Misleading benchmarks, comparing apples to oranges

[m-rei]

Specifically related to the comparison between odin and c, the task being spectral-norm

The c implementation is using openmp (parallel programming, using all the threads), the odin implementation is not!

Furthermore, the odin implementation does allocate in a loop (see a_times_transp()), whereas in the c version, the array is preallocated once.

This is extremely misleading. Thus the highscoring comparison makes no sense.
Very disappointed.