Another piece, originating from the first quarter edition 2012 of the Xcell Journal, is "Embedded Vision: FPGAs’ Next Notable Technology Opportunity".
To access the article, click here: http://www.eetimes.com/design/military-aerospace-design/4376567/Embedded-vision--FPGAs--next-technology-opportunity
There are some very useful concepts covered here and the summary is generally spot on, however it is important to consider the DSP function in the system context.
If other functions are required in addition to the DSP it may push the solution of choice into an FPGA. For example, a low performance DSP function that is part of a sensor or motor control system need not be implemented in a stand alone DSP device. An FPGA might be the right solution in this case.
Further information on how Xilinx System Generator and other HDL code generation tools may be used with MATLAB and Simulink -- including examples, demos, and videos -- are available from http://www.mathworks.com/fpga.
It is too easy to critisize and rush to jugment in haste, when it is so easy to offer the correction of such typo that appears in many publications that are transcribed by non-technical people.
Simply stating the transgression in this case that the index coefficients are transposed for the constatnt term and the discrete variable term, would have accomplished more and been more informative to others that may not have see this simple transgression. Article as good merits otherwise, in my opinion.
COuldn't the same be said in the other direction?, if there are other considerations more suitable for a DSP processor then the tilt can be as well move the other direction.
In my opinion, it has to be a case by case decision of the designer, given a set of resources and time constraints.
Just another point of view