Speed test

[Marc van Breemen]

Hi,

I'm currently a VB.NET 2008 (Framework 3.5) user, but am seriously considering using PB as a second programming tool and on the long run as a replacement of VB.NET.

To convince myself that that is a good thing to do I'm investigation PB's quality in several ways:

• Reading through several threads in PB's forum.
• Reading the newsgroups, although those aren't a great source regarding the longtime 'war' going on there between persons I will not mention .
• Studying the online documentation of PB/Win9.

On other thing I like to do now, and for which I need your help, is to test PB's speed, because that would be one of the main reasons I'm interested. As PB doesn't have a trial version I would kindly ask you to provide me with a compiled version of the code below (created with VB.NET):

        'UInteger = unsigned 32-bit (4-byte)
        'Integer = signed 32-bit (4-byte)
        'Double = double-precision floating point number

        Dim StopWatch As New System.Diagnostics.Stopwatch()
        Dim strString As String = ""
        Dim dblDouble As Double = 0

        StopWatch.Start()

        For intIndex As UInteger = 0 To 100000

            strString = ""
            For intString As Integer = 0 To 200
                strString = "ab" & strString & "de"
            Next
            strString = Replace(strString, "ab", "123")

            dblDouble = 9876.54321
            For intDouble As Integer = 0 To 1000
                dblDouble *= dblDouble
                dblDouble /= ((dblDouble + 100) / 2)
            Next
        Next

        MsgBox(StopWatch.ElapsedMilliseconds.ToString)

On my PC this test, using a compiled EXE, runs between 11165-13066 milliseconds. These are the minimum and maximum times recorded on numerous executions. My PC: XP SP3, 4GB RAM, Intel Core 2 Duo T5870 @ 2.00Ghz.

I certainly don't say that this is the greatest test ever made, but it will give me a clue on string and calculation operations. If you'll have better code examples (with compiled EXE's) I'll be very grateful to receive them from you.

Second question:

To be honest I think that VB.NET is a great programming language (also thanks to the huge framework) and VS2008 is a very comfortable IDE (especially the intellisense feature). On the other hand it can certainly be slow sometimes and sometimes the IDE eats 500MB of my memory.

Besides the speed can you think of any more reasons to switch to PB?.

Regards,
Marc
(once user of the great TurboBasic)

[Patrice Terrier]

#COMPILE EXE
#OPTION VERSION5

DECLARE FUNCTION GetTickCount LIB "KERNEL32.DLL" ALIAS "GetTickCount" () AS DWORD

FUNCTION PBMAIN
    LOCAL strString AS STRING
    LOCAL nIndex, nString, nDouble AS LONG, dDouble AS DOUBLE, dwTick AS DWORD 

    dwTick = GetTickCount

    FOR nIndex = 0 TO 100000

        strString = ""
        FOR nString = 0 To 200
            strString = BUILD$("ab", strString, "de")
        NEXT
        REPLACE "ab" WITH "123" IN strString

        dDouble = 9876.54321
        FOR nDouble = 0 To 1000
            dDouble *= dDouble
            dDouble /= ((dDouble + 100) / 2)
        NEXT
    NEXT

    MsgBox "Ellapsed ticks" + STR$(GetTickCount - dwTick)
END FUNCTION

Result: 9313 ticks
on a notebook dual core 2 duo running VISTA 64.bit

[Michael Mattias]

>As PB doesn't have a trial version

AFAIK they still do. It's not really a 'trial' version, it's "the" version, for which you pay the regular price. But there is a thirty-day-no-questions asked money-back guarantee.

At least that used to be the program. Contact PB's sales department to get official word.

If you're not interested in that ......in both the Source Code Forum and the File Section, there are any number of posts where the poster has supplied not only the source code, but an executable as well. That might give you an idea of performance.

I, too, have been using PB products starting with Turbo Basic. Don't sweat it, the compiler still generates extremely efficient code.

(But as we all know, true performance does not come from hardware or software... it comes from the wetware... you know, that stuff between your ears?)

MCM

[John Gleason]

Iterative string concatenation like above, tho improved a lot by BUILD, is a notoriously slow technique to build up strings. VB.NET may not have the functions like PB to increase efficiency of code in situations like these and others, which is another reason you may like PB. Below I use the MID$ function to further double the speed over even BUILD.

#COMPILE EXE
#DIM ALL

DECLARE FUNCTION GetTickCount LIB "KERNEL32.DLL" ALIAS "GetTickCount" () AS DWORD

FUNCTION PBMAIN
    LOCAL strString AS STRING
    LOCAL nString, nDouble AS LONG, dDouble AS DOUBLE, nIndex, dwTick AS DWORD

    dwTick = GetTickCount

    FOR nIndex = 0 TO 100000

        strString = STRING$(4 * 201, "x")
        FOR nString = 0 TO 200
            MID$(strString, 2 * 201 - 2 * nString - 1) = "ab"
            MID$(strString, 2 * 201 + 2 * nString + 1) = "de"
'            strString = BUILD$("ab", strString, "de")
        NEXT
        REPLACE "ab" WITH "123" IN strString

        dDouble = 9876.54321
        FOR nDouble = 0 TO 1000
            dDouble *= dDouble
            dDouble /= ((dDouble + 100) / 2)
        NEXT
    NEXT

    MSGBOX "Ellapsed ticks" + STR$(GetTickCount - dwTick)
END FUNCTION

[Kev Peel]

The speed tests prove nothing - any compiler can optimize a loop (even VB6). PB's real power is in the finished result, which feels smoother and is a great deal smaller memory footprint than most. I have used almost all languages and nothing gives you the power of PB (except maybe straight "C", which is a painful write-only language). PB's syntax is easy to understand, using plain English words and structured if written by a seasoned programmer.

Once you've learned the Windows API, programming in PB is a breeze, and of course we have this forum for any questions..

[Michael Mattias]

> The speed tests [shown here] prove nothing...

AMEN MY BROTHER!

[Patrice Terrier]

Kev,

I agree with everything you have written

...

[Marc van Breemen]

Result: 9313 ticks
on a notebook dual core 2 duo running VISTA 64.bit

Thx. I'm quite satisfied with the speed, on my PC 9219-10215 ms.

[Marc van Breemen]

>As PB doesn't have a trial version

AFAIK they still do. It's not really a 'trial' version, it's "the" version, for which you pay the regular price. But there is a thirty-day-no-questions asked money-back guarantee.

At least that used to be the program. Contact PB's sales department to get official word.

If you're not interested in that ......in both the Source Code Forum and the File Section, there are any number of posts where the poster has supplied not only the source code, but an executable as well. That might give you an idea of performance.

I, too, have been using PB products starting with Turbo Basic. Don't sweat it, the compiler still generates extremely efficient code.

(But as we all know, true performance does not come from hardware or software... it comes from the wetware... you know, that stuff between your ears?)

MCM

I know about the money-back guarantee, although it's not unconditional according to what I've read. Correct me if I'm wrong. I prefer not to spend shipping costs vica-versa if I can get some test result via this channel.

I know that this test doesn't say everything, it's only a small sign. Other things like database performance may be the bottleneck factor regarding speed and not the program itself.

[Marc van Breemen]

Iterative string concatenation like above, tho improved a lot by BUILD, is a notoriously slow technique to build up strings. VB.NET may not have the functions like PB to increase efficiency of code in situations like these and others, which is another reason you may like PB. Below I use the MID$ function to further double the speed over even BUILD.

Now that's impressive!: 6516-7250 ms.

NB: Funny thing is that the size of your EXE is exactly the same as Patrice's example: 9216 bytes. How's that possible?

[Marc van Breemen]

PureBasic ... OOPS!

Some guy suggested me to look at PureBasic, so I did. I'm totally new with PureBasic, so maybe I'm doing something wrong, but the time of PureBasic is: 29640-29844 ms!.

Below is the code I used in PureBasic. Maybe someone can optimize the code to give PureBasic a honest comparison (if possible). Right now it isn't an alternative (the language looks also much less mature).

StartTime = ElapsedMilliseconds()

strString.s = ""
dblDouble.d = 0
intIndex.i = 0
intString.i = 0
intDouble.i = 0

For intIndex.i = 0 To 100000

    strString.s = ""
    For intString.i = 0 To 200
        strString.s = "ab" + strString.s + "de"
    Next intString.i
    strString.s = ReplaceString(strString, "ab", "123")

    dblDouble.d = 9876.54321
    For intDouble.i = 0 To 1000
        dblDouble.d = dblDouble.d * dblDouble
        dblDouble.d = dblDouble.d / ((dblDouble.d + 100) / 2)
    Next intDouble.i
Next

ElapsedTime = ElapsedMilliseconds()-StartTime 

MessageRequester("","Tijd:" + Str(ElapsedTime), #PB_MessageRequester_Ok)

[Marc van Breemen]

The speed tests prove nothing - any compiler can optimize a loop (even VB6). PB's real power is in the finished result, which feels smoother and is a great deal smaller memory footprint than most. I have used almost all languages and nothing gives you the power of PB (except maybe straight "C", which is a painful write-only language). PB's syntax is easy to understand, using plain English words and structured if written by a seasoned programmer.

Once you've learned the Windows API, programming in PB is a breeze, and of course we have this forum for any questions..

That sounds interesting.

In VB.NET some controls can be a little slow, like the LISTVIEW control. I'm talking about adding items and sorting. Does PB use the same controls or is it in PB different eg. faster/smoother?

[Michael Mattias]

I know about the money-back guarantee, although it's not unconditional according to what I've read. Correct me if I'm wrong

I am not now, have never been, nor am ever likely to be an employee, agent, officer or director of PowerBASIC Inc.

If you want corrrections or clarifications please contact [email protected].

MCM

[Paul Dixon]

Marc,
the following change knocks another 25% off the time in John's code above:
remove:

dDouble AS DOUBLE,

Add the line:

REGISTER dDouble AS EXT

It calculates more accurately than a double as it uses an 80bit extended FP value instead of a 52bit DOUBLE FP value (so may give slightly different results) but it stores dDouble in a register which runs faster.

9216 bytes. How's that possible?

9216 = 9 x 1k. If the block size on disk is 1k then any code between 8193 and 9216 bytes will occupy 9216 bytes on disk.


Paul.

[Marc van Breemen]

I am not now, have never been, nor am ever likely to be an employee, agent, officer or director of PowerBASIC Inc.

If you want corrrections or clarifications please contact [email protected].

MCM

Sorry if you misunderstood me.

I know you're not part of the PB staff and I certainly *not* intended to try changing the policy of PB just for my sake. I was just expressing what I've read.

[Marc van Breemen]

It calculates more accurately than a double as it uses an 80bit extended FP value instead of a 52bit DOUBLE FP value (so may give slightly different results) ....

I really not have a clue what you're saying there (yet) . But I'm sure in a while, after I've studied PB more, I will.

[Paul Dixon]

Marc,

Maybe someone can optimize the code to give PureBasic a honest comparison

You're more likely to get that at the PureBasic forum.

Paul.

[Marc van Breemen]

Marc,

You're more likely to get that at the PureBasic forum.

Paul.

I know, but I won't do that as I doubt that the results will improve that much to go beyond PowerBasic's results so far. Besides that I don't think PureBasic is mature enough language wise to see it as an alternative for VB.NET/PB.

It's cheap (79 euro) with free updates "for life", which sounds better than PB but isn't. I don't think you can have a serious long standing business if you're practically give away your product like they do.

[Paul Dixon]

Marc,

I really not have a clue what you're saying hear (yet)

Don't worry about it. Attched is a program with the one change mentioned so you can see if it is faster or not.

Paul.

[Paul Dixon]

Marc,
just a thought on the speed of your Purebasic example.
If you're using the demo version then I think it might run with the debugger enabled at all times. For a fair speed comparrison you need to run with the debugger turned off.

Paul.