Using a float to store 16 2-bit ints

Projects Programming
1

Hi!

I have some floats (they must be floats because they're used elsewhere, as they're multipurpose) that I would like to use to store 16 2-bit integers.

This would be easy enough using bitwise operators, but those are not available for floats.

Here's some simple code to demonstrate to problem:

volatile float a=0;

void setup() {
  // put your setup code here, to run once:
  a=reinterpret_cast<uint32_t>(a)>>2;
}

void loop() {
  // put your main code here, to run repeatedly:

}

With or without the reinterpret_cast, it throws an error.

I tried with union too, but there's nothing to show for it (union also wouldn't work for what I'm doing, as I need to do left- and right-shifts)

I do not want to use a uint32_t to store these variables, as I'm optimizing for size.

Any suggestions?

2

Fuzzyzilla:
I do not want to use a uint32_t to store these variables, as I'm optimizing for size.

Since float is 32 bits why will uint32 be worse than float ?

4

I am not using floats in place of uint32_t's to save size - I know they're the same. I'm using them because I need floats elsewhere.

What I'm making here is a OS with apps. The apps share a common temporary memory of 10 floats (for the reason that floats can store ints OR decimals when needed). However, in one app, I want to store 16 2-bit ints, hence why I want to recycle one of the general-purpose floats.

Thanks for the quick replies!

6

What I meant is I do not want to add another variable for this one app, as I cannot change the type of the floats.

Can I have a little help with that? I was unable to do it, as I said in the OP.

Should I use a union between float and uint32_t to achieve right- and left-shifts?

Sorry for my confusing posts, but thank you for the help!

7

for the reason that floats can store ints OR decimals when needed)

There is no such thing as a "decimal" nor con a float store an int. It's clear to me that you have no understanding of how data is stored. SO GO LOOK IT UP.

Mark

8

Thanks, oh so helpful sir. You see, this is why I love this forum oh so much.

I KNOW a float cannot store an int - but it comes close.

I KNOW a float cannot store a fractional number, but it comes close.

9 
typedef union
{
  float as_float;
  struct
  {
    uint16_t thing_one;
    uint16_t thing_two;
  }
  as_struct;
  uint16_t as_array[2];
}
thirty_two_bits_of_whatever;
10

Alright, I think I've figured out a way using reinterpret_cast.

volatile float a=0;

void setup() {
  // put your setup code here, to run once:
  uint32_t *binary=reinterpret_cast<uint32_t>(&a);
  uint32_t value=*binary<<2;
}

void loop() {
  // put your main code here, to run repeatedly:

}

EDIT: Thanks Coding Badly! Which should I use: union or reinterpret? (if it makes any difference, I hope to migrate to a 32-bit architecture, and I know some things can have incompatibility issues)

This is untested, but it compiles at least.

11

Using a float to store 16 2-bit ints

If you know it can't why bloody weel ask!

Mark

12

Fuzzyzilla:
Which should I use: union or reinterpret?

As far as I know the only difference regards aliasing. A quick Google search indicates that union is a better choice in such matters.

However, I believe aliasing problems can only occur if you use both pointers (pointer to float and pointer to uint16_t) in the same context.

I tend to favour unions because I believe it makes the code easier to read.

(if it makes any difference, I hope to migrate to a 32-bit architecture, and I know some things can have incompatibility issues)

Packing, padding, and byte order. Three new things for you to learn.