Typing in binary numbers with decimal point

[Former user]

Originally reported on Google Code with ID 445

Typing in the decimal value 1,15625 and showing the binary result gives:
0b1.00101

But there is no way you can type in a binary number with decimal point e.g. 0b10.01
(which should be equal to 2.25).

Can this be fixed?

Reported by kent.kinna on 2013-08-20 13:34:58

[Former user]

Hi, Kent. I'm curious to know what your use case is that requires input of non-integer
binary numbers.

Reported by helder.pereira.correia on 2014-02-18 01:17:01

[Former user]

Hello, thanks for looking into this.

When representing a number with decimals in hardware (e.g. FPGA), the
decimal point could be set at any digit.
E.g.
A binary number:
101001 with 2 decimal bits will be 1010.01 (bin) = 10.25 (dec)
When verifying DSP functionality there might be a model that calculates in
real numbers e.g. MATLAB. Comparing the results between MATLAB and HW
(simulated or measured) might (or will at some ponit) end up in comparing a
MATLAB model with the HW.

I love the speedcrunch feature with fast conversion between different
bases, and conversion between a decimal value to a binary works like a
charm but not the other way around. I can't type in 0b1.1 to get the
decimal value 1,5.

I hope you understand why I found this functionality useful, please respond
of there are any more questions.

Reported by kent.kinna on 2014-02-20 06:23:09

[Former user]

Indeed, being able to play with non-integer numbers in binary-ish bases is useful when
designing fixed-point systems.

Is there a specific reason for this feature not being supported since the earliest
versions of SpeedCrunch ? Maybe I missed something, but the numbers parser in floatio.c
seems to support numbers with decimal points for every radix, and adding this support
to Evaluator::scan() seems to be enough to support the whole feature. I did a quick
and dirty implementation, and it seems to work properly so far.

Reported by [e-mail] on 2014-05-01 17:25:38

[Former user]

Good feedback!
Sounds great if there is a quite quick fix for this.
//
Kent

Reported by kent.kinna on 2014-05-02 05:20:01

[Former user]

@Tey: I don't think there's any reason besides available time and priorities. If you
have a patch, please post it for review as usual :)

Reported by helder.pereira.correia on 2014-05-05 16:50:37

[Former user]

Okay, here's a patch:
https://github.com/Tey/SpeedCrunch/commit/69dd5abd0eea0e05fcadc4f794134477a1e10fbc

I also tried adding support for exponents, but the number parser in the math lib does
not support that (though it can display such numbers).

Reported by [e-mail] on 2014-05-06 15:00:03

[Former user]

A more compact implementation (but which uses a state variable):
https://github.com/Tey/SpeedCrunch/commit/4ca56ea2ca46b98f6303c49c90841189f4c4841e

I forgot to remove the new parsing states from this commit though. Choose the one you
prefer :)

Reported by [e-mail] on 2014-05-06 15:14:38

[Former user]

Thanks, Tey. Added in revision 30a3ff3. I added another change for tests.

Reported by helder.pereira.correia on 2014-05-08 07:19:12 - Status changed: Fixed - Labels added: Type-Enhancement, Milestone-0.12