Hi,
I ran the Sierpinski benchmark again using the new change model.
Compared to what I measured before (see
http://www.eclipse.org/modeling/emft/henshin/examples.php?example=sierpinski)
this is much faster. I am nit sure whether I was doing something
wrong in the old benchmark, because the difference is so huge.
Anyway, what you can definitely see is that the matching time is
more or less the same as before, but the application time is much
better now, which I believe is the result of the new change model.
Level |
Rule applications |
Nodes |
Matching Time |
Application Time |
Total Time |
1 |
1 |
6 |
37ms |
4ms |
41ms |
2 |
3 |
15 |
0ms |
1ms |
1ms |
3 |
9 |
42 |
1ms |
3ms |
4ms |
4 |
27 |
123 |
4ms |
8ms |
12ms |
5 |
81 |
366 |
13ms |
27ms |
40ms |
6 |
243 |
1,095 |
37ms |
67ms |
104ms |
7 |
729 |
3,282 |
73ms |
158ms |
231ms |
8 |
2,187 |
9,843 |
125ms |
195ms |
320ms |
9 |
6,561 |
29,526 |
107ms |
231ms |
338ms |
10 |
19,683 |
88,575 |
185ms |
219ms |
404ms |
11 |
59,049 |
265,722 |
441ms |
723ms |
1,164ms |
12 |
177,147 |
797,163 |
1,309ms |
2,500ms |
3,809ms |
13 |
531,441 |
2,391,486 |
3,627ms |
9,944ms |
13,571ms |
14 |
1,594,323 |
7,174,455 |
14,609ms |
117,470ms |
132,079ms |
Cheers,
Christian
PS: will update the data on the website soon
|