Re: [tip:timers/urgent] ktime: Get rid of the union

Re: [tip:timers/urgent] ktime: Get rid of the union

[Peter Zijlstra]

On Sun, Dec 25, 2016 at 05:26:33AM -0800, tip-bot for Thomas Gleixner wrote:
> Commit-ID:  64dac5c9e0f5a933e4520f0cb689b7b143925207
> Gitweb:     http://git.kernel.org/tip/64dac5c9e0f5a933e4520f0cb689b7b143925207
> Author:     Thomas Gleixner <tglx@linutronix.de>
> AuthorDate: Sun, 25 Dec 2016 11:38:40 +0100
> Committer:  Thomas Gleixner <tglx@linutronix.de>
> CommitDate: Sun, 25 Dec 2016 13:11:50 +0100
> 
> ktime: Get rid of the union
> 
> ktime is a union because the initial implementation stored the time in
> scalar nanoseconds on 64 bit machine and in a endianess optimized timespec
> variant for 32bit machines. The Y2038 cleanup removed the timespec variant
> and switched everything to scalar nanoseconds. The union remained, but
> become completely pointless.
> 
> Get rid of the union and just keep ktime_t as simple typedef of type s64.

All good stuff. One question that remains is why keep the type while
removing the cycles_t type?

Re: [tip:timers/urgent] ktime: Get rid of the union

[Thomas Gleixner]

On Tue, 3 Jan 2017, Peter Zijlstra wrote:

> On Sun, Dec 25, 2016 at 05:26:33AM -0800, tip-bot for Thomas Gleixner wrote:
> > Commit-ID:  64dac5c9e0f5a933e4520f0cb689b7b143925207
> > Gitweb:     http://git.kernel.org/tip/64dac5c9e0f5a933e4520f0cb689b7b143925207
> > Author:     Thomas Gleixner <tglx@linutronix.de>
> > AuthorDate: Sun, 25 Dec 2016 11:38:40 +0100
> > Committer:  Thomas Gleixner <tglx@linutronix.de>
> > CommitDate: Sun, 25 Dec 2016 13:11:50 +0100
> > 
> > ktime: Get rid of the union
> > 
> > ktime is a union because the initial implementation stored the time in
> > scalar nanoseconds on 64 bit machine and in a endianess optimized timespec
> > variant for 32bit machines. The Y2038 cleanup removed the timespec variant
> > and switched everything to scalar nanoseconds. The union remained, but
> > become completely pointless.
> > 
> > Get rid of the union and just keep ktime_t as simple typedef of type s64.
> 
> All good stuff. One question that remains is why keep the type while
> removing the cycles_t type?

That would have been a massive surgery which I was not able to pull off on
top of the other changes.

Thanks,

	tglx

Re: [tip:timers/urgent] ktime: Get rid of the union

[Peter Zijlstra]

On Wed, Jan 04, 2017 at 09:54:12AM +0100, Thomas Gleixner wrote:
> On Tue, 3 Jan 2017, Peter Zijlstra wrote:

> > > Get rid of the union and just keep ktime_t as simple typedef of type s64.
> > 
> > All good stuff. One question that remains is why keep the type while
> > removing the cycles_t type?
> 
> That would have been a massive surgery which I was not able to pull off on
> top of the other changes.

And the reason ktime needs be s64 is because 0 is at boot, and we need
to represent time before boot, right? Might want to stick that in a
comment somewhere near that typedef, so I don't keep asking this ;-)

Re: [tip:timers/urgent] ktime: Get rid of the union

[Frederic Weisbecker]

2017-01-04 10:39 GMT+01:00 Peter Zijlstra <peterz@infradead.org>:
> On Wed, Jan 04, 2017 at 09:54:12AM +0100, Thomas Gleixner wrote:
>> On Tue, 3 Jan 2017, Peter Zijlstra wrote:
>
>> > > Get rid of the union and just keep ktime_t as simple typedef of type s64.
>> >
>> > All good stuff. One question that remains is why keep the type while
>> > removing the cycles_t type?
>>
>> That would have been a massive surgery which I was not able to pull off on
>> top of the other changes.
>
> And the reason ktime needs be s64 is because 0 is at boot, and we need
> to represent time before boot, right? Might want to stick that in a
> comment somewhere near that typedef, so I don't keep asking this ;-)

Aaah, that confused me as well :-)