Linux And Feature Creep

Discussion in 'NZ Computing' started by Lawrence D'Oliveiro, Nov 5, 2010.

  1. Phoronix has done some benchmarking on five years’ worth of Linux kernels,
    and found that in most areas, performance has remained constant
    <http://www.zdnet.com/blog/hardware/adding-new-features-to-linux-has-not-affected-core-performance/10245>.

    Open-source developers know how to add new features without losing
    performance.
     
    Lawrence D'Oliveiro, Nov 5, 2010
    #1
    1. Advertising

  2. Lawrence D'Oliveiro

    Sweetpea Guest

    On Fri, 05 Nov 2010 23:00:10 +1300, Lawrence D'Oliveiro wrote:

    > Phoronix has done some benchmarking on five years’ worth of Linux
    > kernels, and found that in most areas, performance has remained constant
    > <http://www.zdnet.com/blog/hardware/adding-new-features-to-linux-has-

    not-affected-core-performance/10245>.

    That's to be expected.

    If adding functionality to a piece of software results in reduced
    performance of existing features or reduced efficiency in any way then
    that is poor software development.


    --
    "Filtering the Internet is like trying to boil the ocean"
     
    Sweetpea, Nov 5, 2010
    #2
    1. Advertising

  3. Lawrence D'Oliveiro

    Enkidu Guest

    On 06/11/10 07:11, Sweetpea wrote:
    > On Fri, 05 Nov 2010 23:00:10 +1300, Lawrence D'Oliveiro wrote:
    >
    >> Phoronix has done some benchmarking on five years’ worth of Linux
    >> kernels, and found that in most areas, performance has remained constant
    >> <http://www.zdnet.com/blog/hardware/adding-new-features-to-linux-has-

    > not-affected-core-performance/10245>.
    >
    > That's to be expected.
    >
    > If adding functionality to a piece of software results in reduced
    > performance of existing features or reduced efficiency in any way then
    > that is poor software development.
    >

    That's a swinging claim, Lennier. What if your client want you to add a
    feature that you know will slow things down, like dynamically flanging
    the gizmots instead of statically flanging them. Would you refuse to
    make such a change?

    Cheers,

    Cliff

    --

    The ends justifies the means - Niccolò di Bernardo dei Machiavelli.

    The end excuses any evil - Sophocles
     
    Enkidu, Nov 6, 2010
    #3
  4. Lawrence D'Oliveiro

    Sweetpea Guest

    On Sat, 06 Nov 2010 14:47:56 +1300, Enkidu wrote:

    > That's a swinging claim, Lennier. What if your client want you to add a
    > feature that you know will slow things down, like dynamically flanging
    > the gizmots instead of statically flanging them. Would you refuse to
    > make such a change?


    My point is that _adding_ a feature should not impact on the performance
    of the other functions of that piece of software.

    In your example, assuming that there are other features in that software
    other than "flanging the gizmots", a deliberate change from statically to
    dynamically "flanging the gizmots" shouldn't affect the performance of
    the part of the application that "compresses the whatzits".


    --
    "Filtering the Internet is like trying to boil the ocean"
     
    Sweetpea, Nov 6, 2010
    #4
  5. Lawrence D'Oliveiro

    Enkidu Guest

    On 06/11/10 17:19, Sweetpea wrote:
    > On Sat, 06 Nov 2010 14:47:56 +1300, Enkidu wrote:
    >
    >> That's a swinging claim, Lennier. What if your client want you to add a
    >> feature that you know will slow things down, like dynamically flanging
    >> the gizmots instead of statically flanging them. Would you refuse to
    >> make such a change?

    >
    > My point is that _adding_ a feature should not impact on the performance
    > of the other functions of that piece of software.
    >
    > In your example, assuming that there are other features in that software
    > other than "flanging the gizmots", a deliberate change from statically to
    > dynamically "flanging the gizmots" shouldn't affect the performance of
    > the part of the application that "compresses the whatzits".
    >

    Sigh! You've never developed software, have you? Changing what you call
    a feature is almost certainly going to affect another feature. For
    instance "compressing the whatzits" might require "flanging the
    gizmots(*)" in which case a change to the flanging process will affect
    the performance of other parts. It's a fact of life of software development.

    (* I realised after I posted that 'gizmots' is mis-spelled. I meant
    common or garden 'gizmos', not some exotic French variety.)

    Cheers,

    Cliff

    --

    The ends justifies the means - Niccolò di Bernardo dei Machiavelli.

    The end excuses any evil - Sophocles
     
    Enkidu, Nov 6, 2010
    #5
  6. Lawrence D'Oliveiro

    Sweetpea Guest

    On Sat, 06 Nov 2010 22:15:13 +1300, Enkidu wrote:

    >> My point is that _adding_ a feature should not impact on the
    >> performance of the other functions of that piece of software.
    >>
    >> In your example, assuming that there are other features in that
    >> software other than "flanging the gizmots", a deliberate change from
    >> statically to dynamically "flanging the gizmots" shouldn't affect the
    >> performance of the part of the application that "compresses the
    >> whatzits".
    >>

    > Sigh! You've never developed software, have you? Changing what you call
    > a feature is almost certainly going to affect another feature. For
    > instance "compressing the whatzits" might require "flanging the
    > gizmots(*)" in which case a change to the flanging process will affect
    > the performance of other parts. It's a fact of life of software
    > development.


    If you believe that adding new features is impossible without negatively
    impacting existing features, would you care to explain how the vast
    majority of new features in the Linux kernel have not impacted at all
    with the performance of existing features?


    --
    "Filtering the Internet is like trying to boil the ocean"
     
    Sweetpea, Nov 6, 2010
    #6
  7. Lawrence D'Oliveiro

    Enkidu Guest

    On 07/11/10 12:16, Sweetpea wrote:
    > On Sat, 06 Nov 2010 22:15:13 +1300, Enkidu wrote:
    >
    >>> My point is that _adding_ a feature should not impact on the
    >>> performance of the other functions of that piece of software.
    >>>
    >>> In your example, assuming that there are other features in that
    >>> software other than "flanging the gizmots", a deliberate change from
    >>> statically to dynamically "flanging the gizmots" shouldn't affect the
    >>> performance of the part of the application that "compresses the
    >>> whatzits".
    >>>

    >> Sigh! You've never developed software, have you? Changing what you call
    >> a feature is almost certainly going to affect another feature. For
    >> instance "compressing the whatzits" might require "flanging the
    >> gizmots(*)" in which case a change to the flanging process will affect
    >> the performance of other parts. It's a fact of life of software
    >> development.

    >
    > If you believe that adding new features is impossible without negatively
    > impacting existing features, would you care to explain how the vast
    > majority of new features in the Linux kernel have not impacted at all
    > with the performance of existing features?
    >

    You don't know that. All you know is that a) new features have been
    added and b) performance has not noticeably changed in the same period.

    It is likely that a) new features would normally have impacted
    performance b) this has been offset by other unrelated improvements that
    have been made to kernel performance.

    Improvements and changes *have* on occasion made the kernel
    significantly slower, but this has been worked on and fixed in later
    releases. I don't remember the kernel versions involved.

    Cheers,

    Cliff

    --

    The ends justifies the means - Niccolò di Bernardo dei Machiavelli.

    The end excuses any evil - Sophocles
     
    Enkidu, Nov 6, 2010
    #7
  8. Lawrence D'Oliveiro

    Sweetpea Guest

    On Sun, 07 Nov 2010 12:28:48 +1300, Enkidu wrote:

    >> If you believe that adding new features is impossible without
    >> negatively impacting existing features, would you care to explain how
    >> the vast majority of new features in the Linux kernel have not impacted
    >> at all with the performance of existing features?
    >>

    > You don't know that. All you know is that a) new features have been
    > added and b) performance has not noticeably changed in the same period.


    FACT: The vast majority of new features added to the Linux kernel have
    not resulted in a release having poorer performance of existing kernel
    features.


    > It is likely that a) new features would normally have impacted
    > performance b) this has been offset by other unrelated improvements that
    > have been made to kernel performance.


    Demonstrably unlikely - as the history of all kernel releases over the
    last 5 years demonstrates that virtually all existing features for
    virtually all releases were not impacted negatively by features added
    into any new release.


    > Improvements and changes *have* on occasion made the kernel
    > significantly slower, but this has been worked on and fixed in later
    > releases. I don't remember the kernel versions involved.


    Demonstrably not true for almost all features in all kernel releases over
    the last 5 years.


    --
    "Filtering the Internet is like trying to boil the ocean"
     
    Sweetpea, Nov 6, 2010
    #8
  9. Lawrence D'Oliveiro

    Enkidu Guest

    On 07/11/10 12:36, Sweetpea wrote:
    > On Sun, 07 Nov 2010 12:28:48 +1300, Enkidu wrote:
    >
    >>> If you believe that adding new features is impossible without
    >>> negatively impacting existing features, would you care to explain how
    >>> the vast majority of new features in the Linux kernel have not impacted
    >>> at all with the performance of existing features?
    >>>

    >> You don't know that. All you know is that a) new features have been
    >> added and b) performance has not noticeably changed in the same period.

    >
    > FACT: The vast majority of new features added to the Linux kernel have
    > not resulted in a release having poorer performance of existing kernel
    > features.
    >

    FACT: You don't know that that isn't because improvements in the kernel
    have offset any performance hits because of new features.
    >
    >> It is likely that a) new features would normally have impacted
    >> performance b) this has been offset by other unrelated improvements that
    >> have been made to kernel performance.

    >
    > Demonstrably unlikely - as the history of all kernel releases over the
    > last 5 years demonstrates that virtually all existing features for
    > virtually all releases were not impacted negatively by features added
    > into any new release.
    >

    Nope, the figures do not show that.
    >
    >> Improvements and changes *have* on occasion made the kernel
    >> significantly slower, but this has been worked on and fixed in later
    >> releases. I don't remember the kernel versions involved.

    >
    > Demonstrably not true for almost all features in all kernel releases over
    > the last 5 years.
    >

    Demonstrably true according to the graphs here:
    http://kernel-perf.sourceforge.net/results.machine_id=15&options=.html

    Cheers,

    Cliff



    --

    The ends justifies the means - Niccolò di Bernardo dei Machiavelli.

    The end excuses any evil - Sophocles
     
    Enkidu, Nov 7, 2010
    #9
  10. Lawrence D'Oliveiro

    Sweetpea Guest

    On Sun, 07 Nov 2010 18:53:41 +1300, Enkidu wrote:

    >>> You don't know that. All you know is that a) new features have been
    >>> added and b) performance has not noticeably changed in the same
    >>> period.

    >>
    >> FACT: The vast majority of new features added to the Linux kernel have
    >> not resulted in a release having poorer performance of existing kernel
    >> features.
    >>

    > FACT: You don't know that that isn't because improvements in the kernel
    > have offset any performance hits because of new features.


    FACT: the performance of features in the kernel has not changed over the
    last 5 years as new features are introduced - as measured from one
    release to another.


    --
    "Filtering the Internet is like trying to boil the ocean"
     
    Sweetpea, Nov 7, 2010
    #10
  11. Lawrence D'Oliveiro

    Enkidu Guest

    On 07/11/10 23:16, Sweetpea wrote:
    > On Sun, 07 Nov 2010 18:53:41 +1300, Enkidu wrote:
    >
    >>>> You don't know that. All you know is that a) new features have been
    >>>> added and b) performance has not noticeably changed in the same
    >>>> period.
    >>>
    >>> FACT: The vast majority of new features added to the Linux kernel have
    >>> not resulted in a release having poorer performance of existing kernel
    >>> features.
    >>>

    >> FACT: You don't know that that isn't because improvements in the kernel
    >> have offset any performance hits because of new features.

    >
    > FACT: the performance of features in the kernel has not changed over the
    > last 5 years as new features are introduced - as measured from one
    > release to another.
    >

    So, you snipped my reference which proves otherwise. From memory and
    looking at those graphs 2.6.27 was a particularly slow one.

    Cheers,

    Cliff

    --

    The ends justifies the means - Niccolò di Bernardo dei Machiavelli.

    The end excuses any evil - Sophocles
     
    Enkidu, Nov 7, 2010
    #11
  12. Lawrence D'Oliveiro

    Sweetpea Guest

    On Mon, 08 Nov 2010 09:12:52 +1300, Enkidu wrote:

    >>>> FACT: The vast majority of new features added to the Linux kernel
    >>>> have not resulted in a release having poorer performance of existing
    >>>> kernel features.
    >>>>
    >>> FACT: You don't know that that isn't because improvements in the
    >>> kernel have offset any performance hits because of new features.

    >>
    >> FACT: the performance of features in the kernel has not changed over
    >> the last 5 years as new features are introduced - as measured from one
    >> release to another.
    >>

    > So, you snipped my reference which proves otherwise. From memory and
    > looking at those graphs 2.6.27 was a particularly slow one.


    Note: I said "The vast majority of new features added to the Linux kernel
    have not resulted in a release having poorer performance of existing
    kernel features."

    What part of "vast majority" don't you understand!! I deliberately didn't
    say "all new features..." because I was aware of that one release. And on
    one test the performance hit hasn't recovered since then.


    --
    "Filtering the Internet is like trying to boil the ocean"
     
    Sweetpea, Nov 8, 2010
    #12
  13. Lawrence D'Oliveiro

    Enkidu Guest

    On 08/11/10 20:26, Sweetpea wrote:
    > On Mon, 08 Nov 2010 09:12:52 +1300, Enkidu wrote:
    >
    >>>>> FACT: The vast majority of new features added to the Linux kernel
    >>>>> have not resulted in a release having poorer performance of existing
    >>>>> kernel features.
    >>>>>
    >>>> FACT: You don't know that that isn't because improvements in the
    >>>> kernel have offset any performance hits because of new features.
    >>>
    >>> FACT: the performance of features in the kernel has not changed over
    >>> the last 5 years as new features are introduced - as measured from one
    >>> release to another.
    >>>

    >> So, you snipped my reference which proves otherwise. From memory and
    >> looking at those graphs 2.6.27 was a particularly slow one.

    >
    > Note: I said "The vast majority of new features added to the Linux kernel
    > have not resulted in a release having poorer performance of existing
    > kernel features."
    >
    > What part of "vast majority" don't you understand!! I deliberately didn't
    > say "all new features..." because I was aware of that one release. And on
    > one test the performance hit hasn't recovered since then.
    >

    All new features added result in slowing down of a program. Adding new
    features increases the total path length of the program. It can't help
    but be slower. Unless you put it on a faster machine.

    Cheers,

    Cliff

    --

    The ends justifies the means - Niccolò di Bernardo dei Machiavelli.

    The end excuses any evil - Sophocles
     
    Enkidu, Nov 9, 2010
    #13
  14. Lawrence D'Oliveiro

    Sweetpea Guest

    On Tue, 09 Nov 2010 19:52:16 +1300, Enkidu wrote:

    >> Note: I said "The vast majority of new features added to the Linux
    >> kernel have not resulted in a release having poorer performance of
    >> existing kernel features."
    >>
    >> What part of "vast majority" don't you understand!! I deliberately
    >> didn't say "all new features..." because I was aware of that one
    >> release. And on one test the performance hit hasn't recovered since
    >> then.
    >>

    > All new features added result in slowing down of a program. Adding new
    > features increases the total path length of the program. It can't help
    > but be slower. Unless you put it on a faster machine.


    Prove it that any new feature added to any program result in existing
    features in said program working slower!

    I think it's been clearly demonstrated to be not true by the article at
    the URL posted by the original poster in this thread!


    --
    "Filtering the Internet is like trying to boil the ocean"
     
    Sweetpea, Nov 9, 2010
    #14
  15. Lawrence D'Oliveiro

    Enkidu Guest

    On 09/11/10 23:01, Sweetpea wrote:
    > On Tue, 09 Nov 2010 19:52:16 +1300, Enkidu wrote:
    >
    >>> Note: I said "The vast majority of new features added to the Linux
    >>> kernel have not resulted in a release having poorer performance of
    >>> existing kernel features."
    >>>
    >>> What part of "vast majority" don't you understand!! I deliberately
    >>> didn't say "all new features..." because I was aware of that one
    >>> release. And on one test the performance hit hasn't recovered since
    >>> then.
    >>>

    >> All new features added result in slowing down of a program. Adding new
    >> features increases the total path length of the program. It can't help
    >> but be slower. Unless you put it on a faster machine.

    >
    > Prove it that any new feature added to any program result in existing
    > features in said program working slower!
    >

    I did. Longer paths = longer execution times.

    Cheers,

    Cliff

    --

    The ends justifies the means - Niccolò di Bernardo dei Machiavelli.

    The end excuses any evil - Sophocles
     
    Enkidu, Nov 9, 2010
    #15
  16. Lawrence D'Oliveiro

    John Little Guest

    On Nov 10, 9:15 am, Enkidu <> wrote:

    > Longer paths = longer execution times.


    Your tongue is surely in cheek. Arguing from a low level, features
    may be added by extra items in a jump table, with no increase in path
    length. The longer jump table may still fit in a cache line, or part
    of an already two-level structure, with no extra loading time unless
    used.

    Regards, John
     
    John Little, Nov 9, 2010
    #16
  17. Lawrence D'Oliveiro

    Enkidu Guest

    On 10/11/10 11:39, John Little wrote:
    > On Nov 10, 9:15 am, Enkidu<> wrote:
    >
    >> Longer paths = longer execution times.

    >
    > Your tongue is surely in cheek.
    >

    Don't tell Lennier!

    Cheers,

    Cliff

    --

    The ends justifies the means - Niccolò di Bernardo dei Machiavelli.

    The end excuses any evil - Sophocles
     
    Enkidu, Nov 10, 2010
    #17
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. sligo

    XP PRO slows to creep.

    sligo, Apr 22, 2005, in forum: Computer Support
    Replies:
    13
    Views:
    603
    Blinky the Shark
    May 10, 2005
  2. Rich

    Intel "Creep Ahead"

    Rich, Jan 3, 2006, in forum: DVD Video
    Replies:
    1
    Views:
    660
  3. nospam

    Can the whole lot of you spam this creep!

    nospam, Nov 30, 2005, in forum: Computer Information
    Replies:
    5
    Views:
    367
  4. Au79

    Feature: Linux free, easy & secure

    Au79, Jul 21, 2006, in forum: Computer Support
    Replies:
    14
    Views:
    584
    William Poaster
    Jul 26, 2006
  5. Have a nice cup of pee

    Linux... yeah linux.. Linux

    Have a nice cup of pee, Apr 12, 2006, in forum: NZ Computing
    Replies:
    19
    Views:
    673
    Bette Noir
    Apr 17, 2006
Loading...

Share This Page