Slashdot Mirror
Intel Rolls Out "Beacon Mountain" Android Dev Platform For Atom
MojoKid writes "In an effort to coax developers to begin taking Atom seriously as an Android platform, Intel has just released a complete suite of tools that should help ease them into things — especially since it can be used for ARM development as well. It's called Beacon Mountain, named after the highest peak outside of Beacon, New York. As you'd expect, Beacon Mountain supports Jelly Bean (4.2) development, and with this suite, you're provided with a collection of important Intel tools: Hardware Accelerated Execution Manager, Integrated Performance Primitives, Graphics and System Performance Analyzers, Threaded Building Blocks and Software Manager. In addition, Android SDK and NDK, Eclipse and Cygwin third-party tools are included to complete the package."
Microsoft products are vastly better than that garbage.
Oh, whoops I thought it said 'Bacon' mountain. Curse you Intel and your false meaty product names.
Silence is a state of mime.
I hope I wasn't the only one to read that as "Bacon Mountain".
I love bacon!
Charlie! Lets go to Candy Mountain, Charlie!
If you gave me a choice between a printer and a giraffe with explosive diarrhoea, i'll get my ladder and my raincoat
I sure would like to begin developing applications for a platform I don't have! Maybe if I create a program, a major company will create some hardware for it! (at a price I can afford, available on a network in my area)
Microsoft products are...
Android and Chrome head Sundar Pichai has just revealed that Android has passed the milestone of 900 million activations, up from 400 million in 2012 and 100 million in 2011 (to put that in some perspective Windows Installs is about 1.2Billion). Its an incredibly popular OS that people want, on devices people want. The same is not true for the current version on Windows with its new tablet interface, on current PC's, Which is damaging the whole PC industry....and in context of this article why intel wants to be part of this growing wave of devices.
Yea it supports ARM, how is that support and how will it work out for you? Will the support it equally? What is the licensing of this confabulation? Do I have to pay anything if I make a commercial product other than the atom processor, support chips and sundry support components?
I've priced atom with all the needed support chips and compared to arm and it sucks balls on costs. I'm leery of hidden costs in this confabulation over the already sub par costs of atom.
I'd go on a Vegan diet but the delivery time from Vega is too long. --brownkitty
Wrong. Bacon Mountain conveys Awesomeness in Large Quantities.
Just curious, but the initials of Beacon Mountain are B.M. Do you think that's an indicator that Intel pretty committed to Windows still?
Don't you belong in the 1990s? You know, back before x86 became the most powerful and power efficient architecture.
Increasingly, the best Android apps will use C++ and assembler, producing binaries that will NOT run only dodgy x86 versions of Android. There is already an issue of the best 'to-the-metal' apps on Android only running on certain ARM tablets, although this is usually down to laziness or excessive caution by the programmers. ARM provides excellent ways to ensure ARM binaries have sufficient support for the minor variations found amongst the most commonly used ARM CPU cores, the main variation being in the area of vector acceleration facilities for floating point code.
The world doesn't need x86 Android. The world doesn't want x86 Android. The world is only subject to x86 Android because Intel (illegally) PAYS third parties to build x86 Android devices. There is no sane commercial reason for any company to use an Intel chip UNLESS Intel turns up with wheel-barrows full of cash and shed loads of free low end x86 parts. Luckily, getting the devices built doesn't help Intel subvert the marketplace, since no-one chooses to buy them. Buying an Intel Android tablet would be like buying a non-cortex ARM based tablet. Sure, they'll both run 'Angry Birds', and other primitive Java only apps. However, no aware person would choose a non-cortex ARM or x86 CPU unless they wanted to be constantly checking the compatibility of Android software (and at least Android ARM binaries CAN be made compatible with non-cortex ARM v7).
We've seen this before, in the early days of Microsoft NT (now, what you call 'Windows'). Microsoft backed 3 or 4 different CPUs, and provided tools for each. In theory, an app could carry binary pay-loads for each type of CPU in the same package. In practice this NEVER happened. Either an app was a general program for a common x86 based PC, or an app was a highly specialised program for a MIPS machine or whatever. Of course, back then the (supposedly) CPU ISA independent .NET initiative did not exist.
Or again, consider the nintendo Wii U. This console was designed for brainless and cheap ports from the Xbox360. The Wii U has CPU and GPU features that can be considered as supersets of the Xbox360, but in reality things are more complex. The Wii U may have more power than the Xbox360, and 'compatible' hardware (same CPU ISA, GPU form same company), but now almost no Xbox360 developer is creating versions of their games for the Wii U. Intel's argument for Android on x86 is like Nintendo's argument for the Wii U- namely that developers from successful platforms will obviously want to port their apps/games across if the process is 'easy' enough.
In the world of software development 'easy enough' is a buzz phrase designed to fool the 'pointy-haired bosses', and it doesn't even do this. The very reason, for instance, that EA no longer codes ANY games for the Wii U is the self-same reason vanishing few good apps will appear for the x86 version of Android. Testing, supporting, and porting just won't be worth the effort. Developers who support Intel KNOW they are uselessly helping to fragment the Android market, AND support a CPU manufacturer that, if successful, will massively raise the cost of x86 Android CPU parts. Intel's mad dream is to drive ARM out of the mobile market, and then to raise the price of their mobile x86 parts back to notebook levels.
I always enjoy a hearty B.M. every morning and evening, it's a natural healthy process, nothing to be ashamed of.
Yeah right, and we're supposed to believe some guy named Sumdum Pikachu about this?
They put it on a big screen and everything :) http://cdn.androidcentral.com/sites/androidcentral.com/files/imagecache/w680h550/postimages/108579/900m.jpg
http://software.intel.com/en-us/articles/beacon-mountain-v05-frequently-asked-questions
The IDE is provided free of charge, and you view the licenses when you install the IDE. I have no idea what they are because I haven't downloaded it, but you can check for yourself if you are curious.
A little tip: All bibles of all religions were written by primitive men with no grasp of science, not some magical sky daddy.
Apples and oranges. I wish idiots like you would stop trying to compare a telephone "OS" to an actual, real PC OS.
Then Windows 8 should have been a real operating system instead of a tablet one (on machines with lower DPI and less portable), because everyone right now is choosing Android over Windows. Its very much an Oranges vs Oranges comparison (Apple priced themselves out of every market), and its what Microsoft wanted...pushed even with its self styled Ecosystem at least they will make a Billions from the shop :).
The x86 ISA is still inefficient because of the lack of registers. That's why they added more with x86-64. The x86 won primarily because of amazing manufacturing, not because it's an incredible instruction set.
"First they came for the slanderers and i said nothing."
The x86 ISA is still inefficient because of the lack of registers. That's why they added more with x86-64.
Which happened how long ago? Your argument basically boils down to "well they improved it, so that doesn't count". How many other CPU architectures remain stagnant with no updates?
Sorry to break this to you, but the lowest end Intel CPU is powerful enough to emulate the highest end ARM CPU
Its not true ARM chipsets are faster than atom chipsets...and even if they weren't you need a 12x speed in power. The bottom line though is what people need is CPU's cheap, fast enough (for smartphone apps) vs power consumption (at least a day maybe two). The problem till now is Intel didn't have a CPU suitable for mobile...now they do (have for a while), but they are still expensive(because they insist on ludicrous margins...and its helping kill the PC industry), and in comparisons worse than the opposition.
The Atom chipset is generally 32 bit, especially in mobile devices, so it's still using the old architecture, with not enough registers.
"First they came for the slanderers and i said nothing."
A cosmetic change to the start menu does not change that any more than the superbar in Windows 7 did.
I notice this new lie, about the start menu. Its just that a lie. The problem with Windows 8 is that it resembles an embedded OS on locked hardware not a Disk based OS on General Purpose Hardware...and compared poorly to Android and iOS. The problem is metro...the problem is Windows RT.
Uhh, you're years late. The Atom has had 64-bit support since 2008. That's 5 years ago.
Youre welcome to make your own architecture. We await eagerly the amazing innovations you will surely be bringing to the table.
Sorry, I work only on Linux and OS X. Try again.
I only eat MorningStar Farms(R) brand Veggie Bacon Strips. They're delicious and good for you.
I'd like some...
Here is the the tegra 4 (4+1 core) clocked at 1.9Ghz and 2.3GHz respectively...but again that is not really my point the threat is the Allwinner...or the next generation Allwinner.
Uhh, you're years late. The Atom has had 64-bit support since 2008.
Most don't.
"First they came for the slanderers and i said nothing."
Consult Google and educate yourself on the joys of register renaming.
Actually, most of the Atoms in production right now have 64-bit support. How many ARM CPUs out in real use do?
Register renaming doesn't make up for a lack of registers because it's harder for a compiler (or human) to optimize for it.
"First they came for the slanderers and i said nothing."
ARM CPUs don't need it because they already have enough registers.
"First they came for the slanderers and i said nothing."
So what you are saying is that a modern day ARM CPU is better than a decades old x86. Try comparing the two at the same time frame.
Gee, I can't wait for the day that my crappy x86 CPU will be able to run Crysis 3 like my vastly superior ARM CPU can. You know, what with the x86 being horribly inefficient and not being able to handle programs of the same complexity as ARM.
If you want to compare the two at the same time, then we can say that x86 was built for a time, decades ago, when demands on the CPU were different. Intel has been able to keep up with modern technology by clever engineering and piling on bandaids over the chip's eccentricity's, but it's still a pile of baindaids. That's why Intel wanted to dump it and start fresh.
x86 won the desktop wars because of Intel's incredible manufacturing processes, not because it's a beautiful design. The design is a dog.
"First they came for the slanderers and i said nothing."
On my supposedly "archaic" x86 desktop, I download any Linux distro I feel like using and can use the exact same installer to setup a 5 year old desktop or next month's Haswell.
On my "futuristic" smartphone I have to wade through outdated information on sketchy forums to find the exact set of model-specific voodoo in order to unlock the device. Oh.. and I'm aware that not every ARM device comes locked, I was in the first-wave of Raspberry Pi purchasers. But guess what? Even with my Raspberry Pi I have to hunt down images that are tailor made just to booth with the Pi and stepping off the Raspberry Pi software reservation gets real ugly real fast.
Why is the thought of an unlocked x86 tablet that could host the exact same Linux distro that I feel comfortable with on various other computers be considered some type of evil? Why is the idea of having the ability to install a stock Android with no garbage without having to sift throught 2,000 forum posts dedicated to a specific flavor of smartphone for a specific vendor considered "anti-freedom"?
AntiFA: An abbreviation for Anti First Amendment.
The design is a dog
That may be so in your opinion, but it's still the best architecture right now. That is why everyone has switched to it, regardless of whether it was to Intel or AMD.
That may be so in your opinion, but it's still the best architecture right now.
As if you have any understanding of CPU architectures. I know your type, you're the fanboy who will say anything to defend your object of adulation. More interested in fighting than learning.
"First they came for the slanderers and i said nothing."
I wrote x86 assembly language for nearly 10 years in the early 80s to early 90s. You're the one who is talking out of his ass.
And I have to say, that was a blatant cop-out on your part. You are presented with facts that you cannot refute so you resort to personal attack. Fantastic.
An SDA on /.?
But now I am disappointed in the lack of salted porky goodness.
While I used to brainstorm at the Security & Defence Agenda in Brussels, I didn't find it challenging enough so I left.
If the glove fits, wear it. If God came down and gave you the perfect CPU, then you'd still say, "x86 is better."
"First they came for the slanderers and i said nothing."
Lamest logical fallacy ever. The poster brings a valid point, and you respond with nonsensical, empty rhetorics. "Can you do better? No? Then shut up." instead of countering his argument with something factual.
I would? You have to show me this ARM powered brain scanner you've devised some time, because I have a few suggestions for improvement.
The poster brings ignorant, fanboy bullshit
FTFY
I dont think its a fallacy to point out that NOONE has managed to make faster processors than x86-64 arch ones in years despite claims that there are superior processor arches. Appeal to authority is not necessarily a fallacy if all of the biggest authorities in the field agree.
If that was the only concern sane people would use MIPS not arm as it has twice as many registers again.
(Or itanium that has a stupid number).
Clearly it is not the only concern.
"First they came for the slanderers and i said nothing."
> That's why they added more with x86-64.
That's why _AMD_ added more with x86-64.
Intel was still trying to get Itanium working when AMD trumped them by bring out x86-64. Intel had to copy that to keep relevant. The initial Intel x86-64 CPUs couldn't even run Windows-64, an AMD chip was required until Intel fixed theirs.
> back before x86 became the most powerful and power efficient architecture.
x86 (or x86-64) never was, never is, and never will be.
It may well be more cost effective than the more powerful and power efficient CPUs, but that is because of volume production and not because of its architecture.
Considering the bastards don't even remotely plan to support Linux, I'm uninterested.
Moreover, X86 Atom's not a player in the space because of it being X86- most of the apps with NDK binaries (read: about 1/3 of the apps in the store) are ARM ones not X86. It's going to support ARM...big whoop-de-doo... I already HAVE that support for both X86 and ARM already- for free and on ALL OS platforms.
This is just to look "relevant" for the purposes of impressing sharesellers.
Intel wanted to dump it so they could be the sole provider of chips for it. Having AMD able to compete at all was considered worse than AMD just limping along.
First, the "fallacy" referred to was along the lines of "let's see you do better," which is certainly not the same as "nobody has done better."
Second, best != fastest clock. MIPS and ARM can be done in a fraction of the silicon, have simpler ISAs, and use much less power at comparable technologies. And no, "Intel can make it at 22 nm when everyone else is at 32 nm or higher" (may be slightly out of date by now) is not a valid reason for saying "x86-64 is the best arch there is."
Third, this is false unless you count overclocked prcessors. You say NOONE, I say IBM had POWER6 up to 5.00 GHz in 2008.
According to this page,
Also because the x86 instructions are variable length, which means they are more L1 cache efficient. The cost of instruction decoding (transistors / watts) is non-zero, but negligible (even for ultra low power designs).
x86 won over pretenders due to software compatibility.
ARM won over MIPS due to a focus on embedded and more aggressive / better licensing... Which brought them presence and mind share even though MIPS was commonly taught in college.
Intel are very capable. You may not appreciate their instruction set, but you'd be a fool to think ARM are playing at their level.
And don't hold the ARM instruction set up as particularly special. Check out their new 64-bit model... It's a significant change... Some elements you may have loved from ARM32 are gone.
If so, why is ARM doubling the number of registers in ARMv8?
x86 (or x86-64) never was, never is, and never will be.
Except it is. Right now. Name ONE commercially produced microprocessor that can outperform a modern x86 CPU.
Clockspeed doesn't mean shit unless you are comparing identical processors.
How much do you want to bet that my Core i7 3770T 2.5GHz will leave any single POWER based CPU in the dust, even at 5GHz?
Good point, you can never have enough registers.
"First they came for the slanderers and i said nothing."
Ha - actually more registers equates to more state you need to save when you tape an interrupt / exception / context switch. So it's a trade off for sure.
Can't you tell?
I've got about 25 years of assembly experience, and he's right - x86 is dog shit now, and was dog shit back in the 80's as well. Intel wouldn't know what an orthogonal instruction set was if it bit them in the ass.
I've got about 30 years of assembly experience and he's wrong. x86 is great now and it was great back in the 80s as well. Intel invented the microprocessor, so they know how best to approach it.
Despite the fact that Intel has only about 3 phone models released, all on the same platform (a second one is coming), they fail to provide updates to Android 4.2 for all of them.
So this is just another shot in the foot for them. Android is a great ecosystem, but it is not for the faint of heart. If you want to compete, you have to do it properly. Half-hearted attempts will like (just like HTC).
Intel has been able to keep up with modern technology by clever engineering and piling on bandaids over the chip's eccentricity's, but it's still a pile of baindaids. That's why Intel wanted to dump it and start fresh.
By bandaids what do you mean exactly? We all know that support for 16-bit mode and memory segmentation costs silicon, but at most that requires the same number of transistors as the first 8088. If you mean because its not RISC then sorry your favorite design philosophy didn't win the performance war, but it was for the same reason that the opposing design philosophy didn't: Hybrid of CISC and RISC is better than either, but can only be implemented with a design philosophy that permits a large instruction set.
As far as register counts, there are diminishing returns to adding more registers. In practice any out-of-order CPU (ie: performance) already has a pool of registers greater than the number of ones indicated by the instruction set. This pool is used in a register renaming scheme that allows the CPU to manage a long out-of-order pipeline efficiently.
You claim that Intel wanted to dump x86 and start fresh with something simpler, but thats not what we observe to have been implemented with Itanium. Sorry, Intel wanted to dump x86 because they had to share x86, not for whatever bullshit you are imagining.
"His name was James Damore."
The design is a dog
That may be so in your opinion, but it's still the best architecture right now. That is why everyone has switched to it, regardless of whether it was to Intel or AMD.
Anyone, and I mean anyone, who believes that the Intel x86 with its stupid segmentation (when it started) and other ugliness was in any way "the best" architecture compared to say, the Motorola 68K series, with a fairly 'pure and graceful' 32bit architecture needs their head examined. x86 came off the 8086 (a "16bit" architecture), then the 80286, and then the 80386 with all that 16bit "crap" still stuck in it. 68K started off clean as a 32-bit architecture.
Maybe nowadays, with things going 64-bit and the need for greater than 32-bit addressing, even 68K is somewhat ancient, but it is/was a *hell* of a log better 32-bit architecture than x86 *ever* was.
Where Motorola lost out was in IBM picking x86 for the PC, Apple stagnating for a while and then dropping 68K or PPC (still Motorola), and with the PC's success (Intel having far more money rolling in from lots of PC sales, vs. Motorola not so much from the far smaller percentage of Mac sales by Apple) which company do you think had more income to invest in better Fabrication technology, etc?
Register renaming doesn't make up for a lack of registers because it's harder for a compiler (or human) to optimize for it.
Register renaming and the OOE that it enables greatly widens the target that compilers needs to hit. The point of register renaming is that you still win even when the compiler is retarded about register use. Intels latest chips are commonly pulling 2.5+ instructions per clock cycle even when using compilers written in the 1990's, written before there was even such a thing as register renaming.
Intel designs processors that execute existing code efficiently. Thats the metric in use by their engineering team.
In no way, shape, or form does register renaming have either a negative impact on cpu performance or on compiler optimization opportunities. Its the exact opposite.
"His name was James Damore."
Intels latest chips are commonly pulling 2.5+ instructions per clock cycle even when using compilers written in the 1990's, written before there was even such a thing as register renaming.
ARM cores are dual-issue now, so that's just not that exciting any more.
In no way, shape, or form does register renaming have either a negative impact on cpu performance or on compiler optimization opportunities. Its the exact opposite.
Straw man, or at best, you failed to understand the argument. The argument is not that register renaming has a negative impact on cpu performance, the argument is that having more GPRs provides superior performance to register renaming. This is a proven fact; just recompiling some code for x86_64 provides a 15% performance increase for this reason alone (on the same processor.)
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Uhh, you're years late. The Atom has had 64-bit support since 2008.
Most don't.
Intel will "solve" this problem by simply abandoning the old processors. They did it to the first atoms already; preview releases of new Linux distributions that Intel has contributed code to (e.g. Tizen) don't support them already.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
That may be so in your opinion, but it's still the best architecture right now
Learn to read.
Best Typo Ever. (Check the overview slide.)
As far as register counts, there are diminishing returns to adding more registers. In practice any out-of-order CPU (ie: performance) already has a pool of registers greater than the number of ones indicated by the instruction set. This pool is used in a register renaming scheme that allows the CPU to manage a long out-of-order pipeline efficiently.
Do you really think this makes up for not being able to choose registers at compile time?
"First they came for the slanderers and i said nothing."
Do you really think this makes up for not being able to choose registers at compile time?
Uh, who said anything about not choosing registers?
You have imagined an extreme that does not exist.
"His name was James Damore."
This is a proven fact; just recompiling some code for x86_64 provides a 15% performance increase for this reason alone (on the same processor.)
Note to self:
A CPU that has completely unused silicon in one mode performs better when all of its silicon is being used by the other mode.
You claim that I failed to understand the argument, but its you that are failing to make an argument that doesnt have a hole big enough to drive a clue truck through.
"His name was James Damore."
In an effort to coax developers to begin taking Atom seriously as an Android platform
I think that part was a joke, because TFA states it's windows only.
The [very few] windows devs I know, use MS tools like C#, etc. Are there actually any android devs that run windows out there?
A CPU that has completely unused silicon in one mode performs better when all of its silicon is being used by the other mode.
We're talking about GPRs here, in 32 bit mode it uses less of them. There are other differences, but it's still using all the functional units in 32 bit mode, which makes your argument laughable at best.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
People prefer to write programs that use between four and sixteen registers. Too many GPRs wastes stack space. Too few requires constant copy instructions that are done quickly by a CPU that secretly has more registers. x86 famously only has four. PPC famously has 32. AMD looked at the history of register use and chose to have 16.