[Arm-netbook] Any ARM SoC has Open-Source access to hardware video decoder ?

[lkcl luke]

On Sun, Jul 1, 2012 at 5:46 PM, Mark Wilczynski
<mark_wilczynski at hotmail.com> wrote:
>> From: al4321 at gmail.com
>
>> How much the ICube's IC1 costs ?
>>
>> Also one of the BIG problems, at least initially, is that it uses MVP
>> architecture, which would require maintaining a full linux distro for
>> it (Debian?).
>> Is anyone learning this opportunity ?
>> Are there weak points in this design ?
>>
>
> I find it hard to believe that a software based GPU/VPU will offer
> competitive performance vs. fixed function hardware rasterizers, samplers,
> and decoders.  It has been tried before and failed.  Unless their
> instruction set is radically different from MMX, SSE, NEON, etc.

 it's radically different from MMX, SSE and NEON.  a hardware engineer
(icobecorp's chief hardware engineer) who used to work for ATI would
not bother wasting his time nor be part of a team that managed to get
significant venture capital unless he actually had a clue.

 translation: several hundred instructions have been added to the UPU.
 possibly as many as 2,000.

 but that's not the end of the story: we know full well that such
advanced instructions are utterly useless unless they are deeply
embedded into the compiler toolchain in an easy-to-use manner that
boils down at its crudest to "just hit compile".  that's why they have
a compiler expert on board with at least 20 years experience in
compiler technology.

 oh: i just remembered one weak point of their design (which points to
a long - and protracted - development cycle): it's 32-bit.  i believe
they *may* be working on 64-bit addressing for IC2.

 l.