Which is faster? - GameCreators Forum

Author

Message

george++

AGK Tool Maker

16

Years of Service

User Offline

Joined: 13th May 2007

Location: Thessaloniki, Hellas

Posted: 6th Jan 2021 14:45 Edited at: 6th Jan 2021 14:46

Link

+ Code Snippet

type AType
  aninterger as integer
endtype

global myType as AType

myType.aninteger = 1

or

+ Code Snippet

global aninterger as integer

aninteger = 1

Back to top

Profile PM Email

PSY

Developer

7

Years of Service

User Offline

Joined: 3rd Jul 2016

Location: Laniakea Supercluster

Posted: 6th Jan 2021 16:02

Link

As you're declaring the variable only once, it won't make any difference.

Theoretically, your second varible assignment is slightly faster than your first one.

I did 10 million assignments using myType.aninteger = 1 vs 10 million assignments using aninteger = 1, and the difference was about 50-100 milliseconds.

That's a difference of 0.000005 to 0.00001 milliseconds per assignment...

PSY LABS Games
Coders don't die, they just gosub without return

Back to top

Profile PM Website

Raven

19

Years of Service

User Offline

Joined: 23rd Mar 2005

Location: Hertfordshire, England

Posted: 6th Jan 2021 19:05 Edited at: 6th Jan 2021 19:06

Link

This post has been marked by the post author as the answer.

// Project: Test 
// Created: 20-12-11

// show all errors

SetErrorMode(2)

// set window properties
SetWindowTitle( "Test" )
SetWindowSize( 1024, 768, 0 )
SetWindowAllowResize( 1 ) // allow the user to resize the window

// set display properties
SetVirtualResolution( 1024, 768 ) // doesn't have to match the window
SetOrientationAllowed( 1, 1, 1, 1 ) // allow both portrait and landscape on mobile devices
SetSyncRate( 0, 1 ) // 30fps instead of 60 to save battery
SetScissor( 0,0,0,0 ) // use the maximum available screen space, no black borders
UseNewDefaultFonts( 1 )

#Constant CountMax 999999

Type Stopwatch
	Start As Float
	Stop As Float
	Duration As Float
EndType

Global Time As Stopwatch

Type AType
	AnInteger As Integer
EndType

Global myTypeValue As AType
Global myValue As Integer

Time.Start = Timer()
For Count = 0 To CountMax
	// Benchmark Loop to remove the For..Next Overhead
Next
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global ForNext As Float : ForNext = Time.Duration

Time.Start = Timer()
Count = 0
Repeat
	// Alternative For..Next
	Inc Count
Until Count > CountMax
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global ReUntil As Float : ReUntil = Time.Duration

Time.Start = Timer()
For Count = 0 To CountMax
	myTypeValue.AnInteger = 1
Next
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global Typed As Float : Typed = Time.Duration

Time.Start = Timer()
For Count = 0 To CountMax
	myValue = 1
Next
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global Basic As Float : Basic = Time.Duration

Repeat
	Print( "ForNext Time: " + Str(ForNext, 2) + "ms" )
	Print( "RepeatUntil Time: " + Str(ReUntil, 2) + "ms" )
	Print( "Typed Value: " + Str(Typed - ForNext, 2) + "ms : Total ( " + Str(Typed, 2) + "ms )" )
	Print( "Basic Value: " + Str(Basic - ForNext, 2) + "ms : Total ( " + Str(Basic, 2) + "ms )"  )
	Sync()
Until GetRawKeyPressed(27) // wait for escape key

+ Code Snippet

// Project: Test 
// Created: 20-12-11

// show all errors

SetErrorMode(2)

// set window properties
SetWindowTitle( "Test" )
SetWindowSize( 1024, 768, 0 )
SetWindowAllowResize( 1 ) // allow the user to resize the window

// set display properties
SetVirtualResolution( 1024, 768 ) // doesn't have to match the window
SetOrientationAllowed( 1, 1, 1, 1 ) // allow both portrait and landscape on mobile devices
SetSyncRate( 0, 1 ) // 30fps instead of 60 to save battery
SetScissor( 0,0,0,0 ) // use the maximum available screen space, no black borders
UseNewDefaultFonts( 1 )

#Constant CountMax 999999

Type Stopwatch
	Start As Float
	Stop As Float
	Duration As Float
EndType

Global Time As Stopwatch

Type AType
	AnInteger As Integer
EndType

Global myTypeValue As AType
Global myValue As Integer

Time.Start = Timer()
For Count = 0 To CountMax
	// Benchmark Loop to remove the For..Next Overhead
Next
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global ForNext As Float : ForNext = Time.Duration

Time.Start = Timer()
Count = 0
Repeat
	// Alternative For..Next
	Inc Count
Until Count > CountMax
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global ReUntil As Float : ReUntil = Time.Duration

Time.Start = Timer()
For Count = 0 To CountMax
	myTypeValue.AnInteger = 1
Next
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global Typed As Float : Typed = Time.Duration

Time.Start = Timer()
For Count = 0 To CountMax
	myValue = 1
Next
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global Basic As Float : Basic = Time.Duration

Repeat
	Print( "ForNext Time: " + Str(ForNext, 2) + "ms" )
	Print( "RepeatUntil Time: " + Str(ReUntil, 2) + "ms" )
	Print( "Typed Value: " + Str(Typed - ForNext, 2) + "ms : Total ( " + Str(Typed, 2) + "ms )" )
	Print( "Basic Value: " + Str(Basic - ForNext, 2) + "ms : Total ( " + Str(Basic, 2) + "ms )"  )
	Sync()
Until GetRawKeyPressed(27) // wait for escape key

This should show you the benchmark figures... it's slower, but change the MaxCount Value; you'll notice the performance difference reduces with more iterations.
Obviously though the key issue is that the more iterations the slower the longer overall it takes.

I also included a quick'n'dirty benchmark for 2 different approaches to iterations.
For...Next is SLOWER than you might think; but I can't tell you why.

Arrays also come with a similar performance penalty to Typed Data, so keep that in mind; but I'm sure you can add a benchmark run for that.
Now as a note... if you want more accurate data, I'd recommend Multiple Runs (3-5) that are then averaged; as the more times then with averaging approaches you can remove unusually high/low scores to give a more aggregated (expectable) result.

[edit]
I should note that with very minor changes this code will also run in Dark BASIC Professional., should you be curious to see the performance difference.
Trust me, it's a very interesting result that you might not expect.

Back to top

Profile PM Email Website

PartTimeCoder

AGK Tool Maker

9

Years of Service

User Offline

Joined: 9th Mar 2015

Location: London UK

Posted: 6th Jan 2021 19:25

Link

Declaration of type variables is slightly slower than basic globals but access speeds are pretty mush the same, for me readability trumps performance, types tidy up the code, organize vars and such.

I like to modularize everything when I write code and types help with this, if I take a tiny performance hit but have modular code I can drop into other projects then its well worth a few hundreds of a millisecond overhead lol

Back to top

Profile PM

george++

AGK Tool Maker

16

Years of Service

User Offline

Joined: 13th May 2007

Location: Thessaloniki, Hellas

Posted: 6th Jan 2021 22:45 Edited at: 6th Jan 2021 22:52

Link

Thank you very much

Quote: "As you're declaring the variable only once, it won't make any difference."

I declare it only once ~~but I use the typed variable as an argument in a function call within the main application loop~~ but I make assignments within the main application loop

@Raven: thank you very much for the code snippet. I should have thought of it before posting it

Quote: "I should note that with very minor changes this code will also run in Dark BASIC Professional"

Errr... I was a Blitz3D guy and it was super fast. I know nothing about DBPro

, but I can imagine that DBPro is much faster than AppGameKit since it is a compiler, right?

Quote: "types tidy up the code, organize vars and such."

This is my dilemma... Easier code reading (and maintenance) or faster execution...?

Back to top

Profile PM Email

Raven

19

Years of Service

User Offline

Joined: 23rd Mar 2005

Location: Hertfordshire, England

Posted: 7th Jan 2021 09:51

Link

george++ wrote: "This is my dilemma... Easier code reading (and maintenance) or faster execution...?"

Well., it all depends on how much you need to read-write values.
As you'll see from the basic benchmark code I wrote., realistically the main performance overhead ISN'T from Reading / Writing the Variables themselves.

Sure, Structured Types are considerably slower ... but the biggest performance overhead stems from the Loops themselves.

I've said this in the past, and I'll repeat this again; but the biggest issue with AppGameKit Script is performance.
Performance is abysmal, even for a Scripting Language.

I'd recommend not worrying to much about using Structured Types., instead just focus on smaller projects that don't need to rely on a lot of logic or in-flight data.

Back to top

Profile PM Email Website

Kevin Picone

21

Years of Service

User Offline

Joined: 27th Aug 2002

Location: Australia

Posted: 8th Jan 2021 02:05

Link

All instructions have different performance, so If you need to time something it's best to unroll the operation rather than jam in within a tight loop. So your timing more of the operation and less loop overhead.

Time.Start = Timer()
For Count = 0 To CountMax
myTypeValue.AnInteger = 1
myTypeValue.AnInteger = 2
myTypeValue.AnInteger = 3
myTypeValue.AnInteger = 4
myTypeValue.AnInteger = 5
myTypeValue.AnInteger = 6
myTypeValue.AnInteger = 8
myTypeValue.AnInteger = 9
myTypeValue.AnInteger = 10
Next
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global Typed As Float : Typed = Time.Duration

Time.Start = Timer()
For Count = 0 To CountMax
myValue = 1
myValue = 2
myValue = 3
myValue = 4
myValue = 5
myValue = 6
myValue = 7
myValue = 8
myValue = 9
myValue = 10
Next
Time.Stop = Timer()

also, since reading is more prevalent than writing it's handy to time the read performance also.

Time.Start = Timer()
For Count = 0 To CountMax
temp =myTypeValue.AnInteger
temp =myTypeValue.AnInteger
temp =myTypeValue.AnInteger
temp =myTypeValue.AnInteger
temp =myTypeValue.AnInteger

temp =myTypeValue.AnInteger
temp =myTypeValue.AnInteger
temp =myTypeValue.AnInteger
temp =myTypeValue.AnInteger
temp =myTypeValue.AnInteger

Next
Time.Stop = Timer()
Time.Duration = (Time.Stop - Time.Start) * 1000.0
Global Typed As Float : Typed = Time.Duration

Most gains from runtimes/VM are found in caching results that tend to be consistently computed on demand..

etc etc

PlayBASIC To HTML5/WEB - Convert PlayBASIC To Machine Code

Back to top

Profile PM Website

Arch-Ok

AGK Developer

4

Years of Service

User Offline

Joined: 11th Jul 2019

Location: Bursa/TÜRKIYE

Posted: 12th Jan 2021 19:51 Edited at: 12th Jan 2021 20:50

Link

second one is faster

mytype.mydata = 1

+ Code Snippet

Asm
mov eax, 1
mov ebx, [mytype]
mov [ebx + _MYDATA_OFFSET_], eax ;_MYDATA_OFFSET_ is 0 here
End Asm

mydata = 1

+ Code Snippet

Asm
mov [mydata], 1
End Asm

Back to top

Profile PM

george++

AGK Tool Maker

16

Years of Service

User Offline

Joined: 13th May 2007

Location: Thessaloniki, Hellas

Posted: 12th Jan 2021 21:24

Link

@Arch-Ok: is this a disassembly?. Where did you find this?

Back to top

Profile PM Email

Arch-Ok

AGK Developer

4

Years of Service

User Offline

Joined: 11th Jul 2019

Location: Bursa/TÜRKIYE

Posted: 12th Jan 2021 21:38

Link

@george++ no it is not. very simple way of showing the difference. Also, the difference gets bigger, if you have array of types.

Back to top

Profile PM

george++

AGK Tool Maker

16

Years of Service

User Offline

Joined: 13th May 2007

Location: Thessaloniki, Hellas

Posted: 13th Jan 2021 12:55

Link

When I was young, I was programing the 8088 on an Amstrad 1512, for sprite handling.
Thanks to your post I remembered the old good days...

Back to top

Profile PM Email

Sorry your browser is not supported!

AppGameKit Classic Chat / [SOLVED] Which is faster?