SignalX - The fastest signal for Connect and Fire today,

[bro_proxiomFun (Proxiom)]

although it uses more memory than the rest of the signals

Creator Hub - Store

Hello Devs!

I would like to introduce you to SignalX, my new signal system implementation designed specifically for maximum performance in Roblox. I’ve working on optimizing this module for a long time(4 days lol), and Im get very good resuls, so I want to share them with you.

What is SignalX

SignalX is a high–performance replacement for standard Roblox signals, designed for use in performance-critical systems. It uses

• Direct memory access: Minimal overhead due to the use of buffers to store connection information.
• Bit operations: Efficient memory usage and fast flag operations.

---

• Thread Pool: Reused streams to minimize delays in triggering signals.
  (This is what makes my code more memory-demanding, but in my opinion, speed is important for productive systems, not “oh no, it took an extra 10 megabytes on the server.”)

---

What is it for?

• Games with intensive input processing (for example, first-person shooters).
• Any systems where every millisecond counts.

Benchmarks! (The most interesting part)

I ran a series of tests comparing SignalX with others signals:

• 1000 Creations
  

  
  
  And the SignalPlus is very good here!
  here my module is as bad as possible due to the use of buffers. Alas, I can’t improve it in any way.

• 1000 Fires no connections
  

  

• 1000 Connections no fires
  

  
  
  caveat: sometimes performance in the benchmark can slip, I’m not sure why this is, but it doesn’t seem to happen in the game itself.

• 1000 Connections + Disconnections
  

  

• 10 Fires 1 Connect
  

  
  
  to be fair, I’ll make this benchmark again when we get the new version of SignalPlus. Maybe I’m doing something wrong.
  as for poolThread, it’s just one thread that sits in your memory. In my opinion, this is a much better implementation than those that were in GoodSignal, and so on.

How To use?

Download: https://create.roblox.com/store/asset/127353288593114/SignalX

local Signal = require(path)
local mySignal = Signal.new()

-- alternatively, you can specify the power of two with the first argument, which should be > than the 
-- number of signals you have. This minimizes buffer expansion inside the code. If you specify the 
-- second argument, I do not recommend specifying more than 8 or 16
  
local connection = mySignal:Connect(function(...)
     print(...)
end)
mySignal:Fire("Hello from signal") -- prints "Hello from signal"
connection:Disconnect()

Important! Limitations:

• there is literally nothing but Fire and Connect, because I created the module only recently.
• The module is under active development, and API changes are possible.

The future of SignalX:

I plan to continue working on SignalX, adding new features and optimizations. I will be glad to hear your feedback and suggestions!

If you have any questions, feel free to ask them in this thread. I’ll try to answer everything.

Thanks for your attention!
Donate if u think it cool

By Proxiom

[AlexanderLindholt (Alexander)]

What are you benchmarking? There’s no info on that.

Also please don’t create duplicates. I believe that’s against the guidelines.
Just update your original post instead.

[bro_proxiomFun (Proxiom)]

[AlexanderLindholt (Alexander)]

Sure. Add yield support and then use the newest version of Signal+.

[bro_proxiomFun (Proxiom)]

Did I add support for this? Well, yes, of course, I agree with you, she’s literally a crutch right now, you’re right, I’m going to do it now!

[bro_proxiomFun (Proxiom)]

Version number 2

• Significantly improved performance
• Asynchronous signal dispatching (thread pool)

lol, I described adding 120 lines in 2 lines (Although I did it for a long time)

[athar_adv (athar_adventure)]

Hi, are you using version 63 for simplesignal?

[bro_proxiomFun (Proxiom)]

Yes, its version 63 strong text

[athar_adv (athar_adventure)]

I’ll do my own benchmarks rq just to see if it’s real ong

[bro_proxiomFun (Proxiom)]

I just clicked ctrl z in the benchmark code.

[athar_adv (athar_adventure)]

Ahh yeah there was some oversight causing the module to repeatedly preallocate threads I’ll benchmark it with a patched version really quick

[bro_proxiomFun (Proxiom)]

It’s great to see that you’ve found the problem! Fix your module soon and let me test it. We’ll have to redo the benchmarks again.

[athar_adv (athar_adventure)]

Hi, unfortunately something came up and i have to sleep right now, the patch is done but I havent benchmarked it yet Will be sending tomorrow

[roalex2008 (Dottik)]

I thought I had commented on this module and left my penny; anyway, while you preallocate, its still unwise to copy the buffer repeatedly on connect if the size is not enough (because you haven’t made it so it grows better), misc, but worth noting.

[AlexanderLindholt (Alexander)]

Hi there.

The benchmarks are not quite right visually.
But don’t worry, it’s not your fault! It’s BoatBomber’s fault.

There have been many bugs in BoatBomber’s Benchmarker plugin for a while, even though it’s paid.
I have just recently reported the issues, which he says will be looked into as soon as he has time.

I’m obviously not allowed to send even a modified version of the plugin, since it is not free.
Just know that one of the biggest issues is that the flamechart (the chart you highlight in your benchmarks) isn’t accurate. Look at the numbers, not the length of the horizontal bars — because for example in the 1000 Connections no fires benchmark of yours it looks like SignalX is ~10% faster than Signal+ but in reality they’re the same speed.

[bro_proxiomFun (Proxiom)]

Version 2.0.1
added MIT license

[TheYusufGamer (Asta)]

@bro_proxiomFun

1000 Creations

And the SignalPlus is very good here!
here my module is as bad as possible due to the use of buffers. Alas, I can’t improve it in any way.

I have edited the module slightly to improve it; however, I don’t own the Boatbomber Benchmarker plugin.

So I did a simple benchmark comparing my edited vs the original vs SignalPlus.

This is only for the 1000 creation time, and was able to achieve a 33% (avg) reduction in time

  22:28:03.402  Starting benchmark: 1000 instances x 5 trials for SignalX, NewX, and SignalPlus...  -  Server - Script:26
  22:28:03.418    Trial 1: SignalX = 11210.30 µs | NewX = 4515.40 µs | SignalPlus = 16.00 µs  -  Server - Script:59
  22:28:03.770    Trial 2: SignalX = 3782.70 µs | NewX = 3927.10 µs | SignalPlus = 17.60 µs  -  Server - Script:59
  22:28:03.893    Trial 3: SignalX = 3712.20 µs | NewX = 3741.60 µs | SignalPlus = 17.10 µs  -  Server - Script:59
  22:28:04.020    Trial 4: SignalX = 3676.40 µs | NewX = 3692.90 µs | SignalPlus = 18.40 µs  -  Server - Script:59
  22:28:04.130    Trial 5: SignalX = 3733.90 µs | NewX = 3660.10 µs | SignalPlus = 15.00 µs  -  Server - Script:59
  22:28:04.131  ---------------------------------------------------------  -  Server - Script:92
  22:28:04.131  -- BENCHMARK RESULTS --  -  Server - Script:93
  22:28:04.131  -- Iterations per Module per Trial: 1000  -  Server - Script:94
  22:28:04.131  -- Number of Trials: 5  -  Server - Script:95
  22:28:04.131  ---------------------------------------------------------  -  Server - Script:96
  22:28:04.131  [SignalX] Average Total (1000 iter): 5223.10 µs  -  Server - Script:99
  22:28:04.131  [SignalX] Average Per Instance:      5.2231 µs  -  Server - Script:100
  22:28:04.131  ---  -  Server - Script:101
  22:28:04.131  [NewX] Average Total (1000 iter):    3907.42 µs  -  Server - Script:104
  22:28:04.131  [NewX] Average Per Instance:      3.9074 µs  -  Server - Script:105
  22:28:04.131  ---  -  Server - Script:106
  22:28:04.131  [SignalPlus] Average Total (1000 iter): 16.82 µs  -  Server - Script:109
  22:28:04.131  [SignalPlus] Average Per Instance:   0.0168 µs  -  Server - Script:110
  22:28:04.131  ---------------------------------------------------------  -  Server - Script:111
  22:28:04.131  -- Comparison Ranking (Fastest First) --  -  Server - Script:128
  22:28:04.132    1. SignalPlus: 16.82 µs total (0.0168 µs per instance)  -  Server - Script:130
  22:28:04.132    2. NewX: 3907.42 µs total (3.9074 µs per instance)  -  Server - Script:130
  22:28:04.132       -> 3890.60 µs (23130.80%) slower than SignalPlus  -  Server - Script:144
  22:28:04.132    3. SignalX: 5223.10 µs total (5.2231 µs per instance)  -  Server - Script:130
  22:28:04.132       -> 1315.68 µs (33.67%) slower than NewX  -  Server - Script:144
  22:28:04.132  ---------------------------------------------------------  -  Server - Script:152

[TheYusufGamer (Asta)]

SignalX.rbxl (70.2 KB)

Place the file to check the benchmarks, I also added some extra stuff to the module. However, don’t take it as a fully working module, I just wanted to try and optimise some stuff, do your testing and debugging, etc. I added comments in the code as well

Callback ID Management (Memory)
Extra Thread Creation (Performance & Memory)
Buffer Resizing (Performance)
Finding Empty Slots (Minor Optimization)

Those are some chaged I made

[bro_proxiomFun (Proxiom)]

Version 2.0.2
Fixed a bug with passing only 1 argument

[towerscripter1386 (Tower)]

Hi, with my own benchmark against my own module, I found a fatal bug in your module that pretty much causes it to not execute functions after they yielded:

--!optimize 2
--!strict

-- yield test
local BadSignal = require(script.BadSignal)
local SignalX = require(script.SignalX)

local loops = 100
local Signal1 = BadSignal:new(1) -- 1 is to tell to make module dynamically allocate slots
local Signal2 = SignalX.new() -- empty to make it do the same as above

local t = os.clock()
for i = 1,loops do
	Signal1:Push(true,function(b:any) -- true because there is a yielding function
		local a = i*b::number
		task.wait()
		print("pTest1")
	end)
end
print("Bad signal connect: ",os.clock()-t)
t = os.clock()
for i = 1,loops do
	Signal2:Connect(function(b:any)
		local a = i*b::number
		task.wait()
		print("pTest2")
	end::any)
end
print("SignalX connect: ",os.clock()-t)

local t = os.clock()
for i = 1,10 do
	Signal1:Fire(2)
end
print("Bad signal fire: ",os.clock()-t)
local t = os.clock()
for i = 1,10 do
	Signal2:Fire(2)
end
print("SignalX fire: ",os.clock()-t)

The output in theory should have 1000 pTest1 and 1000 pTest2. But on tests there are only 1000 pTest1.
Also, I tested mine against non-yielding operations, and from the tests, once again, sometimes some of the functions don’t even fire:

--!optimize 2
--!strict

-- no yield test
local BadSignal = require(script.BadSignal)
local SignalX = require(script.SignalX)

local loops = 100
local Signal1 = BadSignal:new(1) -- 1 is to tell to make module dynamically allocate slots
local Signal2 = SignalX.new() -- empty to make it do the same as above

local t = os.clock()
for i = 1,loops do
	Signal1:Push(false,function(b:any) -- false because there are no yielding functions
		local a = i*b::number
		print("pTest1")
	end)
end
print("Bad signal connect: ",os.clock()-t)
t = os.clock()
for i = 1,loops do
	Signal2:Connect(function(b:any)
		local a = i*b::number
		print("pTest2")
	end::any)
end
print("SignalX connect: ",os.clock()-t)

local t = os.clock()
for i = 1,10 do
	Signal1:Fire(2)
end
print("Bad signal fire: ",os.clock()-t)
local t = os.clock()
for i = 1,10 do
	Signal2:Fire(2)
end
print("SignalX fire: ",os.clock()-t)

Bad signal connect:  0.0000092999980552122
SignalX connect:  0.00009100000170292333
▶ pTest1 (x1000)
Bad signal fire:  0.10646739999356214
▶ pTest2 (x600) 
SignalX fire:  0.08270960000663763 

As for cases when all of the functions fire, it loses against Bad Signal:

Bad signal connect:  0.000005299996701069176 
SignalX connect:  0.00010699999984353781
▶ pTest1 (x1000) 
Bad signal fire:  0.09846180000022287 
▶ pTest2 (x1000) 
SignalX fire:  0.10529619999579154 

(Note that those are singleton tests; they can vary but most of the time Bad Signal wins.)