Floating-point conversion error

[ekiwi]

There is a bug in the tests provided by #2318.

tests/correctness/ieee754-float/parse-and-print.futil.data provides the input value 15.1239. However, at the end of the test, the value has changed to 15.12 in tests/correctness/ieee754-float/parse-and-print.expect.

This is even though a 32-bit IEEE float should be able to represent 15.1239. So there appears to be a bug, most likely in how the data is converted back after the simulation.

[rachitnigam]

Argh, thanks for catching this! Also tagging @jiahanxie353 in case he can get to this before me.

[EclecticGriffin]

Probably an easy fix? Best guess is that the problem comes from this line:

calyx/fud/fud/stages/verilator/json_to_dat.py

Line 45 in 6b48969

return fp.to_dec(round_place=2)