[Aegean]

I've researched this topic for a while in university, and the idea was to keep a general purpose FPGA as a programming device, and load preconfigured images to it when needed. For example load up an encryption module when encryption intensive operations took place, or load a video encoder/decoder whenever needed.

There were a few problems:

* Each FPGA vendor has its own architecture and such a technology requires momentum from these vendors. A 3rd party cannot do much about it.

* Dedicated ASIC circuits often perform better with less power consumption. So keeping a general purpose FPGA may not bring sufficient benefit to develop the technology.

[count]

Would it be possible to blend something like this into, say, GCC, so that when you compile your code, easily hardware accelerated sections (for whatever definition of that you like) can be generated into FPGA loadable files and used at runtime?

[Aegean]

Yes I thought about this idea. For example it would be nice to have a .fpga section in the ELF files and one could load images as part of OS file execution.

It is a nice idea and I would try to push this as a technology if I was Xilinx or Altera. But it requires that momentum to become a reality because this creates a new model and ABI for software - one that doesn't exist today.