Scratch

(I've posted this in All About Scratch because I don't think any of this is *too* technical, and it's I think it will be useful to most Scratchers.)

UPDATE - 1/21/12 - Here I have more results with testing in Turbo Mode.

Introduction

The Flash Player has confused me for a very long time. What makes it so fast? Many games that were incredibly slow (in Scratch 1.4 and the java player) can now be played very easily in the Flash Player. For example, check out Scratch Kart Demo by RHY3756547. In the java player, it's unplayable (doesn't actually load). In Scratch 1.4, it's playable, but doesn't live up to it's full potential. In the flash player, it runs at a much greater speed, and is very enjoyable.

Experiment

What I decided to do is set up an experiment. My purpose was to find out what features of the flash player are faster than Scratch 1.4 or the java player. My background research was... well there wasn't any xD. My hypothesis was that the flash player would be faster, as that is what I have found in my general experience. My procedure was to create a project that tested the different features of Scratch, and to test it in all 3 platforms (flash player, java player, and Scratch 1.4).

This is the project with the tests. The 7 tests were all within repeat (100) blocks, while using the timer to determine how long each one took. I repeated the project 3 times in each platform and averaged the times together for an accurate result. Inside the repeat (100) blocks, I had the following:

- nothing
- broadcast and wait (with an accompanied "when received" block)
- pen drawing in a square
- rotating square
- increasing size of Scratch Cat
- increasing ghost effect of Scratch Cat
- stamping Scratch Cat at random positions

Results

Here are my results:

Scratch 1.4

Test 1: 2.502 seconds
Test 2: 7.504 seconds
Test 3: 2.501 seconds
Test 4: 2.646 seconds
Test 5: 2.545 seconds
Test 6: 2.502 seconds
Test 7: 2.588 seconds

Java Player

Test 1: 3.000 seconds
Test 2: 6.000 seconds
Test 3: 3.000 seconds
Test 4: 3.000 seconds
Test 5: 3.000 seconds
Test 6: 3.000 seconds
Test 7: 3.000 seconds

Flash Player

Test 1: 0.000 seconds
Test 2: 3.276 seconds
Test 3: 3.287 seconds
Test 4: 3.288 seconds
Test 5: 3.316 seconds
Test 6: 3.315 seconds
Test 7: 3.308 seconds

Conclusion

Let's start with Scratch 1.4. A repeat block takes about 0.025 seconds for one loop, and a broadcast block takes 0.05 seconds to be sent and received. These are the underlying causes of lag in Scratch. Apart from that, test 4, 5, and 7 have a bit of a difference, but nothing important.

Now let's take a look at the java player. A repeat block takes about 0.03 seconds for one loop, and a broadcast block takes 0.03 seconds to be sent and received. The results of these tests were very consistent. So, the broadcasts are faster than those in Scratch 1.4, but the repeat blocks are slower.

Lastly, the flash player. The repeat block takes 0 seconds for one loop, and -- wait what!? It seems that the flash player is smart enough to see through this trick, and simply ignores the empty repeat block. The rest of the results are fairly similar, varying from 3.275 to 3.325 seconds.

This is confusing however, because these times are greater than those of Scratch 1.4. Yet the flash player is faster? Clearly my hypothesis is wrong. What I have concluded is that I have done a lot of rendering which is slower than Scratch 1.4, but I have not tested calculating, which I guess is faster. I am most likely going to make another experiment to test this.

Well that's the end! I hope you've learned a bit about the flash player. If you have any comments about the results, or suggestions for the next experiment (to test calculations and long scripts), then please do post!

And of course, Scratch On! 

Last edited by Lucario621 (2012-01-21 16:33:49)