Printable View
Okay so i tested hannes' code ..
I had to loop it 1 000 000 times to get 172ms.
and my code looped 49 times to get a comparable time ...!!!!Code:Dim T As Integer
Dim J As Integer
Dim I As Integer
T = GetTickCount
J = CInt(TextBox1.Text)
For I = 1 To j
Me.BackgroundImageLayout = ImageLayout.Tile
Me.BackgroundImage = PictureBox1.Image
Next I
Me.Text = GetTickCount - T
thats in excess of 20400 times faster :eek: ...
Well...partially... (without having looked at the full code or any other evaluation)...Quote:
Originally Posted by GremlinSA
The performance is probably based on two distinct factors.
1) The .NET routines are implemented significantly better and at a lower level.
2) The .NET libraries contain alot of "You are asking me to do something I have already done, so I am going to ignore you" optimizations. Better known as comparing existing state to requested state.
To compensate for the second part (which is still a benefit since you dont have to even think about it) re-work BOTH programs to alternate between two different images inside the loop....
..very interested in seeing THOSE results....
Okay so a breakdown so far.
code loops 100 times, on same sized form (Critical)
VB6 - API (Wizbang) = 94ms
VB.NET - GDI (GremlinSA) = 640ms
VB.NET - Std Control (Hannes) = 1.38ms(back calculated from 10000000 loop)
Perhaps we could reopen the Codewar forum's for this task.. ???
Gremmy..
--Edit--
Updated scores..
Okay initial testing has revealed you are 100% correct here...Quote:
Originally Posted by TheCPUWizard
I'm now calculating the results via comparisons..
Most interesting Gremmy :)
I agree that alternating images should give a better idea of performance.
Perhaps the attached sample will help, as it draws boxes rather than tiling an image. Now there's no built-in function of any language to skew results. Only a coded algorithm will do the job.
Gotta love it... YOU select an example.... YOU write the "best" (or at least "standard best practices") implementation.....Quote:
Originally Posted by WizBang
Then you say that it is not a good example of "performance".
Because Hannes was able to write a shorter version that did not require knowledge of anything outside the "included" material and performed orders of magnitude faster it is a clear indication that there are distinct benefits to developing in .NET (at least for certain cases).
Here is the results of my testing ...
Note all times are for 1000000 loops ...
Test 1 - Set tile mode only.. = 140
Test 2a - Set Center then stretch = 6942 (Control test)
Test 2b - Set Tile then Stretch = 7020
Test 2c - Set Tile then center = 6615
Test 3 - Set Image only = 78
Test 4a - Set image then nothing = 16849 (Control test)
Test 4B - Set Image 1 then Image 2 = 20998
From test 2 we can deduce a few things, Setting the layout to Stretch takes the longest however, Center and Tile have comparable times. hence setting tile layout time is 3307ms (1/2 of tile-center time)
From test 4 we can deduce that setting the image to nothing is faster than setting the image ( Dah, common sence), and by switching between two identical images (i used two pictureboxes with the same image), the time to set the image is 10499ms
so the time to set tile mode and the image is 13806ms.. (thats for 1000000 loops) do some maths and we get 1.38ms for 100 loops..
all tests done on same size form as the previous two, and as a control both previous sample tests were done before and after again to ensure simmular system responce. (with near identical results)
all the above tests were done a number of times and the best repetitive time was used..
IE: results for 8 runs of test 4 .. :21045,20920,20998,21060,20999,21135,20998,21045
20920 was the fastest however 20998 occured 3 times (20999 included as tick times are updated at 15 - 16ms intervals)
Gremmy..
looking back at the results i forgot why i did test 1 and test 3..
It was a control test to see the time used for the acctual looping..
Hmm so lets calculate it..
Total looptime all included =172ms (previous test)
Test 1 = 140ms
Test 3 = 78ms
Total = 218ms.
218 - 172 = 46ms ..
So the loop overhead is 46ms
However because of the calculations used the loop overhead was canceled out ..
Doing all the calculations over, removing the overhead time, then adding it to the end returns identical results. This was purely by luck, because both our times were halved and added to eachother we did not need to compensate..
Gremmy...
Not quite. The method Hannes implemented wasn't equivalent to what I wrote, since an object handled the actual tiling, not the code. Consider if it was the VB6 code which used such an object and you had to write the loop in .net. Needless to say it wouldn't be equivalent either, and you'd rightfully be pointing it out.Quote:
Originally Posted by TheCPUWizard
Now, my second example avoids this. Neither language can simply hand the task to something else to do the work in one shot. Both must loop through and so forth.
No this is the FUNDAMENTAL POINT where we disagree. From a practical standpoiint the issue is to accomplish a goal, by what ever means the selected environment supports (identifying any EXGTERNAL components used) and using the BEST practice for that environment/language.Quote:
Originally Posted by WizBang
It is NOT about forcing a specific implementation.
If I gave two developers the task. Write me a program which fills a windows of a specific size with a tiled image, my concerns are (in order):
1) How robust is the program?
2) How maintainable is the program?
3) How long did the program take to write?
4) How long does the program take to run?
As a general rule (with plenty of exceptions) a cleanly written short program is more robust (less chance of bugs), more maintainable (less code to understand).
I also belive that it took Hannes less time to type in and debug the program than it took you to to do the same.
You also had to make use of other material to accomplish your goal.....
And ... oh yeah.... Your program was SIGNIFICANTLY slower...
I would find it very difficult to "sell" your program as a better alternative to Hannes".....
The very reason I used a loop was to get a measure of speed at which it would be executed. Even if VB6 had a tile mode, I still wouldn't have used it.Quote:
Originally Posted by TheCPUWizard
Similarly, if I had copied the image one pixel at a time, that's what I'd want to see done in .net. Using BitBlt is fine when you're writing for money, but it would hardly make for a good comparison in this case. What I (and others) really want to know is how VB6 and .net compare for speed, doing the same work. I'm all for using built-in stuff when appropriate, and IF all I wanted was to tile an image on the form in .net, I'd very likely do as Hannes did. But again, that's not the point. The point is to have some sort of simple code (so those rewriting don't have to spend all day) which takes a measurable amount of time to execute. Otherwise we really don't get a measure of the speed of execution of said code, but that of an object which is obviously operating at a lower level.
Let's see how the second sample goes. I'm sure you aren't going to suggest it isn't a fair test. And unless I miss my guess, .net doesn't have an object to do it all for us. So both implementations will need to actually loop though and do it one box at a time. And no, producing an image beforehand, and simply applying it to the form at runtime would also invalidate the test.
This is the sticking point with me.Quote:
Originally Posted by WizBang
As myself and others have suggested, it isn't about the coding in the same style between VB6 and .Net that matters, it's about getting the equivalent work done. Chances are if you code .Net using a VB6 approach you are likely taking the wrong approach, or at the very least you aren't taking full advantage of what the newer framework has to offer.
Quote:
Originally Posted by Arjay
Exactly.
I suppose we could draw each tiled object a bit at a time, but we could also run each in a separate thread, which would be faster?Quote:
Originally Posted by WizBang
I think it's obvious.
Using one thread per pixel? ;)Quote:
Originally Posted by dglienna