mirror of
https://github.com/torvalds/linux.git
synced 2024-11-08 21:21:47 +00:00
f927da1786
add sched_clock() hook to paravirtualize sched_clock(). ia64 sched_clock() is based on ar.itc which isn't stable on virtualized environment because vcpu may move around on pcpus. So it needs paravirtualization. Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp> Signed-off-by: Tony Luck <tony.luck@intel.com>
46 lines
1.4 KiB
C
46 lines
1.4 KiB
C
#ifndef _ASM_IA64_TIMEX_H
|
|
#define _ASM_IA64_TIMEX_H
|
|
|
|
/*
|
|
* Copyright (C) 1998-2001, 2003 Hewlett-Packard Co
|
|
* David Mosberger-Tang <davidm@hpl.hp.com>
|
|
*/
|
|
/*
|
|
* 2001/01/18 davidm Removed CLOCK_TICK_RATE. It makes no sense on IA-64.
|
|
* Also removed cacheflush_time as it's entirely unused.
|
|
*/
|
|
|
|
#include <asm/intrinsics.h>
|
|
#include <asm/processor.h>
|
|
|
|
typedef unsigned long cycles_t;
|
|
|
|
extern void (*ia64_udelay)(unsigned long usecs);
|
|
|
|
/*
|
|
* For performance reasons, we don't want to define CLOCK_TICK_TRATE as
|
|
* local_cpu_data->itc_rate. Fortunately, we don't have to, either: according to George
|
|
* Anzinger, 1/CLOCK_TICK_RATE is taken as the resolution of the timer clock. The time
|
|
* calculation assumes that you will use enough of these so that your tick size <= 1/HZ.
|
|
* If the calculation shows that your CLOCK_TICK_RATE can not supply exactly 1/HZ ticks,
|
|
* the actual value is calculated and used to update the wall clock each jiffie. Setting
|
|
* the CLOCK_TICK_RATE to x*HZ insures that the calculation will find no errors. Hence we
|
|
* pick a multiple of HZ which gives us a (totally virtual) CLOCK_TICK_RATE of about
|
|
* 100MHz.
|
|
*/
|
|
#define CLOCK_TICK_RATE (HZ * 100000UL)
|
|
|
|
static inline cycles_t
|
|
get_cycles (void)
|
|
{
|
|
cycles_t ret;
|
|
|
|
ret = ia64_getreg(_IA64_REG_AR_ITC);
|
|
return ret;
|
|
}
|
|
|
|
extern void ia64_cpu_local_tick (void);
|
|
extern unsigned long long ia64_native_sched_clock (void);
|
|
|
|
#endif /* _ASM_IA64_TIMEX_H */
|