Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
12 years ago

Floating point format support

I am trying to determine what kind of support is available in Altera tools, and IP for floating point formats. From what I have seen so far, it appears single precision and double precision is mostly supported. Although there is a warning regarding double precision when using OpenCL. I also see that user defined precision formats are also supported in some megafunctions.

One particular floating point format I am interested in is half-precision (16 bit). Is this format offered? (I would expect it is for at least the IP that support custom/user defined precision formats.

Thanks

2 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Arria 10 has hardware IEEE-754 support.

    What kind of application are you working on which requires the dynamic range of FP? Most people I talk to seem to require FP since that's what they are used to in their previous DSP, their Matlab model uses FP, etc.