R2011b slower than R2010a!

Why should this be true? Same computer (Win XP SP3 32 bit)... overall bench scores of 38-40 under 2011b, and 60 under 2010a, which are repeatable. LU and FFT scores are much lower under the old version.

[Jan]

Are you talking about the built-in function BENCH or about the FEX submission?

Answer 1

[Robert]

• I see that the bench function actually changes between versions. So, for example, array lengths used in LU and FFT, are different, hence the difference in speed.

[Jan]

Isn't it a bad idea to change the problem size in a bench function?!