Bit Width Specification

Every integer you declare in the C code can have a nonstandard bit width tailored to your application. The supported floating point bit widths are 16, 32, and 64. The choice of cores instantiated in the datapath will be based upon the bit width of the variables passed to them. Smaller bit width variables will be extended to take advantage of the larger cores unless no such core exists, in which case the variables will be truncated to use the largest core available.

Specifying the specific bit width is done by declaring a typedef at the beginning of your program. This typedef must be in the form of ROCCC_intX where X is any positive number. This can then be used to declare any variable with the appropriate size.

The idea is similar when specifying float bit sizes. The typedef must be of the form of ROCCC_floatX where X is either 16, 32, or 64. You can typedef as many different bit sizes as needed.