If ... just if ...

[Trev]

If I had the money to burn, I'd just open source what I could (not GPL, probably something less restrictive) and set it free. But it's really not about buying "Amiga"--most of the core IP is worthless, and what isn't worthless has been replicated fairly well by others.

It's really about taking the Amiga trademark and turning it into something users can be proud of. When you look at Amiga OS, MorphOS, AROS, and AmigaAnywhere, what you see is a standard API, implemented in the same way UNIX vendors implement POSIX standards or the Single UNIX Specification. Whether or not that API is really the best way to do things isn't what's important. What's important is allowing it to grow and prosper. It really can't do that in a small, disparate community relying on mostly closed standards.

[Trev]

The things that made the original Amiga innovative are now ubiquitous. To move forward, the "Amiga" would have to make strides in parallelization and related areas. A complete paradigm shift is unlikely as users and the industry are too invested in the computing memes of our time--windowed GUIs, the mouse, peer-to-peer networking (not file sharing--peer-to-peer protocols like TCP/IP), etc. Much of what is innovative today revolves around new uses for existing technology, not innovations in the technology itself.

[Trev]

@da9000

Decomposition. Yes. Although I'd move that into software, probably the compiler. Now it's separate from the hardware, and the compiled code can scale appropriately depending on the target.

Everyone take note: without a major engineering breakthrough, processors are not going to get faster than they are today. It's all about rethinking how we solve problems.

@darule

That sounds like a great machine for continuing the classic legacy (although I'd ditch PCI and use PCIe and probably drop Zorro altogether); however, it's not going to move things forward.

Here's an excellent example of innovation using existing technology:

1. Ageia releases the PhysX SDK and companion hardware for accelerating physics calculations in games and other software.
2. nVidia buys Ageia, ports the PhysX middleware to its existing GPUs, and ends up with a solution that runs PhysX software an order of magnitude faster than the original companion hardware.

Here's my point: you don't need special purpose hardware. You just need a configuration flexible enough to allow your developers to make the best use of existing hardware. (Yes, "GPUs" are now general purpose processors.) That's what makes Cell, Tesla, and other low-cost, high-performance processor subsystems so attractive.

[Trev]

@darule

Microsoft wouldn't be the only developer that preferred CISC to RISC. Regardless, today's RISC processors are more like CISC processors than they are like the first generation RISC processors--lots of high level instructions that could have been implemented by the software developer as a series of low-level instructions. But really, that sums up any modern processor, doesn't it? Instructions are microcode that translate into primitives implemented in hardware. It keeps the hardware smaller, regardless of the core architecture.