Re: [myhdl-list] Dynamic data type
Brought to you by:
jandecaluwe
From: Brendan R. <bre...@gm...> - 2008-05-22 18:10:46
|
Blubaugh, David A. <dblubaugh <at> belcan.com> writes: > To Whom It May Concern, > > I was wondering that if one were to include floating-point support for MyHDL by developing a support module for such number processing. I was wondering if MYHDL will be able to handle dynamic data type processing. Such as if I were to multiply a complex number (real and complex values are integers) with a floating point number and then finally dividing with an integer number. Will MYHDL and or python be able to handle the multiple different data types to create a final result of a floating-point complex number. Has anyone out there done any relevant work on this topic that will allow for myhdl to create synthesizable Verilog, that handles these multiple variable data type processing?? > Thanks, > > David Blubaugh Hi David, Including direct support for floating point conversion doesn't make sense. Floating point implementations tend to be architecture (device) specific. Anything that attempts to be generic would not be what people want anyway. What might make sense is something that helps you to convert your floating point algorithms to fixed point, though I'm not convinced that it should be a part of MyHDL. MyHDL should to being an excellent HDL experimentation and generation tool. Along the lines of floating-to-fixed conversion, I found this online: http://users.ece.utexas.edu/~bevans/projects/wordlength/converter/index.html Tom Dillon's company also has some pretty good information on this topic: http://www.dilloneng.com/ingenuity/fixed-vs-floating-point Cheers, - Brendan |