Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.os.linux.embedded > #430

Re: Linux: Pressure on Intel to ditch x86 and move to ARM chips

Cross-posted to 2 groups.

Followups directed to: comp.os.linux.advocacy, comp.os.linux.embedded

Show all headers | View raw

Snit wrote:

>> Linux: Pressure on Intel to ditch x86 and move to ARM chips
>> -----------------------------------------------------------
>> 
>> http://seekingalpha.com/article/1090941-intel-making-apple-s-arm-chips-
will-leave-samsung-behind
>> 
>> The world + dog now eats from only ARM.
>> So the pressure is on for Intel to ditch its out dated x86 and
>> focus on building ARM.
> 
> Even if Intel does start making ARM chips again (it used to make the
> StrongArm chips), they would not be ditching the x86 architecture.
> 
>> 
>> However there are a number of problems with ARM chips.
>> 1. No software headers to define registers and bit fields.
>>    So no portability of software from one version of the
>>    CPU to the next. Absolute ton of work has to be done
>>    to port Linux from one CPU to next even from same manufacturer.
> 
> That would explain why so few people use Linux on ARMs, or use ARM
> devices in embedded systems.

Liar!

Nearly ALL arms that are SoCs run Linux.


> Apart from the minor issue that the assumption is bollocks - there /are/
> headers available defining all the necessary parts.


You will now prove this by posting one header file where
the registers and bit fields are define!!!!!!!!!!!!!!!!

BWAHAHAHAHAHAHAHAHAHAHAHHAHAHAHAHAHAHAHAHAHAAAA!!!!!!!!!!!!!!!


> One issue that has existed for Linux on ARMs is that there have been
> many variants of both chips and boards - so even though the details are
> documented and the headers exist, the differences between them have made
> it inconvenient to support new boards and devices.


Fake issue.

The issue is that ARM Holdings forgets to give out a library which
allows registers and bit fields to be captured from the IP they sell.

They realize their mistake and have issued the CMSIS library
but it only goes half way and names only some registers for some CPUs,
and no bit fields. Impossible as ever to port software from one
CPU to next without hard coded numbers which change with each CPU.
Unlike a PIC where one single header file defines everything; so when a CPU
changes to a new variant, only one device specific file changes.


> But recently the ARM
> support in the kernel has been consolidated to give more consistent
> support - and ARM themselves have been a big driving force here.

I doubt their sincerity. The CMSIS files were written by amateurs.
They only so far cover about 5 or 10 CPUs out of the billions they
have sold.

 
>> 2. Everything is under NDA - so no chance of developers
>>    getting hold of datasheets and getting open source
>>    support to port drivers such as SATA because there
>>    are no software headers in the first place to start
>>    this work, and if you discussed the registers and their
>>    bit fields from NDA'd data sheets to get help
>>    then you are in violation of the NDA.
> 
> That would explain why so few people use Linux on ARMs, or use ARM
> devices in embedded systems.



Liar!

Nearly ALL arms that are SoCs run Linux.


> Apart from the minor issue that the assumption is bollocks - most
> information is easily available without any sort of agreements.


Liar!

You can't even get the full BGA diagram for most CN, Taiwan, Korea
produced chips, or the full register set, or even the header files
they use for their own chips.

 
> There /are/ sometimes NDAs or other licensing requirements for some
> parts of the information.  Some chips are made specifically for a single
> customer, and only that customer gets the details.  Other chips have
> things like cryptography or security features that are not publicly
> documented.  And some companies like to keep details of graphics
> controllers hidden and provide only binary blobs.
> 
> 
>> 
>> So you see a lot of projects with limited functions
>> such as USB + wifi but no ethernet, or ethernet+SATA but no wifi.
> 
> That would be because the project only needs limited functionality.
> Embedded systems developers don't waste time and money on features they
> don't need.

Liar!

Thats not what the management and sales people said.

 
>> Unusual combinations despite being same chip, because
>> the engineers in that particular factory could not
>> solve their software problems on their own
>> or reach out and download open sourced working
>> code because NDAs prevent source code being released despite such
>> NDAs being illegal and GPL code hoarding is also illegal.
>> 
> 
> That would be because the project only needs limited functionality.
> Embedded systems developers don't waste time and money on features they
> don't need.


Liar!

Thats not what the management and sales people said.


>> Despite all that wrong doing, look around you and 90% of the
>> world is powered by Linux
>> gadgets for all gadgets worth $30 or more.
>> 
> 
> If anything you claim about NDAs and secrets were remotely true, this
> would not be the case.
> 
>> There is a way out and that is for companies like NXP,
> 
> NXP make ARM chips...
> 
>> Microchip or Intel to make their own chip, or licensed ARM chip,
>> and do it in such a way they also create proper
>> header files for registers and bit fields, and then make
>> it into a consistent family that have no NDAs and open source
>> friendly documentation so that anyone can go build
>> projects with chips that are not encumbered with NDAs.
>> 
> 
> There are plenty of companies that make ARM devices with lots of
> detailed information and header files, as well as the massive amounts of
> documentation and header files available from ARM themselves.


Liar!

Thats not what the management and sales people said.

 
>> The first person to do it properly ditching all NDAs
>> and making open source friendly chip will be the winner
>> because they would gobble up the entire market for
>> Internet of Things worth at least 20 billion devices
>> in as short a period as 5 years.
>> 
>>

Back to comp.os.linux.embedded | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Linux: Pressure on Intel to ditch x86 and move to ARM chips 7 <email_at_www_at_enemygadgets_dot_com@enemygadgets.com> - 2013-01-14 00:18 +0000
  Re: Linux: Pressure on Intel to ditch x86 and move to ARM chips David Brown <david@westcontrol.removethisbit.com> - 2013-01-14 11:06 +0100
    Re: Linux: Pressure on Intel to ditch x86 and move to ARM chips 7 <email_at_www_at_enemygadgets_dot_com@enemygadgets.com> - 2013-01-14 12:15 +0000
      Re: Linux: Pressure on Intel to ditch x86 and move to ARM chips David Brown <david@westcontrol.removethisbit.com> - 2013-01-14 13:21 +0100

csiph-web