REGISTER | LOGIN
Breaking News
News & Analysis

ARM Acquisition: What's In It for SoftBank?

7/19/2016 00:01 AM EDT
20 comments
Page 1 / 3 Next >
More Related Links
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
KarlS01
User Rank
Author
Re: Why?
KarlS01   9/9/2016 11:24:14 AM
NO RATINGS
@Wilco1: "A "Jump To Next Curly Brace" instruction, I love it! That'll teach those stupid C programmers to use correct braces rather than leaving them out all the time. I presume there will also be a "Jump To Next BEGIN" to support Pascal? Please?"

Are BEGIN/END required if only one statement/assignment is present?

I should have said to jump to next statement/keyword since curly braces enclose compound statements.  If no curly braces either do next statement or skip it. 

Yes, the source parser can be modified to use begin/end keywords in place of curlys.  What is important is to extract the control logic structure then corresponding hardware control logic can be generated. 

My ultimate ISA would "do what I want, not what I said'.

As far as your scepticism -- I do have a simple test case running.

Wilco1
User Rank
Author
Re: Why?
Wilco1   9/9/2016 5:21:19 AM
NO RATINGS
A "Jump To Next Curly Brace" instruction, I love it! That'll teach those stupid C programmers to use correct braces rather than leaving them out all the time. I presume there will also be a "Jump To Next BEGIN" to support Pascal? Please?

Directly executing source code to remove all those terribly inefficient and buggy compilers, assemblers and linkers would be a huge step forward for humankind. However I would suggest an even better solution.

My perfect ISA has just one instruction, called "Do What I Want" (patent pending), combined with an infallible AI that senses exactly what you want. Programming will never be the same again, anyone can do it without needing to learn a programming language. All of the world's computing troubles will be solved and it'll create a new era that's only limited by your mind.

uzicohen10
User Rank
Rookie
Was I SoftBank...
uzicohen10   7/25/2016 5:26:03 AM
NO RATINGS
ARM is actually a monopoly in the RISC market. in every mobile phone today you'll probably find more than 10 ARM powered chips. And they've justly earned it, with great product and market resposnse (Intel? are awake?)

So if I were SoftBank, i would complete the acquisition, and raise royaltees, just a little bit, and start enjoying this marvelous cash cow.

 

KarlS01
User Rank
Author
Re: Why?
KarlS01   7/22/2016 4:14:45 PM
NO RATINGS
@p123a: "Better yet, it is possible to implement the IR representation in hardware and take care of IR output from any architecture and gain the performance of native execution"

You can go one step farther and implement Statements, Assighnments, and function calls to skip the compiler bytecode/intermediate language.

Lets call it RRISC for REAL RISC. 

There is absolutely no real reason to have to compile down to every ISA.  C syntax defined loops, if/else, function calls, pointers, and assignments so a compiler could expand the source to assembly.  And of course the curly braces identifiy the jump targets for the relational conditions,

p123a
User Rank
Author
Re: Why?
p123a   7/22/2016 4:25:42 AM
NO RATINGS
Apple is perhaps not running bytecode which is the transport vehicle for Java programs. Apple uses the LLVM infrastructure project compiler Clang (C/C++) and other tools under that umbrella. Clang emits an intermediary code called IR or bitcode whichI think stands for Immediate Representation. IR is not the same as Bytecode but can be used by any with an LLVM backend to translate the IR code to an actual target and is a clever way to obtain architecture portability. Better yet, it is possible to implement the IR representation in hardware and take care of IR output from any architecture and gain the performance of native execution.      

Susan Rambo
User Rank
Author
Re: Junko -- is ANYBODY moderating these messages ????
Susan Rambo   7/20/2016 1:25:35 PM
NO RATINGS
Thank you, Sranje.  We deleted the offending content.

perl_geek
User Rank
Author
Re: A wake-up Call to EETimes and Junko !!
perl_geek   7/20/2016 1:04:37 PM
NO RATINGS
I'll second that. There's an incoherence there that definitely makes the brain hurt.

sranje
User Rank
Author
A wake-up Call to EETimes and Junko !!
sranje   7/20/2016 12:49:11 PM
NO RATINGS
Is anybody moderating comments posted?  Is everybody asleep???

What / why is that posted?  Such misuse by the same "generator" has been going on for many months.   Can ANYBODY at EETimes help???

The embarassment of bad practice
Bruzzer   7/19/2016 11:39:44 AM

realjjj
User Rank
CEO
Re: .....
realjjj   7/20/2016 7:19:43 AM
NO RATINGS
Do note that i did not focus on IoT and i do get your point, more on that later.

On a TAM basis, server plus networking are almost the size of mobile computing - that includes phones, tabs and laptops.

Embedded and that includes IoT could also be about that size in a few years.

There are some other segments too ofc and note that you factor in ASPs when it comes to IoT but forget to do so for cars and servers. For cars the TAM is maybe 10 billion now, not quite sure. Volumes will grow a bit ,maybe to 100 million cars per year but the content will grow by quite a bit more. Wouldn't be unreasonable to expect for the TAM to double by 2025 or even a bit sooner.

Glasses will beat (and replace) phones. Higher volumes, higher content, the timing is difficult to guess. Robots for businesses and consumer will be high ASP, a lot of silicon content and volumes will be huge but timing is even harder.

 

As for IoT, yes the volumes are limited for now and ASPs must be low. for volumes to increase  I also don't like the idea of having 1000 pointless IoT things, it's not efficient. Would be better if our PCs (glasses) and robots do most of what those 1000 devices are supposed to do.  Maybe the biggest problem with IoT is that we need a million great devices. With glasses you only need one single great device to start the avalanche. Same for robots.  It's not easy to make one great product, creating many is so much harder.  It is possible though that ARM might aim at more than just hardware when it comes to IoT. Anyway, i am not counting on any IoT boom when saying that ARM can still grow a lot inthe next decade.


Everybody is focusing on phones and that's somewhat relevant for the short term as it is the only market where ARM has very high share. However, ARM has been gaining share in many other markets and if they stay on that trajectory there is good groth ahead. Add to that new markets and more iP and , if you actually do the math, it adds up.

dromdrom
User Rank
Author
Re: .....
dromdrom   7/20/2016 4:24:34 AM
NO RATINGS
"but it's wrong to focus just on phones."

@realjjj: Well I don't know about you but unlike mobile phones I nor many people I know buy a new car and server every year, I also do not know about any other product with a $5+ CPU that sells in the mobile space volume that could come close to the mobile volumes (and generate comparable CPU royalties).

IoT will only reach the many billion volumes if the price is very low. That implies the royalty income for ARM or any other CPU supplier will not be as impressive as Son is hoping for.

Page 1 / 2   >   >>
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed