SoC: Back and forth!

[Natsu]

^If soft-float libraries are included then it's possible to have a compatibility layer.


I stated my reasons before, but If anX86 SoC is chosen, I don't plan on getting a Pyra.

Can I get a link to those reasons?

 

[Hồng Thất Công]

I'm grabbing a chair and popcorn....This gonna be interesting (all over again)...

 

[TrashyMG]

Can I get a link to those reasons?

Well if the forum's search wasn't garbage.. But I'll recap


1.) I find ARM interesting, it's a new frontier and find it fun to work with. Not too interested in things that have been already done.


2.) well I kind of summarized it in the first bullet.


3.) Also I bought an F'ing Omap5 Dev board.. before everyone got wishy washy on the whole SoC choice.

 

[Null]

 

[Natsu]

Can I get a link to those reasons?

Well if the forum's search wasn't garbage.. But I'll recap


1.) I find ARM interesting, it's a new frontier and find it fun to work with. Not too interested in things that have been already done.


2.) well I kind of summarized it in the first bullet.


3.) Also I bought an F'ing Omap5 Dev board.. before everyone got wishy washy on the whole SoC choice.

*devil's advocate

1.) Intel new chips aren't a new frontier?

2.) Well, that helps.

3.) Jump the gun. :lol:

 

[Wrath Of Khan]

(1) Sell omap 5 dev-board.

(2) wait and see what soc is chosen for pyra.

(3) Buy appropriate dev-board when soc is finalised.

(4) Winning!!!

 

[TrashyMG]

Let me fix this:

(1) Sell omap 5 dev-board.

(2) wait and see what soc is chosen for pyra.

(1) wait and see what soc is chosen for pyra.

(2) Sell omap 5 dev-board. if omap5 isn't chosen

(3) Buy appropriate dev-board when soc is finalised.

(3) Weep silently if an ARM SoC isn't chosen..

(4) Winning!!!

 

[Hồng Thất Công]

Nah, I wouldn't worry about it.  When the smokes clear, aside from some legs laying around , it will be ARM

 

[wej]

1.) I find ARM interesting, it's a new frontier and find it fun to work with. Not too interested in things that have been already done.

New frontier? ARM has been used in mobile devices for ages, much unlike modern x86, which is pretty new in this context.

3.) Also I bought an F'ing Omap5 Dev board.. before everyone got wishy washy on the whole SoC choice.

If we had known that, nobody would have dared to even think about using another SoC, let alone suggesting to look at other options.

 

[TrashyMG]

New frontier? ARM has been used in mobile devices for ages, much unlike modern x86, which is pretty new in this context.

mobile device or not doesn't matter to me as much as the platform, x86 isn't new to native linux/software development in general.. When it comes to Linux, ARM is still fairly new and hasn't completely matured yet.. 


This will spur someone to write a ton of crap how everything will be easier and such with x86.. I don't go down easy roads.

 

[wej]

When it comes to Linux, ARM is still fairly new and hasn't completely matured yet..

Linux runs on ARM for more than 15 years (e.g. on the good old StrongARM SA-1110 from 1998), which I don't consider new.

 

[vcoleiro1]

What exactly where the reasons we are going with ARM?.  

I seem to remember most people didn't want x86 because they thought it would eat too much power, then that was disproved with the Baytrail SOC's , people started to come around, . Then there was a discussion on recompiling old programs, NEON, compatibility etc which Exophase addressed . It then looked like x86 was a real shot - then Notaz said he wasn't to keen on x86 . Finally there was a post saying we chose ARM.  

It was a bit weird in the end, in regards to having a clear picture on the decision other than that abstract overview. In the end, It came across to me as, yeahh ok, x86 SOC's seem to be ok now, but we have already chosen and have our mind set on ARM.

Dont get me wrong, not saying one way or the other.  Just would be nice to know the explanation of the decision.

 

[ekianjo]

What exactly where the reasons we are going with ARM?.  

I seem to remember most people didn't want x86 because they thought it would eat too much power, then that was disproved with the Baytrail SOC's , people started to come around, . Then there was a discussion on recompiling old programs and compatibility which Exophase addressed . It then looked like x86 was a real shot - then Notaz said he wasn't to keen on x86 . Finally there was a post saying we chose ARM.  

It was a bit weird in the end, in regards to having a clear picture on the decision other than that abstract overview.

let's not start this discussion here please, there's a whole bunch of threads about that, and it's not about "WE chose", it's "The Pyra Team" - this is not a vote or democratic kind of decision, someone has to make a call sooner or later. 

 

[TrashyMG]

Linux runs on ARM for more than 15 years (e.g. Linux runs on the good old StrongARM SA-1110 from 1998), which I don't consider new.

and your point? A small select devices used that and only the companies that made the device and a select few uber geeks could compile anything back then on it.. Right now ARM is starting to become more of a commonality.. over recent years ARM and other platforms have been getting more notice in the opensource community and to seriously look at portability instead of blindly building for x86 only. That is what is new..

 

[wej]

What exactly where the reasons we are going with ARM?.  

I seem to remember most people didn't want x86 becuase they thought it would eat too much power, then that was disproved with the Baytrail SOC's , people started to come around, it looked like x86 was a real shot - then there was a post saying we chose ARM.  

Back in the Pandora days, x86 wasn't really an option, so back then the choice for ARM was pretty easy. Things have changed since then, those Baytrail SoC are really competitive in terms of both power consumption and cpu power. Also it runs with open source drivery only. No blobs needed.

Probably the only valid argument for ARM that I can think of, is the existing emulator software for the Pandora which has been optimized for ARM. It is not like it is impossible to get good performance on emulators on x86, but the work that has gone into those arm optimizations would certainly not be usable on x86.

Another argument is binary compatibility with the Pandora. Although it might be possible to some degree, I don't think that would be a good reason for sticking with ARM, as you would have to keep all those old Pandora libraries to maintain compatiblity. For performance reasons, you would want to recompile most software anyway.

 

[wej]

Linux runs on ARM for more than 15 years (e.g. Linux runs on the good old StrongARM SA-1110 from 1998), which I don't consider new.

and your point? A small select devices used that and only the companies that made the device and a select few uber geeks could compile anything back then on it.. Right now ARM is starting to become more of a commonality.. over recent years ARM and other platforms have been getting more notice in the opensource community and to seriously look at portability instead of blindly building for x86 only. That is what is new..

Even then lots of embedded devices have been running Linux. You may not have noticed it, though. Toolchains were not that much different from today's toolchains either.

When writing software for Linux, the actual CPU architecture usually does not matter much anyway, if it matters at all. There are a few cases, where the architecture does matter (e.g. emulators which you want to optimize for the target architecture), but unless you are into writing ARM assembly, the CPU architecture really does not matter much. When talking about portability, writing software to run on different operating systems is usually much more effort, than recompiling software which runs on Linux x86 to run on Linux ARM.

 

[vcoleiro1]

@ekianjo

How's it going with the Snapdragon 800 discussion with Qualcomm, is ED talking to the appropriate contact now?

 

[Natsu]

^or the 805.

 

[TrashyMG]

Even then lots of embedded devices have been running Linux. You may not have noticed it, though. Toolchains were not that much different from today's toolchains either.

Okay, I was still learning the ropes of Linux back in 1998 so I'll take your word for it.  

When writing software for Linux, the actual CPU architecture usually does not matter much anyway, if it matters at all. There are a few cases, where the architecture does matter (e.g. emulators which you want to optimize for the target architecture), but unless you are into writing ARM assembly, the CPU architecture really does not matter much. When talking about portability, writing software to run on different operating systems is usually much more effort, than recompiling software which runs on Linux x86 to run on Linux ARM. 

Okay I agree for the most part, most things do just compile easy to ARM.. Then there are the things that don't.. I find that challenge to be fun. 

 

[wej]

I find that challenge to be fun.

There is nothing wrong with that.