>
> And on an experiment in spectral identification that mainly does F64 inner products, I get with the default build
>
> (time (run-experiment-and-graph 5))
>   785617 ms real time
>   785590 ms cpu time (785040 user, 550 system)
>   33 collections accounting for 442 ms real time (350 user, 80 system)
>   3216061748 bytes allocated
>   30338 minor faults
>   no major faults
>
> and with the gcc-4.8 build I get
>
>> (time (run-experiment-and-graph 5))
> (time (run-experiment-and-graph 5))
>    739829 ms real time
>    739790 ms cpu time (739620 user, 170 system)
>    33 collections accounting for 292 ms real time (310 user, 0 system)
>    3216060588 bytes allocated
>    17663 minor faults
>    no major faults

using -fstrict-aliasing didn’t help the spectral identification code much:

> (time (run-experiment-and-graph 5))
(time (run-experiment-and-graph 5))
    733849 ms real time
    733750 ms cpu time (733330 user, 420 system)
    33 collections accounting for 219 ms real time (210 user, 0 system)
    3216086556 bytes allocated
    16938 minor faults
    no major faults


Awesome project! Raspberry Pi is truly a wonderful machine. Thank you for posting your results.
Since I don't have a reference performance, is this 20x difference expected or more than expected? Based on other tests with very different environments on mobile ARM, I would say that it should be on the 10x-50x ballpark.