Geek Answers Handbook

Test vectors for a floating point unit

I developed a 32-bit floating-point multiplier with one precision, as well as a separate Verilog add / subtract block.

I have a script in MATLAB that I wrote that generates a bunch of random floating point numbers and creates a test bench in Verilog.

I compared ncverilog simulation results with data from MATLAB and verified that my multiplier was working correctly and that my add / subtracter was working correctly.

However, I want to know a list of test cases that stress my block floating point. Is there such a list? I know some programs, such as Softfloat, but it checks the floating point of the computer, instead of combining it with the verilog test bench.

I should also mention that my block does not support exceptions such as overflow, overflow and does not support denormalized values. Rounding mode is rounded to the nearest.

Thanks!

+6
source share
2 answers

The canonical [published] set of test vectors for FP are those from Jerome Kuenen’s PhD thesis. They are available (on tape!) From IEEE, but I suspect that you can find them in a more convenient format with a little search.

+1
source

More articles:


All Articles