SBCL quicker than C?

After reading a nice conversation on comp.lang.lisp I decided to play a little bit with SBCL and Debian's Shootout Benchmarks.

I considered the spectral norm benchmark and compared the C and SBCL implementations.

I started re-running the tests without any further optimization and the results were similar to the ones reported with C beating SBCL by a factor of 2. (for the C optimization used, see here).

Then, I noticed that there was no (declaim (optimize (speed 3) (safety 0) (space 0))) in the SBCL file! I've added it and rerun the test. Here are my results:

N=500
gcc 0.15u 0.00s 0.17r
sbcl 0.08u 0.02s 0.21r

N=3000
gcc 5.60u 0.00s 5.69r
sbcl 5.18u 0.01s 5.41r

N=5500
gcc 18.81u 0.01s 19.12r
sbcl 17.42u 0.02s 17.76r

SBCL implementation is actually faster than C! You can find the code for the tests here.