Announcement

Collapse
No announcement yet.

SphinUX OS Claims To Be ~150% Faster Than GNU/Linux

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • kraftman
    replied
    Originally posted by mrugiero View Post
    Actually, you said it's useless crap...
    [/QUOTE]

    I said this about Haiku not about multi threaded GUI.

    Leave a comment:


  • mrugiero
    replied
    Originally posted by kraftman View Post
    Qnx isn't a toy and I wasn't talking about Qnx. When I said small operating systems I meant toys like Haiku or BeOS. There are embedded Linux versions which are also very small, but they're not the toys as well. My original point was size doesn't matter, but design.
    Haiku is a toy (hobby) OS, BeOS is just an obsolete real use OS. In its moment was of marginal use, but it was actually a professional project, and was modern (of course, most of its features were rendered obsolete with time, due to its commercial failure).

    Originally posted by kraftman View Post
    I've never said it's bad (but I'm not saying it's good, because there are reasons why bigger operating systems aren't using same thing).
    Actually, you said it's useless crap...
    [/QUOTE]

    Leave a comment:


  • sdack
    replied
    Half of this thread is proof of why Phoronix should not have published their article. It is sensationalism and unprofessional. I blame the summer heat.

    Leave a comment:


  • ciplogic
    replied
    Originally posted by kraftman View Post
    Great, but why you're still talking about QNX? It's a different level and I wouldn't put it next to toys like Haiku. Furthermore, it should be compared to RTLinux not generic kernel.
    (...)
    When the benchmark is valid one I have nothing against it. For example Intel video driver is faster in Windows and that's ok. However, when some idiots are doing benchmarks and draw stupid conclusions it's something I cannot agree with.
    I agree with you, there are many true facts (like any toy OS cannot beat any "big" OS, by this I think you mean mature, well established and well supported, and full featured). Similarly there are people doing it better or worse, but it is very unlikely excluding for trivial "benchmarks" when some guys in garage beat a corporation which cares about performance.

    So what it remains to settle from your side, is to give quotations when I was "blaming" Qt or Java. I tried to support most of points of views with goods and bads. The "worst" things I found myself close, it was to benchmark multiple applications startup, when someone said something like that "Mono is a pig" in startup, and Qt is so good - http://www.youtube.com/watch?v=cH-VrnIV-rs Other topic (again, I will let you to find it), I remember that there were applications written in C# which were more lightweight than their C++ equivalent (SharpDevelop vs Visual Studio), and I remember I specified that at least on my machine.

    But anyway, the burden is on you to say where I was criticize Qt (unfairly). Or maybe Java.
    Last edited by ciplogic; 10 June 2013, 02:34 PM.

    Leave a comment:


  • kraftman
    replied
    Originally posted by ciplogic View Post
    I didn't. If you look, at least on low-spec hardware (the strongest CPUs out there), QNX has a showing (the license of the articles doesn't let me post numbers or such). It is sometimes even faster (than Linux). But you didn't discuss about "OSes that are toys" compared with "OSes that are not toys", but the "big OSes" compared with "small OSes", right?
    Great, but why you're still talking about QNX? It's a different level and I wouldn't put it next to toys like Haiku. Furthermore, it should be compared to RTLinux not generic kernel.

    You said: "They've used one of the slowest Linux file systems and ufs2 in BSD which is comparable to Ext2 which doesn't provide nearly any protection. " Did you read the link/article, or just the numbers?
    I have only looked at the numbers. I wouldn't ever suppose they've made a compile time benchmark and draw conclusions about locking. Those guys are simply clueless and they're just proving how incompetent they are. They've made a compile time benchmark which only means Haiku compiles faster on some OS configuration. They were thinking BSD folks made locking optimizations and that was the reason - in their opinion - why Haiku compiled faster out there. While it was just a file system which made a difference. It doesn't matter if ReiserFS was the fastest on Linux in this task, because it's still much different than UFS. I bet they didn't even try Ext2 which would make comparison more fair. It also seems they didn't provide what package versions were used which could also have some impact on compilation time.

    About your definition of better/useless (which you seem to adhere), when you find a benchmark where Linux loses, you think that is always the fault of the benchmark? Or you think that people must misconfigure Linux? When Windows wins in video drivers benchmarks (and in games) does it make Linux useless for games?
    When the benchmark is valid one I have nothing against it. For example Intel video driver is faster in Windows and that's ok. However, when some idiots are doing benchmarks and draw stupid conclusions it's something I cannot agree with.

    No, the impact is in the fact that BeOS (kernel) was already built with GCC 2.x compared with a GCC 4.x compiled Linux kernel. This means that more optimizations are applied in the first place to the entire Linux kernel. Many would not matter, but maybe some would.
    That's true, but when I was writing about this I thought it wasn't a compile time benchmark, but disk performance one. However, a different file system will still make a much bigger difference than compiler.

    Leave a comment:


  • ciplogic
    replied
    Originally posted by kraftman View Post
    Qnx isn't a toy and I wasn't talking about Qnx. When I said small operating systems I meant toys like Haiku or BeOS. There are embedded Linux versions which are also very small, but they're not the toys as well. My original point was size doesn't matter, but design.
    (...)
    It seems you're mixing multi-core performance with real time operating systems. You're simply amazing.
    I didn't. If you look, at least on low-spec hardware (the strongest CPUs out there), QNX has a showing (the license of the articles doesn't let me post numbers or such). It is sometimes even faster (than Linux). But you didn't discuss about "OSes that are toys" compared with "OSes that are not toys", but the "big OSes" compared with "small OSes", right?


    Originally posted by kraftman View Post
    Yay! They've used one of the slowest Linux file systems and ufs2 in BSD which is comparable to Ext2 which doesn't provide nearly any protection. And.. Haiku is indeed the slowest out there! You're simply a genius. Furthermore, there were many more benchmarks showing Linux outperforms BSD in 2010. Just check out Phoronix:

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite


    So, according to my logic nobody should be using BSD. Keep in mind Ubuntu had debugging enabled, so it makes BSD to look even worse. Facts are against you Ciplogic. You were doing much better in the past.
    You said: "They've used one of the slowest Linux file systems and ufs2 in BSD which is comparable to Ext2 which doesn't provide nearly any protection. " Did you read the link/article, or just the numbers?

    The article stated explicitly that it uses the:
    For Linux I had initialized the partition with ReiserFS 3.6, since that is supposedly still the fastest file system for compiling Haiku. It is also the only Linux file system that offers xattr support well enough to be usable by the Haiku build process to store the many custom file attributes.
    So, in comparison to what someone is writing here, the benchmarks (at least for compiling Haiku with) were as good as they could be for Linux. If you read at a deeper level you could find that many slow parts were improved inside Haiku so it was faster (at least in this benchmark) like 5x that it was before. 2x slower than FreeBSD/ 1.8 x times slower than Linux!? Who cares? (you do, but out of this discussion)

    Did you try other more up-to-date benchmarks of FreeBSD? Which do not "use debug" for Ubuntu? Like this one: http://www.phoronix.com/scan.php?pag..._preview&num=2

    Yes, Linux still wins, but not by such of a big margin, isn't it?

    About your definition of better/useless (which you seem to adhere), when you find a benchmark where Linux loses, you think that is always the fault of the benchmark? Or you think that people must misconfigure Linux? When Windows wins in video drivers benchmarks (and in games) does it make Linux useless for games?


    Originally posted by kraftman View Post
    PS. I doubt if there's any impact on jam when comes to different GCC versions.
    No, the impact is in the fact that BeOS (kernel) was already built with GCC 2.x compared with a GCC 4.x compiled Linux kernel. This means that more optimizations are applied in the first place to the entire Linux kernel. Many would not matter, but maybe some would.

    Leave a comment:


  • kraftman
    replied
    Originally posted by ciplogic View Post
    One Tiny OS is Qnx, and some cars are using it: "The QNX CAR Application Platform is running in over 20 million vehicles as of mid-2011". Of course, it runs as BB10 too.
    Qnx isn't a toy and I wasn't talking about Qnx. When I said small operating systems I meant toys like Haiku or BeOS. There are embedded Linux versions which are also very small, but they're not the toys as well. My original point was size doesn't matter, but design.

    And as you may say: "I do care just about multi-core performance", of course there is a citation about the performance:
    http://onqpl.blogspot.com/2011/10/qn...enchmarks.html
    It seems you're mixing multi-core performance with real time operating systems. You're simply amazing.

    Haiku/BeOS has a clearly described open architecture (read it here, is free: http://oreilly.com/openbook/beosprog/book/index.html ) When you dismiss others' views because you didn't bother to fact check, says something about your willingness to improve yourself. In my best understanding, BeOs has for every important system a server (similar with X11, or the print spooler, etc.) and applications are not blocking any of these servers (when in X11/Windows preVista you can), so when you draw something, and it is CPU intensive, this will not freeze your UI, and you can move still the windows, etc. This gives at least the feeling of being very responsive about BeOS (or Haiku).
    I've never said it's bad (but I'm not saying it's good, because there are reasons why bigger operating systems aren't using same thing).

    So let's use a citation:

    Did anyone said anything before you about BeOS, Haiku or DOS? Mark45 simply stated about space/speed tradeoff.
    Ok, I have to admit I was mistaken here, so I apologize.

    If you talk about BS, Haiku (nightly) as far as it is stated here (and I trust as being a higher chance of being true, compared with most of what you say), uses GCC 4.7 so the code on the same CPU that does a lot of computations would run (close to) identical on Linux and Haiku. You said: "Haiku, ... (is) much slower than bigger operating systems". If you move the goalpost and state: "it is slower in database benchmarks" or anything of this sort (which was again, not so clear in your initial statement), and you will try to look for a specific benchmark, maybe you will find this link:
    In these exciting times, during which Ingo Weinhold is making great progress with some performance optimizations in the Haiku kernel, I felt this strong urge to conduct some benchmark results, even if that caused me great deal of pain in setting up …



    It looks that at least around year of glory 2010, you should not use your beloved Linux box but FreeBSD, as Linux "is useless" (as someone told you, in your logic, the opposite to better is useless).

    For benchmarks, Haiku is using the outdated GCC 2.xx line, so the comparison may not be that fair.

    Enjoy doing acrobatics with the facts kraftman
    Yay! They've used one of the slowest Linux file systems and ufs2 in BSD which is comparable to Ext2 which doesn't provide nearly any protection. And.. Haiku is indeed the slowest out there! You're simply a genius. Furthermore, there were many more benchmarks showing Linux outperforms BSD in 2010. Just check out Phoronix:

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite


    So, according to my logic nobody should be using BSD. Keep in mind Ubuntu had debugging enabled, so it makes BSD to look even worse. Facts are against you Ciplogic. You were doing much better in the past. PS. I doubt if there's any impact on jam when comes to different GCC versions.

    Leave a comment:


  • ciplogic
    replied
    Originally posted by kraftman View Post
    There's no need for any citations. None of these tiny OSes is being used in serious environments where performance matters.
    One Tiny OS is Qnx, and some cars are using it: "The QNX CAR Application Platform is running in over 20 million vehicles as of mid-2011". Of course, it runs as BB10 too.

    The reason for quotations has to be understood fairly easily: many thing come with "feelings", and less about measure.

    Originally posted by kraftman View Post
    However, it was someone that has claimed smaller systems are faster, so it's him that has to provide proofs.
    And as you may say: "I do care just about multi-core performance", of course there is a citation about the performance:
    A blog about QNX, QNX CAR, automotive, embedded systems, infotainment, M2M, medical devices, PlayBook, RIM, safety-critical software, and telematics.

    And the relevant quote (for me):
    The performance benchmarks, which are especially rigorous, gauge whether an OS has what it takes for applications that have zero tolerance for missed deadlines. These tests include thread-switch latency and interrupt latency.
    Originally posted by kraftman View Post
    When comes to BeOS or Haiku optimizations it was mentioned many times on Phoronix. Some other people would explain this better to you, but as far as I remember it's related to multi threaded GUI.
    Haiku/BeOS has a clearly described open architecture (read it here, is free: http://oreilly.com/openbook/beosprog/book/index.html ) When you dismiss others' views because you didn't bother to fact check, says something about your willingness to improve yourself. In my best understanding, BeOs has for every important system a server (similar with X11, or the print spooler, etc.) and applications are not blocking any of these servers (when in X11/Windows preVista you can), so when you draw something, and it is CPU intensive, this will not freeze your UI, and you can move still the windows, etc. This gives at least the feeling of being very responsive about BeOS (or Haiku).

    Originally posted by kraftman View Post
    My point wasn't off topic, because I was talking about SpinUX not about BeOS or Haiku alone.
    So let's use a citation:
    Originally posted by kraftman View Post
    I didn't hear such bullshit since a long time. BeOS, Haiku, DOS are much slower than bigger operating systems. It's because of architecture not because of being small or not. Furthermore, Linux doesn't use 90% of its features on a common PC, so your point is invalid.
    Did anyone said anything before you about BeOS, Haiku or DOS? Mark45 simply stated about space/speed tradeoff.

    If you talk about BS, Haiku (nightly) as far as it is stated here (and I trust as being a higher chance of being true, compared with most of what you say), uses GCC 4.7 so the code on the same CPU that does a lot of computations would run (close to) identical on Linux and Haiku. You said: "Haiku, ... (is) much slower than bigger operating systems". If you move the goalpost and state: "it is slower in database benchmarks" or anything of this sort (which was again, not so clear in your initial statement), and you will try to look for a specific benchmark, maybe you will find this link:
    In these exciting times, during which Ingo Weinhold is making great progress with some performance optimizations in the Haiku kernel, I felt this strong urge to conduct some benchmark results, even if that caused me great deal of pain in setting up …


    FreeBSD 8.0:
    real 11m53.918s
    user 17m11.611s
    sys 2m39.864s
    (713.9 seconds)

    Linux 2.6.31:
    real 13m32.431s
    user 17m10.099s
    sys 2m49.717s
    (812.4 seconds)

    OpenSolaris 2009.06:
    real 14m20.792s
    user 18m36.871s
    sys 5m39.549s
    (860.8 seconds)

    Haiku r35024:
    real 17m18.436s
    user 27m22.108s
    sys 5m0.447s
    (1038.4 seconds)
    It looks that at least around year of glory 2010, you should not use your beloved Linux box but FreeBSD, as Linux "is useless" (as someone told you, in your logic, the opposite to better is useless).

    For benchmarks, Haiku is using the outdated GCC 2.xx line, so the comparison may not be that fair.

    Enjoy doing acrobatics with the facts kraftman
    Last edited by ciplogic; 10 June 2013, 08:47 AM.

    Leave a comment:


  • ciplogic
    replied
    Originally posted by kraftman View Post
    (...)Said the one who was blaming java and qt in past. It seems you don't follow your own rules.
    Where? I would love specifics

    Let me look for some quotes I wrote about Java:
    Here :
    Source: http://phoronix.com/forums/showthrea...ht=#post325362

    At last, you seem to know to insult everyone, and I think that Java fans out-there would say that are many instances when Java is faster than C: http://scribblethink.org/Computer/javaCbenchmark.html
    Or this quote:
    I know that NBody (in the your shown case) is not multi-core aware, but even the previous month at work, I had some native code and as it had to do a massive processing task (basically to crack passwords), moving into multi-core was done much easier with Java code and at the end, just using basically "Java -server" and by using all cores it was running basically 6-8 times faster. Of course 1 core C++ vs 4 HT (4 x 2) cores Java.
    Maybe I was bashing Qt !?:

    I have no issues with Qt per-se, but the development cycle in C# is most of the times faster than the QMake cycle in Qt. If you can afford Resharper (or CodeRush) you will get static code analysis that warrants higher quality code.

    About Qt and issues I faced:
    - MSSQL uses a sub-par database driver (http://doc.qt.digia.com/stable/sql-driver.html)
    - also Qt is slower in calling signals than C# code and sometimes it has issues with thread safety: http://stackoverflow.com/questions/1...nd-performance
    - if you use a Metacity Window Manager (at least in Qt 4.4 cycle) there were bugs in creating Windows. Qt was stested against KWM mostly (I had issues at my work regarding those issues)
    Maybe here?


    What Qt offers and Mono doesn't?
    - if we talk about the runtime part, I can say that Mono offers more:
    + GC: simply means that you write simpler algorithms and you don't look for reference counted cycles
    + dynamic runtime with inline caching ( http://stackoverflow.com/a/7478557 ),
    + very easy to embed plugins (compare with the QtPlugin, or with dlopen/LoadLibrary mess),
    + a very easy way to do network remoting, Xml processing, class & fied annotations that can work with reflection to automate configuration in complex systems
    What Qt offers easily:
    + C++ integration with foreign codebases (that are old ones)
    + Two paradigms for controls: QWidget and Qml that are both working cross platform
    + arguably a bit better performance
    PyQt offers most of Mono (excluding the toolkit part) with a performance hit if you don't use PyPy and no static checking (meanining performance hit and bugs in typing).
    So, may you say the post where was I "blaming"? Even I did, for some factual items, certainly I never use a sub-par language (and I don't mean Java compared with C# ) like the following quote:

    You:
    Mono on Linux is desirable for m$ only. Linux will become the main platform when it gets games, so m$ wants to pollute it with mono. Ever heard of m$ patents FUD? Did you ever wondered where those 200+ patents violations lay? They lay in mono. Just say fu*k you Icaza, m$ and have a nice day.
    Read the answer:
    Originally posted by directhex View Post
    ...
    - did your butt hurt? It seem that when it was about getting factual, you dissipated like a smoke

    Leave a comment:


  • mrugiero
    replied
    To kraftman. Before reading my answers, keep in mind this:
    I don't care about Haiku. I think it's a nice hobby project, but I don't really see what would be its niche, and I don't follow or defend it. Even though, I've found some arguable statements in your posts.

    Originally posted by kraftman View Post
    There's no need for any citations. None of these tiny OSes is being used in serious environments where performance matters. However, it was someone that has claimed smaller systems are faster, so it's him that has to provide proofs. When comes to BeOS or Haiku optimizations it was mentioned many times on Phoronix. Some other people would explain this better to you, but as far as I remember it's related to multi threaded GUI.
    You have no idea how fallacious your first statement is. No one, read, NOBODY will use in any production scenario (be it or not one particular one where performance matters) any kind of software tagged as alpha or beta, independently of how fast or slow it might be, because that tag implies it's unstable and might even be unsecure.

    Originally posted by kraftman View Post
    It's useless, because nearly any other Operating System will do same job and in most cases it will probably do it better (more available apps, better performance and hardware support).
    The opposite of better is not useless, is worse, you know...
    Maybe Linux can do better, well, great for Linux.
    Windows does it better for the desktop, at least for most people's criteria (I don't know how would that criteria be, but they constantly choose it as their main or maybe only desktop OS), so Linux is, by your logics, useless in the desktop.

    Originally posted by kraftman View Post
    By real world I mean benchmarks that really matters like I/O and multi threading performance.
    Startup time is REALLY important for casual users, who just turn the PC on to read some mails and then shut it down.
    Of course, it doesn't matter for hardcore uses (except in the case it's really f*cked up and it takes minutes or hours, but that's not the case in any OS I know of), like gaming, servers, complex math processing and such, since it will be running at least hundreds of times the time it takes to boot (which means it's an overhead of less than 1%).

    Leave a comment:

Working...
X