BUGS:
Tests spin on 32 bit x86 compile with gcc-- longjmp test spins forever.
CPU_NAME is wrong on earlier (than 10) and later (than 10) versions of Solaris
(parsing psrinfo output is fragile)
TimerRes detection on some Solaris 8 SPARC systems is wrong and falsely
detects resolution as 1nsec
TODO:
.o files should probably be separated from generated binaries (obj-sun4u?)
Move more of the system info data gathering into tattle
Would be nice to show the platform name (uname -i) if available
Multiview's "red" can get a little too bright.
Consider recoding multiview to use a different performance indicator than
color (smileys and bombs?) -- today a 20% perf difference is hard to
spot (it's very very light pink).
FreeBSD ports collection has patches to make libmicro work on FreeBSD.
Fold those in (or reimplement)