Number of bytes required to encode F32

[ThibFrgsGmz]

Hi,

I am structuring my messages in ASN.1 and I see an increase of about 50% in the structures where there are F32s.

If I define a message containing only one F32, I see that it takes 13 bytes to encode it.

Is this normal? I use the default encoding uPER.

What is the procedure to get only 32 bits to encode the floats please?

[maxime-esa]

Assuming your type is defined with this range (corresponding to a 32bits float):

MyF32 ::= REAL (-3.4e38 .. 3.4e38)

For the binary encoding you can choose between various schemes when you invoke the compiler: either uPER (ASN.1's unaligned Packed Encoding Rules) with the -uPER flag or ACN with -ACN.

For real numbers, PER is not an optimal choice because it needs at least 13 bytes whatever precision you need. In fact it theoretically provides "infinite precision" which your do not need most of the time.

So to solve your issue, first create an ACN file for your grammar (f32.asn) like this:

$ asn1scc -ACND f32.asn

Edit the file that was generated (f32.acn) and add the following directive to specify that you want a single-precision encoding using 32 bits only:

MyF32[encoding IEEE754-1985-32]

... and check the resulting encoding layout:

$ asn1scc -icdAcn out.html -ACN f32.asn f32.acn

The resulting file (out.html) should contain this nice table:

In order to generate automatic encoders and decoders for this scheme in C, use the -c flag.

The complete user manual detailing all ACN directives is here:
https://taste.tuxfamily.org/wiki/index.php?title=Technical_topic:_ASN.1_-_An_introduction_to_ACN

[usr3-1415]

Moreover, you can specify the order of bytes in the encoding stream

MyF32[encoding IEEE754-1985-32, endianness little]
or
MyF32[encoding IEEE754-1985-32, endianness big]

[ThibFrgsGmz]

Thanks guys!