When generate HDL codes, can not support fixed point division. My dividend has 40 bits of fraction, and the divisor has 20, the result have 20 fractional bits.

Maybe the question is "En_35"? But I can not find more information in help center.Please help me.

[Zora Aria Jade]

The model can successfully realize function after using fixed-point tool.Just can not generate HDL code.

Answer 1

[Walter Roberson]

ufix40 divided by ufix20 would give a result with (40-20)=20 bits, but you are requesting en35 which is 35 bits.

You can create a ufix20 result and potentially convert it to _en35 as a separate step, understanding that the bottom 15 bits will not contain useful information.

[Zora Aria Jade]

I have tried creating a ufix20 result and converting it to _en35 .But then my model cannot work. Because the result is on the order of 1e-5, which cannot be roundoffed.the convert block change it to 0.

[Walter Roberson]

ufix40 is a 40 bit unsigned integer. ufix20 is a 20 bit unsigned integer. ufix40 divided by ufix20 gives a 20 bit unsigned integer, ufix20 . The smallest non-zero value that can handle is the integer 1.

Your target is ufix20_En35 which is a 20 bit integer with a fraction length of 35. The largest value that can handle is (2^20-1)*2^(-35) = 1048575/34359738368 which is about 3.05175490211695e-05

Therefore the only ufix20 value that ufix20_En35 can represent is the exact value 0. All other ufix20 results of the division would either generate an error or overlow to 1048575/34359738368 = 3.05175490211695e-05

If you want to get 35 bits of meaningful fractional value then you will need to convert your ufix40 and ufix20 before the division.

[Zora Aria Jade]

I feel extremely grateful!

[Zora Aria Jade]

Successfully!Do not need a convert block. Change the data type of "2" from ufix2 to unfix41_En36.