Programming

This is very platform-specific stuff. What platform are these odd floats coming from?

Kinda. This data basically come out from embedded devices as messages and I later have to process them. The spec, which I don't have at my hand right now, actually specify how many bits are for the mantissa. The smallest is actually 8 bits in my case. I must say that spec is just a draft yet, so it might change.

Thanks for all the links and tips to all so far.

Always eager to learn,
S.

The mantissa is just a portion of the number, so the whole thing could be quite a bit larger.

Is the physical memory on that embedded device so scarce that you need to roll your own for floats...also the memory savings will slow down your application because of all the extra CPU cycles needed to do the conversion. Can you say how much phys mem is going to be on that device. This wasn't an issue even a couple of decades ago when there was a real shortage of memory in embedded systems.

If I understood this correctly you are getting messages from an embedded device presumable in some kind of encoded struct but you are not actually coding on/for the embedded device.

If you are post-processing this on an "normal" PC you should just be decoding the messages and using the decoded content to do your processing. For example if you have an 8 bit float of some kind coming out from the device you would convert that and use a standard 32 bit float on your pc and ignore the fact that the data type is much bigger than you actually need.

Likewise is you need to send the information back to the device, you would need to reencode the message into a sequence of bytes and send that.

Adapting/converting the messages rather than reimplmenting the basic functionality will give you much better mileage.

That's right.

I'm not actually processing the data. I just have to display them (for analyzing),
but still I need a variable to contain them when I read the data from the file.

I don't need to resend it anywhere. I just need to contain them, and eventually display them.

I'm actually having the same problem with the integer being 128 bits long. stdint.h is
only able to handle up to uint64_t .

Thanks,
S.

Can you show the struct encoding...maybe it needs to be decoded properly before displaying. An embedded device sending a 128-bit int...what OS and platform are you working on. Also provide some details on the embedded device in use.