Why dynamic partial reconfiguration is still not there?

If I had a great idea in this area, I would demonstrate it in simulation and then ring up a venture capitalist.

-- Mike Treseler

Researchers are always very enthusiastic about many things and so are engineers when hearing about new ways and possibilities. :-)

Practical engineers furthermore see the requirements of development processes as well, such as time to market, costs, development complexity, documentation, testability etc ...

For me, working (among others) for the medical products industry too, I can say, that up to now there are many (even simpler) techniques, methods and devices, which are easier to handle and to test, which are not accepted by the customer and thus not realized. I once liked to integrate an embedded webserver in one deivce and found, that the chip "is not designed for medical product" - which does in fact mean, that it was exclude just for security purposes - the device might have worked nicely!

technology in fields, where there are high demands of safety and security and where much testing has to be performed before beeing able (or allowed) to release a product. Think of e.g. the problem of system tests: Every change of the basic function of a health care product by a change of software / firmware upgrade (and changing FPGA code is nothing else) requires a test and formal recalibration of the instrument. Sometimes it has to be made sure, that only specific persons are allowed to do this, and have to do this by intention. Therefore, many processes are established inside software and hardware (also FPGA) to PROHIBIT such changes. For example, in my current project, there are redundancy mechanisms, which disallow the change of certain parameters (only parameters !!!) by a CPU , because the CPU could be mislead and do this coincidentially (theoretically). Much FMEA has to be done to proove this and documentate it.

Now your idea is, to "open a door" for reconfiguration - meaning "replacing funcionality" ? I can imagine the shocked faces of Dr X in the clinc Y who is responsible for the formal correct processes according GMP and similar regulations and laws.

Also, many medical products have to be checked by a clinical test, which takes much time and is very expensive, too. Even if you were able to establish a safe system for reconfiguration and could convince the FDA inspectors to accept it - you probably will have to test your device under all possible configurations. Nobody will do and poy this. So it is (much) easier to integrate all functions into a big FPGA , test it once and completely and the just do not use some functions. This makes about some $200 for a bigger or a second FPGA but saves money around a Million.

FPGA reconfiguration might in interessting for consumer products, possibly. But I cannot point you to an application, either.

Yes.

If every beneficial technology were commercially exploitable by a small startup company I think the computing world would be quite a different place.

- a

Hmm. I don't remember saying that.

Not every idea is a great one, and it is not up to the FPGA makers to make hardware the way I happen to prefer it.

-- Mike Treseler