++: 42fps

Well, I always assume the “10x slower than C++” blueprint performance too until I did the test. It’s a simple level with only 1 blueprint so don’t take the speed comparison too seriously, but it proves that “10x slower” is indeed **rough **rule-of-thumb depending on lots of factors.

anonymous_user_1ca92533 · August 17, 2015, 11:17am

They say its 10x slower but it shouldn’t be as i’d assume that it is writing the script and placing each piece of code as your placing your nodes in real-time. E.g Bool.

I think if its working like that if then the only thing that would be slow would in fact be how fast your placing each node.

Conkerballs · August 18, 2015, 12:31am

Thanks for the numbers there Nate.

Would be nice to get a real stress test doing the same exercise you did.

Something really quite challenging like worst case scenario.

Still looks to be pretty awesome.