Fixed ticks per second detection for Linux.

[sgalaburda]

The ticks per second value was hard-coded in tsc_time.pyx module. This caused the time in shrapnel-based program to notably diverge from the system time. I implemented detection of actual ticks per second based on CPU frequency from /proc/cpuinfo.

[samrushing]

I agree with Markiyan, I would have just had it grab the first one... Do systems exist that have different TSC freqs on different CPU's?

[ehuss]

TSC is generally synchronized on modern systems, with a constant frequency. Some older CPUs did not make these guarantees (particularly when the CPU went into different states). It's also possible for BIOS bugs to cause problems, but I do not actually know of any. Intel has been particularly good about keeping it in sync, whereas some AMD CPUs have not.

You can look at this to see what the Linux kernel does to detect if TSC is stable:
http://lxr.linux.no/linux+v2.6.38/arch/x86/kernel/tsc.c#L859

[sgalaburda]

The CPU detection makes the solution more general. It allows to run shrapnel on systems with unsynchronized TSC (provided the CPU affinity is set).