Data point:
Once upon a time, to study the performance of square versus jagged arrays in .NET I ported to C# the source code of one of the benchmarks mentioned in the article "The NINJA project. CACM 44(10): 102-109 (2001)."
I dug it out today and ran it against the latest version of Microsoft's C# (7.10.3052.4), Mono (Release 0.25, June 25, 2003), and ran the Java version against J# (7.10.3052.0) and Java (1.4.2). I ran each test 3 times on a Motion Computer M1200 Tablet Computer (Mobile P3, 866MHz, 512MB RAM). Timing is done on either side of a function call, so this should not reflect differences in VM start-up time. Results in milliseconds:
|
Run # |
Part 1 |
Part 2 |
| Microsoft C# |
|
1 |
2754 |
3445 |
|
2 |
2744 |
3485 |
|
3 |
2794 |
3515 |
| Mono C# |
|
1 |
3435 |
3785 |
|
2 |
3385 |
3825 |
|
3 |
3395 |
3815 |
| J# |
|
1 |
3375 |
3766 |
|
2 |
3435 |
3825 |
|
3 |
3365 |
3825 |
| Java |
|
1 |
4517 |
4547 |
|
2 |
4546 |
4567 |
|
3 |
4517 |
4566 |
Source code for
C#, here for
J#, here for
Java.