Embedded FPGA processors

[mattmercaldo]

Hi Randy,

Long time!

Have you looked into the embedded FPGA processor domain for isomax?

I know there were talks many moons ago..

matt

[RMDumse]

Hey Matt, Great to hear from you. We should talk further.

The interest to expand IsoMax(TM) to other processors and also hardware developments is still strong. So far we've run it on the HC11, 332, HC12, DSP56F8xx and the LPC ARMs.

So far we haven't taken on the FPGA's and CPLD's. I had just been considering the CoolRunner II CPLD's, as something I'd like to have the ability to encode.

What are your thoughts?

[mattmercaldo]

I think there is a new trend in FPGAs. The mid level spartan and cyclone devices are getting faster and denser at a substantially lower cost.

You could embed a processor into the FPGA and create a morph that facilitates both embedded (slower) state design as well as native (fast) designs that interact.

You could use attributes in the state machine that tell what you expect to be targeting. The attribute could tell the compiler to create standard ISO code OR verilog.

Typical communications could be achieved by standard drop in queues. Queues would be a natural interface between the two domains.

I recently did a complex design with the spartan FPGA and was amazed at what could be done. (MPEG transport stream delivery at 54 MHZ with deep packet inspection)

I have also been thinking that the state machine design tool could be done in somethink like AJAX. Different business model. (not sure if you want to go off line on this with a subsequent discussion).

As the mid level FPGA chips get cheaper and able to do more, it may be reasonable to expect that the mid level processors and even DSPs will be overtaken by this new paradigm.

I think ISOMAX is an ideal candidate to optimize the toolset so that this approach is well, more approachable.

Just some thoughts.

[RMDumse]

I've seen some of the processor development, like the Picoblaze, Microblaze, 8051 IP cores, and others, etc. They are blazingly fast, faster than you can get from 8051 manufacturers, for instance.

The state machine concept of programming has caught on only sparingly. I still think there is a general unawareness of the utility among programmers, while hardware guys with only marginal software skills tend to be very cautious about learning anything new in languages. Well, even more so for the programming folks.

What I think is needed to make a break through where IsoMax(TM) becomes more mainstream, is something like an article explaining the "state" and deeper physical connections of physics (quantum) black holes (horizon area vs. information storage) and AI (state as a fundamental concept of learning and behavior multiplication. That's a pretty tall order.

But yes, I do like your idea of hardware and software states mixed. What can be computed without a CPU can be native gates, and what takes conditional computations to determine transitions and actions left for a software implementation.

Sorry, I know almost nothing about AJAX. I just looked it up to see it was a google promoted programming language. Hadn't crossed my path before now.

But yes, I think IsoMax(TM) is full of potential promise as both a programming and a hardware programming tool, and there are no distinct edges where one couldn't shift in the same paradigm between the software/firmware/hardware boundaries.

Email me if you want to discuss it further. Still rmd@newmicros.com

[RMDumse]

Matt, I tried the most recent email address I had on you, and it bounced. So if we are to have further contact, it needs to come from your side.