[Arm-netbook] [review] SoC proposal

[Alec Ari]

> Date: Wed, 8 Feb 2012 19:58:44 +0000
> From: lkcl luke <luke.leighton at gmail.com>
> Subject: [Arm-netbook] [review] SoC proposal


> the plan is, therefore, to target this CPU at a very very
> wide range
> of markets, based on it having:
> 
> * 8 CPUs at 1.2ghz or above
> * SMP Cache Coherency
> * 32-bit DDR3 1333mhz RAM (with a 2nd version having 2 DDR3
> interfaces)
> * virtually everything software-programmable (with the
> exception of
> CABAC decode)
> 
> meaning that it will do 3D graphics _and_ 1080p Video
> entirely in
> software.

Sometimes having things done in software doesn't actually make things easier. I've been studying the Mesa3D DRI/Gallium infrastructure for about 2 1/2 years or so now, and KMS (Kernel mode setting) code, and I can tell you one thing for sure.

Considering that GPUs such as RadeonHD already have a lot of low-level / hardware support for 3D, asking to have no hardware acceleration and have _EVERYTHING_ done in software is practically suicide.. The open source drivers for Radeon cards are still far behind FGLRX / AMD WDDM drivers (FGLRX shares a lot of the same code with the Windows drivers) and Mesa has been in development since 1993.. The first r300 driver was released years ago (6 years? 5? 7?), and the open drivers for it are still lacking the performance of the closed drivers. Having a purely software accelerated driver for 1080p playback is asking for a lot of trouble.

You're asking for a ridiculously, and unrealistically optimized open source software engine, and to have 1080p playback.... Current x86 CPUs, take the AMD Phenom II for example isn't even that great for Mesa's softpipe (100% software driver for Gallium) and you're talking about a low power ARM processor, when not even a 125W CPU can keep up with such things. To be honest, I'm not even sure how AMD's Bulldozer family of CPUs would do with 1080p playback on softpipe. llvmpipe looks like a newer and faster alternative to softpipe, but even that driver has been in development for at least a year or two, and still has a long way to go, and I'm sure even the people who wrote it can agree with me on this.

Take a look at this page for an example:

http://unlimiteddetailtechnology.com/description.html

It's a lot of work, and I don't think this engine will even serve it's purpose anytime soon, let alone declared stable for usage. A software driver with the level of optimization of this graphics engine hasn't been done, nor close to, and Mesa seems to be the farthest along than anything as far as software drivers go, and even that is lacking, at least compared to the drivers written by thousands of Windows developers and AMD developers for the closed source display drivers.

I really don't see a problem with Mali, and with libv working on open source drivers for it, it seems like a much less of a headache, and we should be grateful for his efforts.

Perhaps I'm misunderstanding something, and I'm thinking I must be..

-Alec