Adds a CPU-cache performance test. The libprof measure is extended into
Iϕ cycles (resolution: 16 Iϕ cycles) to check the 2-ILP parallelism
hypothesis (confirmed on fx-9860G and fx-CG 50).
The time needed to traverse a small buffer multiple times is also
measured with the same method in an attempt to determine cache size. No
definite result could be obtained.
This change includes a plotting utility.
Lephe
