d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
7d2392141d
freebsd-dev/sys/amd64/include/clock.h

106 lines
2.6 KiB
C
Raw Normal View History

Redo Kernel NTP PLL support, kernel side. This code is mostly taken from the 1.1 port (which was in turn taken from Dave Mills's kern.tar.Z example). A few significant differences: 1) ntp_gettime() is now a MIB variable rather than a system call. A few fiddles are done in libc to make it behave the same. 2) mono_time does not participate in the PLL adjustments. 3) A new interface has been defined (in <machine/clock.h>) for doing possibly machine-dependent things around the time of the clock update. This is used in Pentium kernels to disable interrupts, set `time', and reset the CPU cycle counter as quickly as possible to avoid jitter in microtime(). Measurements show an apparent resolution of a bit more than 8.14usec, which is reasonable given system-call overhead.
1994-09-18 20:40:01 +00:00
/*
* Kernel interface to machine-dependent clock driver.
* Garrett Wollman, September 1994.
* This file is in the public domain.
Fix Pentium CPU rate diagnosis: - Don't print out meaningless iCOMP numbers, those are for droids. - Use a shorter wait to determine clock rate to avoid deficiencies in DELAY(). - Use a fixed-point representation with 8 bits of fraction to store the rate and rationalize the variable name. It would be possible to use even more fraction if it turns out to be worthwhile (I rather doubt it). The question of source code arrangement remains unaddressed.
1995-11-29 19:57:22 +00:00
*
Add wall_cmos_clock sysctl variable, needed to manage adjkerntz even for UTC cmos clocks (needed for Local Timezone FSes)
1996-04-05 03:36:31 +00:00
* $Id: clock.h,v 1.10 1996/03/31 04:05:17 bde Exp $
Redo Kernel NTP PLL support, kernel side. This code is mostly taken from the 1.1 port (which was in turn taken from Dave Mills's kern.tar.Z example). A few significant differences: 1) ntp_gettime() is now a MIB variable rather than a system call. A few fiddles are done in libc to make it behave the same. 2) mono_time does not participate in the PLL adjustments. 3) A new interface has been defined (in <machine/clock.h>) for doing possibly machine-dependent things around the time of the clock update. This is used in Pentium kernels to disable interrupts, set `time', and reset the CPU cycle counter as quickly as possible to avoid jitter in microtime(). Measurements show an apparent resolution of a bit more than 8.14usec, which is reasonable given system-call overhead.
1994-09-18 20:40:01 +00:00
*/
#ifndef _MACHINE_CLOCK_H_
Declare the full uglyness of the interfaces to the clock driver (except things declared in machine-independent files).
1994-11-05 22:51:17 +00:00
#define _MACHINE_CLOCK_H_
Redo Kernel NTP PLL support, kernel side. This code is mostly taken from the 1.1 port (which was in turn taken from Dave Mills's kern.tar.Z example). A few significant differences: 1) ntp_gettime() is now a MIB variable rather than a system call. A few fiddles are done in libc to make it behave the same. 2) mono_time does not participate in the PLL adjustments. 3) A new interface has been defined (in <machine/clock.h>) for doing possibly machine-dependent things around the time of the clock update. This is used in Pentium kernels to disable interrupts, set `time', and reset the CPU cycle counter as quickly as possible to avoid jitter in microtime(). Measurements show an apparent resolution of a bit more than 8.14usec, which is reasonable given system-call overhead.
1994-09-18 20:40:01 +00:00
Add Pentium Pro CPU detection and special handling. For now, all the optimizations we have for 586s also apply to 686s...this will be fine- tuned in the future as appropriate.
1995-12-24 08:10:52 +00:00
#if defined(I586_CPU) || defined(I686_CPU)
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
#define I586_CYCLECTR(x) \
__asm __volatile(".byte 0x0f, 0x31" : "=A" (x))
/*
* When we update the clock, we also update this bias value which is
* automatically subtracted in microtime(). We assume that CPU_THISTICKLEN()
* has been called at some point in the past, so that an appropriate value is
* set up in i586_last_tick. (This works even if we are not being called
* from hardclock because hardclock will have run before and will made the
* call.)
*/
Redo Kernel NTP PLL support, kernel side. This code is mostly taken from the 1.1 port (which was in turn taken from Dave Mills's kern.tar.Z example). A few significant differences: 1) ntp_gettime() is now a MIB variable rather than a system call. A few fiddles are done in libc to make it behave the same. 2) mono_time does not participate in the PLL adjustments. 3) A new interface has been defined (in <machine/clock.h>) for doing possibly machine-dependent things around the time of the clock update. This is used in Pentium kernels to disable interrupts, set `time', and reset the CPU cycle counter as quickly as possible to avoid jitter in microtime(). Measurements show an apparent resolution of a bit more than 8.14usec, which is reasonable given system-call overhead.
1994-09-18 20:40:01 +00:00
#define CPU_CLOCKUPDATE(otime, ntime) \
do { \
Fix Pentium CPU rate diagnosis: - Don't print out meaningless iCOMP numbers, those are for droids. - Use a shorter wait to determine clock rate to avoid deficiencies in DELAY(). - Use a fixed-point representation with 8 bits of fraction to store the rate and rationalize the variable name. It would be possible to use even more fraction if it turns out to be worthwhile (I rather doubt it). The question of source code arrangement remains unaddressed.
1995-11-29 19:57:22 +00:00
if(i586_ctr_rate) { \
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
disable_intr(); \
i586_ctr_bias = i586_last_tick; \
*(otime) = *(ntime); \
enable_intr(); \
Redo Kernel NTP PLL support, kernel side. This code is mostly taken from the 1.1 port (which was in turn taken from Dave Mills's kern.tar.Z example). A few significant differences: 1) ntp_gettime() is now a MIB variable rather than a system call. A few fiddles are done in libc to make it behave the same. 2) mono_time does not participate in the PLL adjustments. 3) A new interface has been defined (in <machine/clock.h>) for doing possibly machine-dependent things around the time of the clock update. This is used in Pentium kernels to disable interrupts, set `time', and reset the CPU cycle counter as quickly as possible to avoid jitter in microtime(). Measurements show an apparent resolution of a bit more than 8.14usec, which is reasonable given system-call overhead.
1994-09-18 20:40:01 +00:00
} else { \
*(otime) = *(ntime); \
} \
} while(0)
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
#define CPU_THISTICKLEN(dflt) cpu_thisticklen(dflt)
Redo Kernel NTP PLL support, kernel side. This code is mostly taken from the 1.1 port (which was in turn taken from Dave Mills's kern.tar.Z example). A few significant differences: 1) ntp_gettime() is now a MIB variable rather than a system call. A few fiddles are done in libc to make it behave the same. 2) mono_time does not participate in the PLL adjustments. 3) A new interface has been defined (in <machine/clock.h>) for doing possibly machine-dependent things around the time of the clock update. This is used in Pentium kernels to disable interrupts, set `time', and reset the CPU cycle counter as quickly as possible to avoid jitter in microtime(). Measurements show an apparent resolution of a bit more than 8.14usec, which is reasonable given system-call overhead.
1994-09-18 20:40:01 +00:00
#else
#define CPU_CLOCKUPDATE(otime, ntime) \
(*(otime) = *(ntime))
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
#define CPU_THISTICKLEN(dflt) dflt
Redo Kernel NTP PLL support, kernel side. This code is mostly taken from the 1.1 port (which was in turn taken from Dave Mills's kern.tar.Z example). A few significant differences: 1) ntp_gettime() is now a MIB variable rather than a system call. A few fiddles are done in libc to make it behave the same. 2) mono_time does not participate in the PLL adjustments. 3) A new interface has been defined (in <machine/clock.h>) for doing possibly machine-dependent things around the time of the clock update. This is used in Pentium kernels to disable interrupts, set `time', and reset the CPU cycle counter as quickly as possible to avoid jitter in microtime(). Measurements show an apparent resolution of a bit more than 8.14usec, which is reasonable given system-call overhead.
1994-09-18 20:40:01 +00:00
#endif
Fix Pentium CPU rate diagnosis: - Don't print out meaningless iCOMP numbers, those are for droids. - Use a shorter wait to determine clock rate to avoid deficiencies in DELAY(). - Use a fixed-point representation with 8 bits of fraction to store the rate and rationalize the variable name. It would be possible to use even more fraction if it turns out to be worthwhile (I rather doubt it). The question of source code arrangement remains unaddressed.
1995-11-29 19:57:22 +00:00
#define I586_CTR_RATE_SHIFT 8
Declare the full uglyness of the interfaces to the clock driver (except things declared in machine-independent files).
1994-11-05 22:51:17 +00:00
#if defined(KERNEL) && !defined(LOCORE)
#include <sys/cdefs.h>
#include <machine/frame.h>
/*
* i386 to clock driver interface.
* XXX almost all of it is misplaced. i586 stuff is done in isa/clock.c
* and isa stuff is done in i386/microtime.s and i386/support.s.
*/
extern int adjkerntz;
extern int disable_rtc_set;
Add wall_cmos_clock sysctl variable, needed to manage adjkerntz even for UTC cmos clocks (needed for Local Timezone FSes)
1996-04-05 03:36:31 +00:00
extern int wall_cmos_clock;
Add Pentium Pro CPU detection and special handling. For now, all the optimizations we have for 586s also apply to 686s...this will be fine- tuned in the future as appropriate.
1995-12-24 08:10:52 +00:00
#if defined(I586_CPU) || defined(I686_CPU)
Fix Pentium CPU rate diagnosis: - Don't print out meaningless iCOMP numbers, those are for droids. - Use a shorter wait to determine clock rate to avoid deficiencies in DELAY(). - Use a fixed-point representation with 8 bits of fraction to store the rate and rationalize the variable name. It would be possible to use even more fraction if it turns out to be worthwhile (I rather doubt it). The question of source code arrangement remains unaddressed.
1995-11-29 19:57:22 +00:00
extern unsigned i586_ctr_rate; /* fixed point */
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
extern long long i586_last_tick;
extern long long i586_ctr_bias;
No longer use the cyclecounter to attempt to correct for late or missed clock interrupts. Keep a 1-in-16 smoothed average of the length of each tick. If the CPU speed is correctly diagnosed, this should give experienced users enough information to figure out a more suitable value for `tick'.
1996-01-30 18:56:47 +00:00
extern unsigned long i586_avg_tick;
Declare the full uglyness of the interfaces to the clock driver (except things declared in machine-independent files).
1994-11-05 22:51:17 +00:00
#endif
extern int timer0_max_count;
extern u_int timer0_overflow_threshold;
extern u_int timer0_prescaler_count;
Add Pentium Pro CPU detection and special handling. For now, all the optimizations we have for 586s also apply to 686s...this will be fine- tuned in the future as appropriate.
1995-12-24 08:10:52 +00:00
#if defined(I586_CPU) || defined(I686_CPU)
Declare the full uglyness of the interfaces to the clock driver (except things declared in machine-independent files).
1994-11-05 22:51:17 +00:00
void calibrate_cyclecounter __P((void));
#endif
Add Pentium Pro CPU detection and special handling. For now, all the optimizations we have for 586s also apply to 686s...this will be fine- tuned in the future as appropriate.
1995-12-24 08:10:52 +00:00
#if defined(I586_CPU) || defined(I686_CPU)
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
static __inline u_long
cpu_thisticklen(u_long dflt)
{
long long old;
No longer use the cyclecounter to attempt to correct for late or missed clock interrupts. Keep a 1-in-16 smoothed average of the length of each tick. If the CPU speed is correctly diagnosed, this should give experienced users enough information to figure out a more suitable value for `tick'.
1996-01-30 18:56:47 +00:00
long len;
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
Fix Pentium CPU rate diagnosis: - Don't print out meaningless iCOMP numbers, those are for droids. - Use a shorter wait to determine clock rate to avoid deficiencies in DELAY(). - Use a fixed-point representation with 8 bits of fraction to store the rate and rationalize the variable name. It would be possible to use even more fraction if it turns out to be worthwhile (I rather doubt it). The question of source code arrangement remains unaddressed.
1995-11-29 19:57:22 +00:00
if (i586_ctr_rate) {
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
old = i586_last_tick;
I586_CYCLECTR(i586_last_tick);
No longer use the cyclecounter to attempt to correct for late or missed clock interrupts. Keep a 1-in-16 smoothed average of the length of each tick. If the CPU speed is correctly diagnosed, this should give experienced users enough information to figure out a more suitable value for `tick'.
1996-01-30 18:56:47 +00:00
len = ((i586_last_tick - old) << I586_CTR_RATE_SHIFT)
Fix Pentium CPU rate diagnosis: - Don't print out meaningless iCOMP numbers, those are for droids. - Use a shorter wait to determine clock rate to avoid deficiencies in DELAY(). - Use a fixed-point representation with 8 bits of fraction to store the rate and rationalize the variable name. It would be possible to use even more fraction if it turns out to be worthwhile (I rather doubt it). The question of source code arrangement remains unaddressed.
1995-11-29 19:57:22 +00:00
/ i586_ctr_rate;
No longer use the cyclecounter to attempt to correct for late or missed clock interrupts. Keep a 1-in-16 smoothed average of the length of each tick. If the CPU speed is correctly diagnosed, this should give experienced users enough information to figure out a more suitable value for `tick'.
1996-01-30 18:56:47 +00:00
i586_avg_tick = i586_avg_tick * 15 / 16 + len / 16;
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
}
No longer use the cyclecounter to attempt to correct for late or missed clock interrupts. Keep a 1-in-16 smoothed average of the length of each tick. If the CPU speed is correctly diagnosed, this should give experienced users enough information to figure out a more suitable value for `tick'.
1996-01-30 18:56:47 +00:00
return dflt;
Reduce jitter of Pentium microtime() implementation by letting the counter free-run and doing a subtract in microtime() rather than resetting the counter to zero at every clock tick. In combination with the changes to kern_clock.c, this should eliminate all the immediately obvious sources of systematic jitter in timekeeping on Pentium machines.
1995-10-12 20:39:49 +00:00
}
#endif
Declare the full uglyness of the interfaces to the clock driver (except things declared in machine-independent files).
1994-11-05 22:51:17 +00:00
/*
* Driver to clock driver interface.
*/
void DELAY __P((int usec));
int acquire_timer0 __P((int rate,
void (*function)(struct clockframe *frame)));
int acquire_timer2 __P((int mode));
int release_timer0 __P((void));
int release_timer2 __P((void));
Moved rtcin() to clock.c. Always delay using one inb(0x84) after each i/o in rtcin() - don't do this conditional on the bogus option DUMMY_NOPS not being defined. If you want an optionally slightly faster rtcin() again, then inline it and use a better named option or sysctl variable. It only needs to be fast in rtcintr().
1996-03-31 04:05:36 +00:00
int rtcin __P((int val));
Declare the full uglyness of the interfaces to the clock driver (except things declared in machine-independent files).
1994-11-05 22:51:17 +00:00
int sysbeep __P((int pitch, int period));
#endif /* KERNEL && !LOCORE */
#endif /* !_MACHINE_CLOCK_H_ */
Reference in New Issue Copy Permalink