[PATCH] sched/clock: cleanup, remove wrap_{max|min}().

[PATCH] sched/clock: cleanup, remove wrap_{max|min}().

[Dongsheng Yang]

I am not sure why we need the wrap_{max|min}() in kernel/sched/clock.c.
But I checked the implementation of max() and min() in linux/kernel.h, I think
we can reuse them here rather than introduce a new function named
wrap_{max|min}().

Signed-off-by: Dongsheng Yang <yangds.fnst@cn.fujitsu.com>
---
 kernel/sched/clock.c | 22 ++++------------------
 1 file changed, 4 insertions(+), 18 deletions(-)

diff --git a/kernel/sched/clock.c b/kernel/sched/clock.c
index 3ef6451..36c03a8 100644
--- a/kernel/sched/clock.c
+++ b/kernel/sched/clock.c
@@ -173,20 +173,6 @@ void sched_clock_init(void)
 }
 
 /*
- * min, max except they take wrapping into account
- */
-
-static inline u64 wrap_min(u64 x, u64 y)
-{
-	return (s64)(x - y) < 0 ? x : y;
-}
-
-static inline u64 wrap_max(u64 x, u64 y)
-{
-	return (s64)(x - y) > 0 ? x : y;
-}
-
-/*
  * update the percpu scd from the raw @now value
  *
  *  - filter out backward motion
@@ -212,11 +198,11 @@ again:
 	 */
 
 	clock = scd->tick_gtod + delta;
-	min_clock = wrap_max(scd->tick_gtod, old_clock);
-	max_clock = wrap_max(old_clock, scd->tick_gtod + TICK_NSEC);
+	min_clock = max(scd->tick_gtod, old_clock);
+	max_clock = max(old_clock, scd->tick_gtod + TICK_NSEC);
 
-	clock = wrap_max(clock, min_clock);
-	clock = wrap_min(clock, max_clock);
+	clock = max(clock, min_clock);
+	clock = min(clock, max_clock);
 
 	if (cmpxchg64(&scd->clock, old_clock, clock) != old_clock)
 		goto again;
-- 
1.8.2.1

Re: [PATCH] sched/clock: cleanup, remove wrap_{max|min}().

[Peter Zijlstra]

[-- Attachment #1: Type: text/plain, Size: 532 bytes --]

On Wed, May 07, 2014 at 08:48:58PM +0900, Dongsheng Yang wrote:
> I am not sure why we need the wrap_{max|min}() in kernel/sched/clock.c.
> But I checked the implementation of max() and min() in linux/kernel.h, I think
> we can reuse them here rather than introduce a new function named
> wrap_{max|min}().

wrap is a good hint there.. they're supposed to deal with the clock
wrapping. Of course 2^64 ns is a rather long time (~584 years in fact),
but that doesn't mean we shouldn't care.

And no, min/max don't do the right thing.

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: [PATCH] sched/clock: cleanup, remove wrap_{max|min}().

[Dongsheng Yang]

On 05/07/2014 09:55 PM, Peter Zijlstra wrote:
> On Wed, May 07, 2014 at 08:48:58PM +0900, Dongsheng Yang wrote:
>> I am not sure why we need the wrap_{max|min}() in kernel/sched/clock.c.
>> But I checked the implementation of max() and min() in linux/kernel.h, I think
>> we can reuse them here rather than introduce a new function named
>> wrap_{max|min}().
> wrap is a good hint there.. they're supposed to deal with the clock
> wrapping. Of course 2^64 ns is a rather long time (~584 years in fact),
> but that doesn't mean we shouldn't care.
>
> And no, min/max don't do the right thing.

IMMO,  max() in kernel.h is checking and returning the u64 with typeof() if
we are passing a u64 parameter to max(). And I checked current callers 
of wrap_max(),
they are all passing parameters of u64 type. So I think max/min works 
well at
the place of wrap_max/min().

Maybe I am missing something here, please correct me if I understand it 
incorrectly.

Thanx

Re: [PATCH] sched/clock: cleanup, remove wrap_{max|min}().

[Peter Zijlstra]

[-- Attachment #1: Type: text/plain, Size: 2233 bytes --]

On Wed, May 07, 2014 at 09:04:19PM +0900, Dongsheng Yang wrote:
> On 05/07/2014 09:55 PM, Peter Zijlstra wrote:
> >On Wed, May 07, 2014 at 08:48:58PM +0900, Dongsheng Yang wrote:
> >>I am not sure why we need the wrap_{max|min}() in kernel/sched/clock.c.
> >>But I checked the implementation of max() and min() in linux/kernel.h, I think
> >>we can reuse them here rather than introduce a new function named
> >>wrap_{max|min}().
> >wrap is a good hint there.. they're supposed to deal with the clock
> >wrapping. Of course 2^64 ns is a rather long time (~584 years in fact),
> >but that doesn't mean we shouldn't care.
> >
> >And no, min/max don't do the right thing.
> 
> IMMO,  max() in kernel.h is checking and returning the u64 with typeof() if
> we are passing a u64 parameter to max(). And I checked current callers of
> wrap_max(),
> they are all passing parameters of u64 type. So I think max/min works well
> at
> the place of wrap_max/min().
> 
> Maybe I am missing something here, please correct me if I understand it
> incorrectly.

Trivial test prog; you really should do these things yourself.

$ ./wrap
wrap_min: 18446744073709551615 wrap_max: 0
min:      0 max:      18446744073709551615


---
#include <stdint.h>
#include <stdio.h>

typedef uint64_t u64;
typedef int64_t  s64;

static inline u64 wrap_min(u64 x, u64 y)
{
        return (s64)(x - y) < 0 ? x : y;
}

static inline u64 wrap_max(u64 x, u64 y)
{
        return (s64)(x - y) > 0 ? x : y;
}

#define min(x, y) ({                            \
        typeof(x) _min1 = (x);                  \
        typeof(y) _min2 = (y);                  \
        (void) (&_min1 == &_min2);              \
        _min1 < _min2 ? _min1 : _min2; })

#define max(x, y) ({                            \
        typeof(x) _max1 = (x);                  \
        typeof(y) _max2 = (y);                  \
        (void) (&_max1 == &_max2);              \
        _max1 > _max2 ? _max1 : _max2; })

void main (void)
{
        u64 a = UINT64_MAX, b = a + 1;

        printf("wrap_min: %lu wrap_max: %lu\n", wrap_min(a, b), wrap_max(a,b));
        printf("min:      %lu max:      %lu\n", min(a, b), max(a,b));
}

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]