Talk:Jiffy (time)

1779?

I removed this:

The origin of the word is unknown, but it is believed to have first appeared in 1779.

how do you know it appeared in 1779, if the origin is unknown? Hotdogger 23:04, 4 February 2007 (UTC)

That's about the first attestation. The first appearance of a word in writing (usually) tells you nothing about its origin per se, it only provides a terminus ante quem for its coinage. I've changed the phrasing in the article to make this point clearer. --Florian Blaschke (talk) 19:09, 7 June 2012 (UTC)

Original text from jiffie

The following text used was merged into the 'use in computing' section from jiffie:

---

A jiffie is a count of the number of clock interrupts that have occurred. It is not an absolute time interval unit, since its duration depends on the clock interrupt frequency of the particular hardware platform under review.

Within the Linux 2.6 operating system kernel, on the Intel i386 platform, a jiffie is 1 ms, or 1/1000th of a second. It is different on other hardware platforms. For example, on Sparc, it is 10 ms.

Jiffies are used to report timing stats for processes in the /proc (pseudo-)filesystem in the Linux kernel.

Jiffies are no absolute time-value in Linux

As far as i read in the kernel jiffies are dependant on the timer-frequency, which is adjustable. So I suppose to remove any references to exact time-values from the article.

#define JIFFIES_TO_NS(TIME) ((TIME) * (1000000000 / HZ))
#define HZ CONFIG_HZ          /* Internal kernel timer frequency */

(Source: kernel/sched.c, line 81, /include/asm-i386/param.h, line 5 - 2.6.20 kernel) It would be nice, if somebody could verify this and then edit the article.

Yes, the article should keep these compile time defaults though. Widefox; talk 09:53, 25 October 2013 (UTC)

Yes, it is configurable according to the code.

(Copyright issue: Linux source code is under GPL license, so it is allowed to be pasted here.)

Source: kernel/arch/x86/include/asm/param.h

#ifdef __KERNEL__
# define HZ		CONFIG_HZ	/* Internal kernel timer frequency */
# define USER_HZ	100		/* some user interfaces are */
# define CLOCKS_PER_SEC	(USER_HZ)       /* in "ticks" like times() */
#endif

#ifndef HZ
#define HZ 100
#endif

Source: kernel/include/linux/jiffies.h

extern u64 __jiffy_data jiffies_64;
extern unsigned long volatile __jiffy_data jiffies;

--Yejianfei (talk) 07:34, 18 January 2017 (UTC)

See also

• Interrupt

External links

• KernelAnalysis-HOWTO-6 (Be aware that HZ has gone from 100 to 1000 since that article was written.)

---

Simply seeking clarification over the length of a Jiffy offered in the text, as defined in 1981 by Edward Harrison. According to Rudy Rucker's book, "Infinity and the Mind: The Science and Philosophy of the Infinite," copyright 1982, a Jiffy, in quantum mechanics, is roughly equal to 10 to the negative 44 or negative 43rd power. (Paperback, page 320). The author goes on to state that a further discussion of this is also found in Paul Davies, "Other Worlds", (Simon & Schuster, New York, 1980). Omnivata 20:50, 10 May 2006 (UTC)

DUH!!?!?

Someone wrote "DUH!!" on the first part of the article. I have deleted it. Can anyone tell me if this was intentional? --Ztobor 02:29, 30 June 2007 (UTC)

calculation

if the speed of light is 299,792,458 m/s and if a fermi is 1E-15 then shouldn't a jiffy be about 3.34E-24. I never went to college and haven't been in a classroom in 12years, so if I am wrong let me know. Now that is if a fermi is actually 1E-15, I have no way to really confirm this definition. Mantion (talk) 18:53, 10 December 2007 (UTC)

True. Jɪmp 19:01, 29 December 2007 (UTC)

---

I think the statement that a jiffy is the time for light to cross the radius of an electron must be wrong; electrons are not usually considered to have any measureable 'size' or to have zero size. —Preceding unsigned comment added by 82.32.70.77 (talk) 23:56, 17 February 2008 (UTC)

Disambig

These sections should be broken up into different articles. They have little to do with each other except that they go by the same name. ~ Booya ^Bazooka 04:32, 2 October 2008 (UTC)

TODO: Use in Computing - References

References #2 and #3 appear to point at irrelevant chunks of linux configuration declarations/directory listings. Could someone provide more relevant references? (unsigned entry)

I think you refer to #3 and #4. They are references for statements on Linux (Within the Linux 2.6 operating system kernel, since release 2.6.13, [...] and The jiffy values for other Linux versions [...]) so it's not surprising those references are linux minded. In general the article states "[...] is not an absolute time interval unit [...]" and describes why that is so. What kind of reference are you thinking of in that context? Richard 08:02, 29 April 2010 (UTC)

Perhaps a reference which actually mentions the word "jiffie" or "jiffy"? How about that, wouldn't that make sense? I, too, only get directory listings, but not what you quote. You need to link to a specific file, not a directory. --Florian Blaschke (talk) 19:13, 7 June 2012 (UTC)

Alright, it's dawning on me what the links are trying to show. Note, however, that first, primary sources are discouraged on Wikipedia, and not everyone reading this article will be a Linux (or UNIX) connoisseur and thus able to make sense of the code. Documentation in natural language would be much preferrable. Won't somebody please think of the lay readers?! --Florian Blaschke (talk) 19:24, 7 June 2012 (UTC)

Outdated

Hi Richardw. In tickless Linux (e.g. when idle), no tick -> no interrupt -> jiffy count doesn't get updated. jiffy being the time interval between two consecutive ticks. Without incorporating something about tickless, the assertion in the section is not always correct about periodic interrupts - outdated by several years. I was quite surprised I couldn't park the info somewhere better myself, and so tickless kernel redirects to this section (with possibilities). Widefox; talk 11:47, 25 October 2013 (UTC)

As it was, it polluted this article. Maybe you'd better make tickless kernel an article in itself, change some 'is'-s in 'was'-s in and add a remark like currently, some operating systems offer a tickless kernel. Richard 15:06, 25 October 2013 (UTC)

I have prepared this article. I'll leave it up to you to fill the article on tickless kernels. Richard 15:13, 25 October 2013 (UTC)

Those changes (editorialising and introducing a self-link) only blurred the factual info, and didn't address the outdated or dubious tag so I undid them - they're not factual or source based. It seems to me this is a WP:DABCONCEPT (WP:SUMMARYSTYLE) type article, inclusion of an update to the jiffy situation seems OK to me. Without it the info is outdated as tagged. All current major OSes do not have a fixed jiffy time nowadays, so this info is outdated. A redirect with possibilities means it's setup with the hope of splitting off later as I share your concern - I agree slightly tangential for now, but not really WP:OFFTOPIC. "polluting" could be seen by some as a bit WP:OWN - other editors may change this article you know, especially when it is both wrong and outdated.

So, per WP:BRD, what's your objection to [1]

Tickless

The Linux kernel on s390 from 2.6.6 and on i386 from release 2.6.21 can be configured to turn the timer tick off (tickless or dynamic tick) for idle CPUs using CONFIG_NO_HZ, and from 3.10 with CONFIG_NO_HZ_IDLE extended for non-idle processors with CONFIG_NO_HZ_FULL. The XNU kernel in OS X 10.4 on, and the NT kernel in Windows 8 kernel are tickless.

A compromise of including (something) here until split seems in order, as this is a valuable update to jiffy. Widefox; talk 20:26, 29 October 2013 (UTC)

My main concern was and is that that's a different topic. This article is on jiffy in general, and not on tickless kernels, let alone specific implementations and/or configuration parameters. That subject should have its own article. Also, even if some CPUs are idle, at least one CPU will always be active and therefore "ticking". As long as the system is running, there will always be a system timer. But, since that timer is not always hardware "dictated" you could strike that in

In computing, a jiffy is the duration of one tick of the system timer interrupt. It is not an absolute time interval unit, since its duration depends on the clock interrupt frequency of the particular hardware platform.

If, as you say, implementation of tickless CPUs results in a variable jiffy, it could be added to the paragraph you presented above. After "are tickless" you could add something like "In" (some?) "systems with a tickless kernel, the duration of a jiffy is no longer fixed."

Richard 11:56, 30 October 2013 (UTC)

Agree it appears tangential as written above, but is on-topic as it is about jiffy ...I think we're meant to update the article until it's spun out per WP:AVOIDSPLIT. (Possibly what's lacking here is an original article on "jiffy (computing)" so this one can stay summary style.) Jiffy is not used (as originally correctly written as a periodic interrupt) by all current main OSes MS Win, OS X (2007), and Linux (e.g. Fedora from 14 was tickless 2010). It's outdated info. The challenge is to rewrite the update from a jiffy perspective while keeping to the sources, while also providing info on tickless until a split. You may know better than me if it's correct to call it a "variable jiffy", as it depends how the term has changed with the change of software - sources needed. I don't know if it's correct to say the boot CPU 1s tick is a jiffy for instance. The CPU is certainly not active though - it is idle (maximum power saving) for the remainder of the second until the interrupt. ([2] talks about the removal of jiffies.) The point of including the config is twofold 1. both legacy periodic jiffy and modern tickless are possible, and even in tickless there's a periodic timer tick which I presume is set to the original jiffy 2. what we actually mean by jiffy in this article is HZ which is where 1/jiffy is configured at compile time. We can't strike the "hardware" either as it's both hardware and OS dependent. I hope you're not asleep?! Widefox; talk 15:28, 30 October 2013 (UTC)

The fact that it's both hardware and OS dependent is exactly why we can strike it: it's depended on the _platform_ (hardware and / of software).

The website you reference basically says two things about jiffies:

The periodic timer tick is now implemented with a clockevent as well. It does all of the things the old timer-based interrupt did - updating jiffies, accounting CPU time, etc. - but it is run out of the new infrastructure.

and

The implementation leaves room for further development like full tickless systems, where the time slice is controlled by the scheduler, variable frequency profiling, and a complete removal of jiffies in the future (followed by a comment on the removal of jiffies).

So, that website seems to treat jiffies as objects rather than measuring units. I can't imagine what they mean with "updating a time period", just like "removing a mile" makes no sense to me. Redefining a unit is possible, of course, but that's not what they appear to say. Richard 10:55, 31 October 2013 (UTC)

yes we could strike it, but saying it's dependent on both is more informative as it avoids the fallacy that a computer architecture only runs one OS, and also leads us into explaining it is software configurable within the bounds of the particular computer architecture, which has now been replaced by a different paradigm anyhow. - the different OSes may (if the hardware allows it) have different jiffy times, and the same OS (kernel really rather than OS) may allow different jiffy (HZ in the case of Linux). Anyhow, yes, jiffies are the count of ticks since boot. The new infrastructure allows a new paradigm which coming back to the point - although it can be shoehorned into this article (and I believe should according to AVOIDSPLIT)...what I do think, is that the redirect from tickless kernel here should be resolved as this is quite informative for readers given OS X has just implemented timer coalescing so this is about time WP covered this topic. I'm not about to write an article on it, so AVOIDSPLIT would indicate to put it here for now. Maybe more opinions would help, but as I said, this info is outdated until updated. Widefox; talk 11:24, 4 November 2013 (UTC)

In my opinion, by striking "hardware" it is more clear that it is not entirely hardware (anymore). But, as said, that's my opinion.

The article starts with Jiffy is an informal term for any unspecified short period of time. A time period is not a count of something. You can measure a time period by counting something, but if that something doesn't occur at a regular interval, the measured interval has no fixed length and the measurement is therefore essentially useless for timing purposes. So, if and when the clock frequency of the CPU is no longer fixed, you can't rely on that for measuring time. The system timer needs to be predictable as well, and thus can no longer be based on CPU ticks. However, if you have a reliable system timer (based on whatever), you can use that timer as a base for something else - for instance, a jiffy. Richard 11:54, 5 November 2013 (UTC)

References

Point of reference #5?

Could someone explain how citation #5 supports the sentence "Within the Linux operating system kernel, since release 2.6.13, on the Intel i386 platform a jiffy is by default 4 ms, or 1/250 of a second.[5]"? When I clicked the link I was redirected to http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/x86/configs/i386_defconfig (blob 619e7f7426c6d2cf5416cd7179ed5c37c5d0340e) I didn't see "jiffy", "jiffies", "seconds", etc. in the file. However, I did notice that CONFIG_NO_HZ was set (=y), which seems related to the 'tickless operation' being discussed above. Maybe I'm misunderstanding, but it seems that the reference is either wrong or the statement is wrong. — Preceding unsigned comment added by 75.73.249.54 (talk) 14:59, 2 May 2014 (UTC)

I agree, that link and the next one were not helpful (anymore?). Sind they were first added in 2008, someone else came to the same conclusion and added "clarify". I have taken it a step further and removed the entire sentence, which was about an antique Linux version anyhow. Richard 16:38, 2 May 2014 (UTC)

There's nothing wrong with fixing the dead reference link, and updating the info per above. Removing seems drastic, and the incoming redirect I added still doesn't inform on landing, as it did in the undone version discussed above. Widefox; talk 01:35, 19 November 2014 (UTC)

Fermi

Femtometre is the standard (and clear) way to refer to the length also known as "Fermi". I feel like it would be better practice to use that here. JoDu987 (talk) 13:25, 21 July 2023 (UTC)

Jiffy IRQ in Commodore 128

In the article is this: 1980s 8-bit Commodore computers, such as the C64 and C128, had a jiffy of 1/60 second, which was not dependent on the mains AC or video vertical refresh rate.

While that is true for the C-64, it is not true for the C-128 when running in C-128 (“native”) mode.  In native mode, the C-128’s jiffy IRQ is generated by the VIC-II raster interrupt, which occurred at the NTSC or PAL video frame rate, depending on which version of the VIC-II is installed.  In the C-64 or in the C-128 running in C-64 mode, the jiffy IRQ is generated by timer-A in CIA #1.

216.152.18.132 (talk) 18:18, 23 November 2023 (UTC)

 Done --- thanks for the info, so I temporarily removed C128 until references are found. I need references for PET / VIC-20 / C128 / other commodore 6502-based computers. • Sbmeirow • Talk • 00:56, 24 November 2023 (UTC)