Is Reflection really slow? - c#

This is a common belief that reflection is slow and try to avoid it as much as possible. But is that belief true, in the current situation? There has been lot of changes in the current .net versions like, use of IL Weaving (i.e. IL Emit) etc, as opposed to traditional PropertyInfo and MethodInfo ways of performing reflection.
Is that any convincing proof, that the new reflection is not that slow any more, and can be used. Is there better way to read attribute data?
Thanks,
Bhaskar

When you think about it, reflection is pretty darn impressive in how fast it is.
A cached delegate from a ConstructorInfo or MethodInfo can be called with speed comparable to any other delegate.
A delegate created from ILGenerator.Emit (which incidentally isn't new, it's been in .NET since version 1) can likewise be called just as fast.
An object obtained through emitting or calling a ConstructorInfo's delegate will be just as fast as any other object.
If you obtain an object by loading an assembly dynamically, using reflection to find the method to call, and calling it, and it implements a defined interface through which you call it from that point on, then it'll be just as fast in how it's used as another implementation of that interface.
In all, reflection gives us ways of doing things that without it we would - if we could do them at all - have to use techniques that are slower both to code and to execute.
It also gives us means of doing things that are more complicated, more brittle, less type-safe and less performant than other means too. Pretty much every line of C# code can be replaced by a big chunk of code that uses reflection. This code will almost certainly be worse than the original line in a whole bunch of ways, and performance is the least of them.
Quite possibly the "avoid reflection because its slow" advice stems from the belief that the sort of developer that would go nuts for any new technique just because it seemed cool would be the sort that would be more likely warned off by "it'll be slower" than by "it'll be less idiomatic, more error-prone and harder to maintain". Quite possibly this belief is completely correct.
For the most part though, when the most natural and obvious approach is to use reflection, then it also won't be less performant than a really convoluted attempt to avoid it.
If performance concerns apply to anything in reflection, its really to the uses that are hidden:
Using dynamic can seem sensible in a case where only a little work could avoid it. Here the performance difference may be worth considering.
In ASP.NET, using <%#DataBinder.Eval(Container.DataItem, "SomeProperty")%> is easier but generally less performant than <#((SomeType)Container.DataItem).SomeProperty%> or <%#SomeCodeBehindProvidedCallWithTheSameResult%>. I'll still use the former 90% of the time, and the latter only if I really care about a given page's performance or more likely because doing many operations on the same object makes the latter actually more natural.
So in all, everything remains "slow" in computers while they are bounded by the requirement to work in a single universe in a way that consumes energy and takes time, for some value of "slow". Different reflection techniques have different costs, but so does the alternatives. Beware not so much of reflection in cases where it's the obvious approach as hidden reflection where another approach just slightly less obvious may serve well.
And of course, code wisely with whatever technique you use. If you're going to call the same delegate a hundred times in a row, you should be storing it rather than obtaining it each call, and that would go whether the way to obtain it was through reflection or not.

It all depends.
Yes, using reflection is without any doubt slower than not using reflection, but you have to look at the big picture:
For example, imagine that your code does some reflection and then loads some data from a database over the network.
In this case, you can completely neglect the additional cost for reflection because the database request over the network will likely take much, much longer.

I wouldn't worry about the performance cost that much. Not saying performance doesn't matter, but in a lot of cases, a millisecond or two isn't noticeable or worth choosing a more complicated coding alternative. Take a look here.
I generally prefer this approach:
write simple code
measure performance
optimize the biggest hitter, if needed

Whad do you mean by mentioning a common belief? Reflection is not a solid block. You should consider each method apart. For instance, creating an object through a default constructor is few times slower than a simple call, whilst creating a parametered constructor is tens time slower. So if you want to study the speed, do a benchmark and benchmark the concrete functions you need.
PS. Using C#, you can always create and compile expressions on the fly which if you manage to do it would be much faster than reflection.

yes reflection is slow when we try to do on bulky opertations or with in a loop.
you can try out dynamic options as well . with dynamic option its can be cached and eventually will be faster than reflection.

If you need to look at attributes applied to source code, then you pretty much have to use reflection.
Some reflection operations are fast, such as Object.GetType(), but some operations are relatively slow, such as Type.FindMethod("MyMethod").
In general, I would say that if your application makes occasional use of Reflection, there should be no performance concern. On the other hand, if your application uses Reflection extensively, then you might observe some slowness.

Introspection is an heavy job. Saying slow is relative to a lot of things. Invoking method/constructor through reflection is slow, but using reflection to only retreive metadatas is not.
Keep in mind, reflection must be only use to retrieve metadatas.
If you need to invoke method, running something, just emit dynamic types/methods at initialization time and invoke them through interfaces/delegates.

Related

Is it possible to use a custom memory allocator for LINQ?

Is there a way to use a custom memory allocator for LINQ?
For example when I call:
someCollection.Where(x).SelectMany(y).ToList();
Methods like ToList() or OrderBy() will always create a new array, so lots of GC will happen.
With a custom allocator, I could always use the same List, which will be cleared and refilled every time. Iam aware that reusing buffers could lead to problems with reentrancy.
The background is, my application is a game and GC means stuttering.
Please don't tell me "Use C++ instead" or "Do not use LINQ", I know that :)
(Although you asked not to be suggested against it, I thin this answer could help the community)
LINQ is a facility built on top the CLR, therefore it uses the CLR allocator, and it cannot be changed.
You can tune it a little bit, for example configuring whether or not the GC cycle should be offloaded to a background thread, but you can't go any further.
The aim of LINQ is to simply writing code for certain class of problems sacrificing the freedom to choose the implementation of every building block (that's why we usually choose LINQ).
However, depending on the scenario, LINQ could not be your best friend as its design choices may play against yours.
If, after profiling your code, you identify that you have a serious performance problems you should try at first to identify whether or not you can isolate the bottleneck in some of LINQ methods and see whether you can roll your own implementation, via extension methods.
Of course this option is viable when yuo are the main caller, unless you manage to roll something that is IEnumerable complaint. You need to be very lucky, because your implementation should abide to LINQ rules. Particularly, as you are not in control of how the objects are manipulated, you cannot perform the optimizations you would in your own code.
Closures and deferred execution work against you.
Otherwise, what has been suggested by the comments, is the only viable option: avoid using LINQ for that specific task.
The reason for stepping away from LINQ is that it is not the right tool to solve your problem with performance constraint you require.
Additionally, as stated in the comments, the (ab)use of lambda expressions significantly increase the memory pressure as backing objects are created to implement the closures.
We had performance issues similar to yours, where we had to rewrite certain slow paths. In other (rare) cases, preallocating the lists and loading the results via AddRange helped.

Specification Pattern and Performance

I've been playing around w/ the specification pattern to handle and contain the business logic in our c#/mvc application. So far so good. I do have a question though - since we'll be creating a number of specification objects on the heap, will that affect performance in any way versus, say creating helper methods to handle the business logic? Thanks!
I do have a question though - since we'll be creating a number of specification objects on the heap, will that affect performance in any way versus, say creating helper methods to handle the business logic?
Of course it will affect performance, every line of code you write and design choice you makes affects performance in one way or another. This one is unlikely to be meaningful, be a bottleneck in your application or be worth caring about as this is almost surely a case of premature optimization. These days you should just focus on modeling your domain properly, and writing extremely clear and maintainable code. Focus more on developer productivity than on machine productivity. CPU cycles are cheap, and in nearly limitless supply. Developer cycles are not cheap, and are not limitless in supply.
But only you can know if it will impact the real-world use of your application on real-world data by profiling. We don't, and can't know, because we don't know your domain, don't know your users, don't know what performance you expect, etc. And even if we knew those things, we still couldn't give you as powerful of an answer as you can give yourself by dusting a profiler off the shelf and seeing what your application actually does.
since we'll be creating a number of specification objects on the heap, will that affect performance in any way
Most design patterns trade off some overhead for cleanliness of design - this is no exception. In general, the amount of memory that the specifications add is very minimal (typically a couple of references, and that's it). In addition, they tend to add a couple of extra method calls vs. custom logic.
That being said, I would not try to prematurely optimize this. The overhead here is incredibly small, so I would highly doubt it would be noticeable in any real world application.
If you use NSpecifications lib just as the examples in its GitHub page, you'll get the benefits from both worlds:
Most of these specifications are simply stored in static members therefore it doesn't take much from the heap
These specifications also use compiled expressions so that they can be reused many times with better performance
If you are using ORM to query the database with lambda expressions, that also uses the heap, the difference here is that NSpecifications stores those expressions inside a Spec object so that it can be reused for both business loginc and querying.
Check here
https://github.com/jnicolau/NSpecifications

Long lists of pass-by-ref parameters versus wrapper types

I need to get three objects out of a function, my instinct is to create a new type to return the three refs. Or if the refs were the same type I could use an array. However pass-by-ref is easier:
private void Mutate_AddNode_GetGenes(ref NeuronGene newNeuronGene, ref ConnectionGene newConnectionGene1, ref ConnectionGene newConnectionGene2)
{
}
There's obviously nothing wrong with this but I hesitate to use this approach, mostly I think for reasons of aesthetics and psycholgical bias. Are there actually any good reasons to use one of these approaches over the others? Perhaps a performance issue with creating extra wrapper objects or pushing parameters onto the stack. Note that in my particular case this is CPU intensive code. CPU cycles matter.
Is there a more elegant C#2 of C#3 approach?
Thanks.
For almost all computing problems, you will not notice the CPU difference. Since your sample code has the word "Gene" in it, you may actually fall into the rare category of code that would notice.
Creating and destroying objects just to wrap other objects would cost a bit of performance (they need to be created and garbage collected after all).
Aesthetically I would not create an object just to group unrelated objects, but if they logically belong together it is perfectly fine to define a containing object.
If you're worrying about the performance of a wrapping type (which is a lot cleaner, IMHO), you should use a struct. Current 32-bits implementations of .NET (and the upcomming 64-bits 4.0) support inlining / optimizing away of structs in many cases, so you'd probably see no performance difference whatsoever between a struct and ref arguments.
Worrying about the relative execution speed of those two options is probably a premature optimization. Focus on getting the algorithm correct first, and having clean, maintainable code. When that's done, you can run a profiler on it and optimize the 20% of the code that takes 80% of the CPU time. Even if this method ends up being in that 20%, the difference between the two calling styles is probably to small to register.
So, performance issues aside, I'd probably use a container class. Since this method takes only those three parameters, and (presumably) modifies each one, it sounds like it would make sense to have it as a method of the container class, with three member variables instead of ref parameters.

Is it possible to have too many anonymous delegates?

Lately, I have taken to the pattern of having a lot of diagnostic logging in parts of my code, that makes use of lambda expressions/anonymous delegates like so:
MyEventManager.LogVerbose( LogCategory.SomeCategory, () => String.Format(msg_string, GetParam1(), GetParam2(), GetParam3() );
Notice that the second argument to LogVerbose is a lambda expression which evaluates to a string. The reason for this is that if verbose logging is not actually enabled, LogVerbose should exit having done as little work as possible, in order to minimize performance impact. The construction of the error message string may, in some cases, take time or resources, and if the lambda expression is never evaluated, that performance penalty will not be incurred.
I'm wondering if littering the type system with so many anonymous delegates like this will have some unforeseen consequence for application performance, or if there are any other strategies I should consider.
It should be fine. In particular, if your anonymous function doesn't capture anything, it is cached as a static field (because it can be). If you capture "this" then you'll end up creating new delegate instances, but they're not expensive.
If you capture local variables, that will involve instantiating a nested type - but I'd only worry about this if you saw it actually becoming a problem. As ever with optimisation, focus on readability first, measure the performance, and then profile it to find out where you need to concentrate your efforts.
Whilst I don't actually know for sure the answer to the question, I think its worth considering that a drive to a more functional style of programming in C# would be seriously undermined if there were any suggestion that there would some kind of limit on the use of such expressions.
I've got a solution that's got thousands of anon delegates, and it still works. Sometimes Visual Studio is a little clunky, but whether that's because we've got hundreds of projects or this or some other factor is unknown. The performance of the applications don't seem to be that strongly affected (with quite a bit of perf testing).

C#, Writing longer method or shorter method?

I am getting two contradicting views on this. Some source says there should be less little methods to reduce method calls, but some other source says writing shorter method is good for letting the JIT to do the optimization.
So, which side is correct?
The overhead of actually making the method call is inconsequentially small in most every case. You never need to worry about it unless you can clearly identify a problem down the road that requires revisiting the issue (you won't).
It's far more important that your code is simple, readable, modular, maintainable, and modifiable. Methods should do one thing, one thing only and delegate sub-things to other routines. This means your methods should be as short as they can possibly be, but not any shorter. You will see far more performance benefits by having code that is less prone to error and bugs because it is simple, than by trying to outsmart the compiler or the runtime.
The source that says methods should be long is wrong, on many levels.
None, you should have relatively short method to achieve readability.
There is no one simple rule about function size. The guideline should be a function should do 'one thing'. That's a little vague but becomes easier with experience. Small functions generally lead to readability. Big ones are occasionally necessary.
Worrying about the overhead of method calls is premature optimization.
As always, it's about finding a good balance. The most important thing is that the method does one thing only. Longer methods tend to do more than one thing.
The best single criterion to guide you in sizing methods is to keep them well-testable. If you can (and actually DO!-) thoroughly unit-test every single method, your code is likely to be quite good; if you skimp on testing, your code is likely to be, at best, mediocre. If a method is difficult to test thoroughly, then that method is likely to be "too big" -- trying to do too many things, and therefore also harder to read and maintain (as well as badly-tested and therefore a likely haven for bugs).
First of all, you should definitely not be micro-optimizing the performance on the number-of-methods level. You will most likely not get any measurable performance benefit. Only if you have some method that are being called in a tight loop millions of times, it might be an idea - but don't begin optimizing on that before you need it.
You should stick to short concise methods, that does one thing, that makes the intent of the method clear. This will give you easier-to-read code, that is easier to understand and promotes code reuse.
The most important cost to consider when writing code is maintanability. You will spend much, much more time maintaining an application and fixing bugs than you ever will fixing performance problems.
In this case the almost certainly insignificant cost of calling a method is incredibly small when compared to the cost of maintaining a large unwieldy method. Small concise methods are easier to maintain and comprehend. Additionally the cost of calling the method almost certainly will not have a significant performance impact on your application. And if it does, you can only assertain that by using a profiler. Developers are notoriously bad at identifying performance problems before hand.
Generally speaking, once a performance problem is identified, they are easy to fix. Making a method or more importantly a code base, maintainable is a much higher cost.
Personally, I am not afraid of long methods as long as the person writing them writes them well (every piece of sub-task separated by 2 newlines and a nice comment preceeding it, etc. Also, identation is very important.).
In fact, many times I even prefer them (e.g. when writing code that does things in a specific order with sequential logic).
Also, I really don't understand why breaking a long method into 100 pieces will improve readablility (as others suggest). Only the opposite. You will only end-up jumping all over the place and holding pieces of code in your memory just to get a complete picture of what is going on in your code. Combine that with possible lack of comments, bad function names, many similar function names and you have the perfect recipe for chaos.
Also, you could go the other end while trying to reduce the size of the methods: to create MANY classes and MANY functions each of which may take MANY parameters. I don't think this improves readability either (especially for a begginer to a project that has no clue what each class/method do).
And the demand that "a function should do 'one thing'" is very subjective. 'One thing' may be increasing a variable by one up to doing a ton of work supposedly for the 'same thing'.
My rule is only reuseability:
The same code should not appear many times in many places. If this is the case you need a new function.
All the rest is just philosophical talk.
In a question of "why do you make your methods so big" I reply, "why not if the code is simple?".

Categories

Resources