time: Introduce CLOCK_REALTIME_COARSE
authorjohn stultz <johnstul@us.ibm.com>
Thu, 20 Aug 2009 02:13:34 +0000 (19:13 -0700)
committerThomas Gleixner <tglx@linutronix.de>
Fri, 21 Aug 2009 19:43:46 +0000 (21:43 +0200)
commitda15cfdae03351c689736f8d142618592e3cebc3
tree497fe3f77e27fa9cf0a484422c7bc382031df1bd
parent8cab02dc3c58a12235c6d463ce684dded9696848
time: Introduce CLOCK_REALTIME_COARSE

After talking with some application writers who want very fast, but not
fine-grained timestamps, I decided to try to implement new clock_ids
to clock_gettime(): CLOCK_REALTIME_COARSE and CLOCK_MONOTONIC_COARSE
which returns the time at the last tick. This is very fast as we don't
have to access any hardware (which can be very painful if you're using
something like the acpi_pm clocksource), and we can even use the vdso
clock_gettime() method to avoid the syscall. The only trade off is you
only get low-res tick grained time resolution.

This isn't a new idea, I know Ingo has a patch in the -rt tree that made
the vsyscall gettimeofday() return coarse grained time when the
vsyscall64 sysctrl was set to 2. However this affects all applications
on a system.

With this method, applications can choose the proper speed/granularity
trade-off for themselves.

Signed-off-by: John Stultz <johnstul@us.ibm.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: nikolag@ca.ibm.com
Cc: Darren Hart <dvhltc@us.ibm.com>
Cc: arjan@infradead.org
Cc: jonathan@jonmasters.org
LKML-Reference: <1250734414.6897.5.camel@localhost.localdomain>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
arch/x86/include/asm/vgtod.h
arch/x86/kernel/vsyscall_64.c
arch/x86/vdso/vclock_gettime.c
include/linux/time.h
kernel/posix-timers.c
kernel/time/timekeeping.c