arch/tile: fix __ndelay etc to work better

The current implementations of __ndelay and __udelay call a hypervisor
service to delay, but the hypervisor service isn't actually implemented
very well, and the consensus is that Linux should handle figuring this
out natively and not use a hypervisor service.

By converting nanoseconds to cycles, and then spinning until the
cycle counter reaches the desired cycle, we get several benefits:
first, we are sensitive to the actual clock speed; second, we use
less power by issuing a slow SPR read once every six cycles while
we delay; and third, we properly handle the case of an interrupt by
exiting at the target time rather than after some number of cycles.

Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
This commit is contained in:
Chris Metcalf
2011-02-28 13:21:52 -05:00
parent 04f7a3f12e
commit 1337173148
5 changed files with 34 additions and 11 deletions

View File

@@ -15,20 +15,31 @@
#include <linux/module.h>
#include <linux/delay.h>
#include <linux/thread_info.h>
#include <asm/fixmap.h>
#include <hv/hypervisor.h>
#include <asm/timex.h>
void __udelay(unsigned long usecs)
{
hv_nanosleep(usecs * 1000);
if (usecs > ULONG_MAX / 1000) {
WARN_ON_ONCE(usecs > ULONG_MAX / 1000);
usecs = ULONG_MAX / 1000;
}
__ndelay(usecs * 1000);
}
EXPORT_SYMBOL(__udelay);
void __ndelay(unsigned long nsecs)
{
hv_nanosleep(nsecs);
cycles_t target = get_cycles();
target += ns2cycles(nsecs);
while (get_cycles() < target)
cpu_relax();
}
EXPORT_SYMBOL(__ndelay);
/* FIXME: should be declared in a header somewhere. */
void __delay(unsigned long cycles)
{
cycles_t target = get_cycles() + cycles;
while (get_cycles() < target)
cpu_relax();
}
EXPORT_SYMBOL(__delay);