LONDON The information provided to me so far about the ARM-Xilinx programmable systems collaboration was long on canned quotes and short on detail which of course leads to all sorts of questions and speculations.
From where I am sitting it seems a fairly important announcement. ARM has worked with programmable logic vendors in the past, such as Atmel and Actel, but in heavily prescriptive ways. There have even been versions of older ARM processor cores that could be synthesized and smeared across FPGAs, although in general that was not considered a useful strategy; throwing away the power savings associated with ARM processor cores.
With the latest announcement ARM and Xilinx seem to be getting pretty strategic with each other. Xilinx gets the option to use the whole Cortex range from M0 to A9 (we think) and the two companies will work together on defining the next-generation AMBA bus, so it can be deployed on future FPGAs.
I am sure neither party would want to say that the deal is exclusive but it looks to me like ARM is backing Xilinx and Xilinx is backing ARM to deliver the future of multiprocessor-plus-FPGA-fabric programmable platforms.
ARM provides the processor and the software canon that covers applications from consumer and automotive to industrial. Xilinx provides the peripheral and communications functions, flexibility and reduced IC development cost compared with system-on-chip. This is vital for all those applications that are not expected to go to iPhone-like volumes.
Such an idea is not new. The industry considered the uniprocessor-plus-FPGA-fabric nearly ten years ago, in the form of companies such as Elixent Ltd. (Bristol, England), spun out of Hewlett-Packard Labs and subsequently acquired by Panasonic. There are other examples, indeed Xilinx has been offering PowerPC plus FPGA fabric ICs, but in general the concept has not seen a lot of take up.