Page 4 of 5 FirstFirst ... 2345 LastLast
Results 31 to 40 of 41

Thread: Samsung Set To Open-Source Parts Of The Exynos

  1. #31
    Join Date
    Sep 2011
    Location
    Rio de Janeiro
    Posts
    202

    Default

    Quote Originally Posted by droidhacker View Post
    That is not actually correct. Google placed their trademark *within the code*, which is licensed APL. As long as it is used as an identifier for that code, there is not a single thing that Google can do to stop it. If somebody used it in reference to an operating system that is NOT ANDROID, only then would Google be able to create legal problems.


    "Nexus" *does* happen to be a trademark that they can actually restrict use of. However, it doesn't mean a whole lot -- in fact, it has no meaning AT ALL with respect to upgrades. All it actually means is that it is "google approved".

    It is a terrible misconception that there is some advantage to "nexus" devices when it comes to software upgrades. Nexus 1 is out in the cold already (despite being perfectly capable of running AOSP 4.1.2), those associated with the US CDMA networks are late and unreliable when it comes to upgrades.
    I don't think a trademark can be put within the code, since these are different kinds of IP, obviously regional legislation may vary. But google may have licensed the trademark in a single "package" with the code and might not be able to revoke that license now, ok, i get it.

    I also get your point that nexus may not guarantee the updates, but does it at least mean that the code is mainlined, in order to ease the burden of porting new versions to the device? Please note that by mainline I don't mean necessarily Linus' kernel, might be just google's branch. As far as I know android's kernel still hasn't been fully merged back to linux right?

    If the situation is so caotic that each android device ever made has its own branch, than all the more reason the eagerly await an intel valley view phone.
    Last edited by Figueiredo; 10-23-2012 at 02:01 PM.

  2. #32
    Join Date
    Oct 2009
    Posts
    2,109

    Default

    Quote Originally Posted by Figueiredo View Post
    I agree with you. Current tablets aren't my cup of tea. A netbook/tablet thingie running a fullblown OS (much like x86 surface) would be much better. I don't care if the SoC is behind the screen of behind the keyboard.

    But intel's 32nm 2 GHz atom is already on phones:

    http://www.anandtech.com/show/6307/m...powered-razr-i

    So I expect that a 22 nm variant much more competent at supporting android would be a no-brainer. I would love to see an tablet/netbook with an AMD SoC, but they can't properly support GNU/Linux, much less android, so the only alternative for them is windows.
    There are a few PC's in the hybrid form that are nice.
    This one is really slick... AND ruggedized;
    http://ctl.net/tablets/convertible/2...pc-nl3-ec10is2

    That phone is powervr, and that device's performance is NOT up to current ARM chips.
    Check this benchmark (yeah, moron made a video when all he needed was a screenshot... see 2:12):
    http://www.geekanoids.co.uk/reviews/...oids-test.html

    Look at the CPU portion (blue), it got absolutely WASTED by a Galaxy Nexus, which is just a dual cortex A9. The intel's ONLY strength is the memory portion, which will still be matched or beaten by any Krait or cortex A15. Samsung Relay (dual 1.5 GHz Krait w/Adreno225) scores around 5100 on a quadrant benchmark. Razr-i seems to be hitting only 3900.

    Broken down;
    Sammy (from: http://lazytechguys.zippykidcdn.com/...0_quadrant.png)
    CPU: 8470
    Mem: 7435
    IO: 6373
    2D: 1000
    3D: 2206

    Razr-i
    CPU: 4535
    Mem: 7612
    IO: 4115
    2D: 840
    3D: 2458

    Look at that, compared to the intel, the Sammy hit nearly TWICE the CPU, similar memory, 50% better IO, marginally better 2D, marginally worse 3D.

    On the CPU scores; 1.5 GHz dual core ARM running nearly twice the performance of a single core 2.0 GHz x86.... that's pretty ugly for the intel. On a "per cycles" basis, 4535 * 1.5 (i.e. 2.0 GHz * 1.5 = 3.0 GHz = 2 x 1.5 GHz), we get 6802.5. Just how does that ARM hit 8470???

    Or to put it another way, 8470 / 2 = 4235 (1 core), * 4/3 = 5647 (to bump up to 2.0 GHz in a crude and unreasonable fashion). Can that be true? ARM chips getting better performance per core than intel? LOL. 2.0 GHz on a phone sure is funny.

    In any case, IF the intel chip was coupled with an intel GPU, the low CPU performance may be a worthwhile compromise, but since its powervr junk, I'll stick with adreno, which is a sure hell of a lot closer to running open than powervr will ever be.
    Last edited by droidhacker; 10-23-2012 at 02:15 PM.

  3. #33
    Join Date
    Sep 2011
    Location
    Rio de Janeiro
    Posts
    202

    Default

    Must be carefull with benchmarks. I don't know very well what each one tests, but they can vary a whole lot. That's why I only thrust the overall results. Look at these for example:

    http://www.engadget.com/2012/09/18/m...2ghz-medfield/

    http://blog.gsmarena.com/we-benchmar...e-the-results/

    RAZR i sporting a 32 medfield SoC with a PowerVR GPU seems to be fairly competent overall against dual Krait RAZR M. Imagine what intel can do with a 22nm dual core + gen graphics chip...

    To be honest, I don't even care that much about performance. If the interface is fluid and videos play without hiccups, thats enough for me. I am much more interested in being the owner of my device, instead of simply "renting" it from the manufacturer and hoping he is nice enough to support the device. In this sense, and given what you told about the current mess that is the ARM world, I am all the more convinced that intel is the way to go, until some ARM vendor gets its act together.
    Last edited by Figueiredo; 10-23-2012 at 02:39 PM.

  4. #34
    Join Date
    Oct 2009
    Posts
    2,109

    Default

    Quote Originally Posted by Figueiredo View Post
    I don't think a trademark can be put within the code, since these are different kinds of IP, obviously regional legislation may vary. But google may have licensed the trademark in a single "package" with the code and might not be able to revoke that license now, ok, i get it.

    I also get your point that nexus may not guarantee the updates, but does it at least mean that the code is mainlined, in order to ease the burden of porting new versions to the device? Please note that by mainline I don't mean necessarily Linus' kernel, might be just google's branch. As far as I know android's kernel still hasn't been fully merged back to linux right?
    No, it means nothing like that. ALL it means, is that the configuration AS DELIVERED is "google approved".
    By now, I think Android HAS been fully re-merged. I haven't confirmed that however.

    If the situation is so caotic that each android device ever made has its own branch, than all the more reason the eagerly await an intel valley view phone.
    Those valleyview chips aren't targeting phones, they're targeting tablets, and won't be out for at least another year. By that time, we should start seeing ARMv8's, which are supposed to get the whole hardware fragmentation mess under control. One kernel to run them all.

    There is other work going on in the kernel to get the ARM mess under control.

  5. #35
    Join Date
    Oct 2009
    Posts
    2,109

    Default

    Quote Originally Posted by Figueiredo View Post
    Must be carefull with benchmarks. I don't know very well what each one tests, but they can vary a whole lot. That's why I only thrust the overall results. Look at these for example:

    http://www.engadget.com/2012/09/18/m...2ghz-medfield/

    http://blog.gsmarena.com/we-benchmar...e-the-results/

    RAZR i sporting a 32 medfield SoC with a PowerVR GPU seems to be fairly competent overall against dual Krait RAZR M. Imagine what intel can do with a 22nm dual core + gen graphics chip...

    To be honest, I don't even care that much about performance. If the interface is fluid and videos play without hiccups, thats enough for me. I am much more interested in being the owner of my device, instead of simply "renting" it from the manufacturer and hoping he is nice enough to support the device. In this sense, and given what you told about the current mess that is the ARM world, I am all the more convinced that intel is the way to go, until some ARM vendor gets its act together.
    Those results you linked to don't look competent at all. From all the past information everyone has been feeding us, razr intel should be absolutely DESTROYING anything ARM. That its coming in mediocre to low is really saying that these intel chips are NOT up to par.

    BTW: Those broken down to "CPU" are testing several aspects of the CPU, so they are, in fact, "overall". "Overall CPU". "Overall memory". "Overall IO". "Overall graphics". Its also why I included the overall-overall numbers, which put the intel a whole 1200 points behind the samsung/krait.

    Of the links you gave, the only test that the intel came out strongly on was the sun spider javascript test. Can you guess why? Single core processor at a very high frequency along with a single core test that was written with x86 processors in mind.

    Now as for what intel can do with 22 nm dual... imagine what ARM can do with 18 nm. I say 18 because krait chips are 28 nm, the size-reduction game is nothing unique for intel. As theirs get smaller, so will everyone elses. ARM chips are, at the moment, ahead of intel in the size reduction game.

    I've pretty much crossed intel off the list of potential chip suppliers. ARM is not as bad as you make believe it is, and Intel plays the same games. You don't seriously believe that you can run a bone stock redhat or debian kernel on razr-i and have the thing actually work, do you? No, it will take all the same games as any ARM chip takes.

  6. #36
    Join Date
    Sep 2009
    Posts
    114

    Default

    Quote Originally Posted by droidhacker View Post
    It took them the better part of a YEAR to move on to AOSP 4.
    I'll certainly concede that a *final* release took a long time for Ice Cream Sandwich. That said, there were nightly releases that were more stable than most other distributions out there. The reason it took so long to name a final release was because they were still trying to align all supported devices to release at the same time. This was definitely a mistake and a symptom of growing too fast to keep up with the previously used release model. Starting with Jelly Bean the release model changed to allow devices that have good support to be released when they're ready rather than waiting for all the other devices to be ready.


    It is nothing but a MASSIVE pile of broken bloat. Go read the source if you don't believe me. What they have BARELY resembles AOSP.
    I read it nearly everyday. I'm not sure what you see that is such a dramatic change? There is a good deal of extra code that is #ifdef'd for various architecture support that is not found in AOSP. However, the majority of this is pulled out at compile time and doesn't add extra "bloat." The exception to this is some of the changes needed in the Java portions but I've never heard someone honestly say they feel CM9 or CM10 feels "bloated."

    AOSP doesn't require any changes for device compatibility.
    Do you honestly believe this or are you a troll? Do you think you can take the Galaxy Nexus target and compile for the HTC Thunderbolt or a Samsung Galaxy S? Qualcomm and TI release changes needed to support their development boards and its a hell of a lot of changes. One of CyanogenMod's most time consuming tasks is to consolidate all of these disparate changes and make one framework that can support all the major architectures. In addition, there is plenty of work needed to support devices that Qualcomm, TI, etc no longer support. The HAL must either be modified or written from scratch for these devices.
    Granted if you are just building AOSP for the Nexus devices or Pandaboard then none of this matters. However, there's plenty of work that needs to be done to support multiple architectures and legacy architectures that the SoC makers no longer support well.

    They barely touch anything at all at the hardware level. They do NOT do anything valuable. They BREAK AOSP. They add pointless graphics. They pretend that what they release is somehow THEIRS when the only FUNCTIONAL parts are copied verbatim from AOSP.
    Where do you think the device support comes from? The majority(albeit not all) of the device bring up to support AOSP is done by CyanogenMod developers. Do you think Samsung releases all of their code so you can just type one command and compile AOSP for the device? This requires not only what I listed above about adding various architecture support to AOSP but also device specific bring up. I'm not sure what you consider "hardware level" but much of the work CyanogenMod does is dealing with hardware support that is below the Android framework.

    Where and what are all these broken features and pointless graphics? What is broken about the shortcuts in the notification shade or the ability to add a search button to the navigation bar? Where do you see a bunch of pointless graphics? Lets see, there's the boot animation, additional wallpapers, additional icons for new buttons and settings, etc. Are these really a problem? 2.5+ million active users seem to disagree with you.

    My guess is that you already know all of this but want to disregard it due to some vendetta you have against CyanogenMod. To each his own, but I think the majority of Android enthusiasts understand the value of CyanogenMod and other developers.
    Last edited by dalingrin; 10-23-2012 at 02:59 PM.

  7. #37
    Join Date
    Sep 2009
    Posts
    114

    Default

    Quote Originally Posted by Figueiredo View Post
    Must be carefull with benchmarks. I don't know very well what each one tests, but they can vary a whole lot. That's why I only thrust the overall results. Look at these for example:

    http://www.engadget.com/2012/09/18/m...2ghz-medfield/

    http://blog.gsmarena.com/we-benchmar...e-the-results/

    RAZR i sporting a 32 medfield SoC with a PowerVR GPU seems to be fairly competent overall against dual Krait RAZR M. Imagine what intel can do with a 22nm dual core + gen graphics chip...

    To be honest, I don't even care that much about performance. If the interface is fluid and videos play without hiccups, thats enough for me. I am much more interested in being the owner of my device, instead of simply "renting" it from the manufacturer and hoping he is nice enough to support the device. In this sense, and given what you told about the current mess that is the ARM world, I am all the more convinced that intel is the way to go, until some ARM vendor gets its act together.
    I mostly agree with DroidHacker on this. I think the overall advantage in the mobile sector is with ARM and will be for the foreseeable future. It is really an uphill battle for Intel to match the performance of ARM architectures while keeping both the cost and the power demand down. I personally think RISC processors are simply fundamentally better suited for the mobile domain. This is not to say that Intel doesn't stand a chance. By virtue of their R&D resources they could eventually become very competitive with ARM and perhaps even out muscle them.

    There are some major architecture changes coming down the pipe for ARM that will improve performance by huge amounts while still being able to dominate performance/watt.

  8. #38
    Join Date
    Oct 2009
    Posts
    2,109

    Default

    Quote Originally Posted by dalingrin View Post
    I'll certainly concede that a *final* release took a long time for Ice Cream Sandwich. That said, there were nightly releases that were more stable than most other distributions out there. The reason it took so long to name a final release was because they were still trying to align all supported devices to release at the same time. This was definitely a mistake and a symptom of growing too fast to keep up with the previously used release model. Starting with Jelly Bean the release model changed to allow devices that have good support to be released when they're ready rather than waiting for all the other devices to be ready.
    It took them the better part of a year to make the FIRST DEMO release of the FIRST DEVICE. (following the instructions on the google website to download AOSP and type "make" don't count)

    Where do you think the device support comes from? The majority(albeit not all) of the device bring up to support AOSP is done by CyanogenMod developers. Do you think Samsung releases all of their code so you can just type one command and compile AOSP for the device? This requires not only what I listed above about adding various architecture support to AOSP but also device specific bring up. I'm not sure what you consider "hardware level" but much of the work CyanogenMod does is dealing with hardware support that is below the Android framework.
    ZERO device support comes from those sacks of crap. The device support comes out of the various hardware vendors source and binary releases. The only job for whoever is building is to collect the proper components and stick them together. NOTHING TO IT.

    Where and what are all these broken features and pointless graphics? What is broken about the shortcuts in the notification shade or the ability to add a search button to the navigation bar? Where do you see a bunch of pointless graphics? Lets see, there's the boot animation, additional wallpapers, additional icons for new buttons and settings, etc. Are these really a problem? 2.5+ million active users seem to disagree with you.
    We can start with boot animations. Complete waste of time (albeit yes, obviously this is fairly minimal). Part of that, however, is BRANDING.
    I take exception to them trying to pretend that they added something useful and claiming ownership of something that all they added was DAMAGE.
    2.5 million means nothing. Not my fault that there are 2.5 million retards using it. Guess what? Over 69 million retards voted for obama in 2008. IT ISN'T HARD TO FIND IDIOTS.

    My guess is that you already know all of this but want to disregard it due to some vendetta you have against CyanogenMod. To each his own, but I think the majority of Android enthusiasts understand the value of CyanogenMod and other developers.
    To take your figure of 2.5 million and place it into proper perspective... At the BEGINNING of this year, the figure representing the total number of Android devices in use was 250 million.

    Wow, 1% of Android users use crappy broken mod.
    Last edited by droidhacker; 10-23-2012 at 03:20 PM.

  9. #39
    Join Date
    Oct 2009
    Posts
    2,109

    Default

    Quote Originally Posted by dalingrin View Post
    I mostly agree with DroidHacker on this. I think the overall advantage in the mobile sector is with ARM and will be for the foreseeable future. It is really an uphill battle for Intel to match the performance of ARM architectures while keeping both the cost and the power demand down. I personally think RISC processors are simply fundamentally better suited for the mobile domain. This is not to say that Intel doesn't stand a chance. By virtue of their R&D resources they could eventually become very competitive with ARM and perhaps even out muscle them.

    There are some major architecture changes coming down the pipe for ARM that will improve performance by huge amounts while still being able to dominate performance/watt.
    You agree with me??? Wow, and I thought you were just being a dick...

  10. #40
    Join Date
    Sep 2009
    Posts
    114

    Default

    Quote Originally Posted by droidhacker View Post
    It took them the better part of a year to make the FIRST DEMO release of the FIRST DEVICE. (following the instructions on the google website to download AOSP and type "make" don't count)
    1) Ice Cream Sandwich source code hasn't even been out for a year yet.
    2) The first "demo" builds came out a month or two after source release.

    I'm not going to argue that the CM9 release cycle was short but your biases are very clear here. It was decided that most of high level changes from CM7 would be scrapped and cleaner and fewer features would go into CM9+. This took a good chunk of time. On top of this it took a good deal of time for the hardware vendors to release binaries and source changes for Android 4.0. The biggest mistake in the CM9 release cycle, in my opinion, was taking too long to release something for devices that were easy to support like the Nexus devices.

    ZERO device support comes from those sacks of crap. The device support comes out of the various hardware vendors source and binary releases. The only job for whoever is building is to collect the proper components and stick them together. NOTHING TO IT.
    This is either ignorance or arrogance, I'm not sure which. Device bring up for AOSP takes a non-trivial amount of time on top of the time it takes to workaround old proprietary binaries and keeping many architectures and device quirks working together happily in one codebase. If it were so easy then AOSP would support more than 5 or so devices directly from Google.

    We can start with boot animations. Complete waste of time (albeit yes, obviously this is fairly minimal). Part of that, however, is BRANDING.
    I take exception to them trying to pretend that they added something useful and claiming ownership of something that all they added was DAMAGE.
    2.5 million means nothing. Not my fault that there are 2.5 million retards using it. Guess what? Over 69 million retards voted for obama in 2008. IT ISN'T HARD TO FIND IDIOTS.
    This is a silly argument in my opinion. Should Arch Linux not brand their distribution as Arch? By your reasoning they clearly piggy back on upstream projects and simply collect vanilla packages. All they do is make it easier for users to get and run a Linux operating system. How dare they claim "ownership" of their distribution. I guess you think Ubuntu is more deserving of their name because they patch the hell out of upstream? Is this really a good metric to use? What is the minimum number of changes to warrant something to have a name?
    I would somewhat understand your argument if CyanogenMod was a company rebranding something to make money off of it with minimal or no changes. CyanogenMod is simply a community working on an AOSP distribution in their limited free time. How terrible of them to put a name on it. Isn't it easier to give each distribution a name to differentiate their codebase? How utterly confusing would it be if no Linux distribution had a name?

    On one hand you claim the work done by the CyanogenMod community is trivial and does not deserve to have a name or branding. On the other hand, you act like it is some gigantic monster of bloat with all the changes to AOSP. Clearly you have no agenda...

    To take your figure of 2.5 million and place it into proper perspective... At the BEGINNING of this year, the figure representing the total number of Android devices in use was 250 million.

    Wow, 1% of Android users use crappy broken mod.
    Yes, obviously the userbase for aftermarket installations on mobile devices is pretty small, not unlike the number of desktop users of Linux is relatively small compared to Windows and OS X. Does this mean we should discount the Linux desktop? If so, why are you here in the first place then?

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •