This is an example showing how to implement an arbitrary precision floating point multiplication core using SmartHLS. The source file fmult.hpp contains the function fmult() that implements floating point multiplication. The function works on any floating point format, with arbitrary mantissa and exponent width. The inputs to the function are of the class ap_fp (arbitrary precision floating point), a template class where the template parameters define the mantissa and exponent width.

This function also displays the use of the ap_uint class (arbitrary precision unsigned integer), which is a SmartHLS provided template class that represents unsigned integers of any bit width (defined by the template parameter).

The fmult_test.cpp file contains tests of the multiplication function. This is done using float and double types, as well as a half precision floating point format, implemented by the third party library in half.hpp.

Please follow the user guide to install SmartHLS and learn more about the features of this HLS tool:

• Homepage: https://www.microchip.com/en-us/products/fpgas-and-plds/fpga-and-soc-design-tools/smarthls-compiler
• User guide: https://onlinedocs.microchip.com/v2/keyword-lookup?keyword=fpgahls&redirect=true&version=latest

After the installation is done, you can try out this example through either GUI or command-line.

Using SmartHLS IDE

1. Launch SmartHLS IDE.
2. Import the example HLS project.
   • From the menu select: File -> Import.
   • In the import window select General -> "Existing Projects into Workspace". Click Next.
   • Click the Browse button beside "Select root directory". Navigate to where you cloned this fpga-hls-examples git repository.
   • Under Projects you should see the fp_mult project checked. Click Finish.
3. Try "Run Software", "Compile to Hardware", and "SW/HW Co-simulation".

Using SmartHLS Commandline

1. Make sure "$(SMARTHLS_INSTALLATION_DIR)/SmartHLS/bin is on the $(PATH)" environment variable to get access to the "shls" script.
2. run "shls sw" to run software.
3. run "shls hw" to compile to hardware.
4. run "shls cosim" to run sw/hw co-simulation.