Is the Verilog real data type synthesisable for a generic Xilinx FPGA? If not, what alternatives do I have for real number manipulation in synthesisable Verilog?
\$\begingroup\$
\$\endgroup\$
2
asked Apr 11, 2012 at 14:26
-
\$\begingroup\$ It depends on what tool you're using for synthesis... \$\endgroup\$– Ben VoigtApr 11, 2012 at 14:55
-
\$\begingroup\$ @BenVoigt: I'm using ISE. \$\endgroup\$– RandomblueApr 11, 2012 at 15:00
Add a comment
|
2 Answers
\$\begingroup\$
\$\endgroup\$
Short answer: No.
Your best bet would be to build your own floating-point block. There are open-source examples. While you can make it IEEE754 compatible, there is no need to. It all depends on your application.
Some applications requires the range provided by single/double-precision IEEE754 numbers. However, other applications only require a small range that can be emulated with fixed-point numbers instead.
\$\begingroup\$
\$\endgroup\$
No, but some FPGAs have a core that works with built-in hardware for floating point. This Altera white paper discusses some details. However you would have to instantiate them; they can't be inferred by ISE or most other tools.
answered Apr 11, 2012 at 16:00