sc_signed_subref and sc_unsigned_subref

[DavidC]

Hi,

I have a trivial piece of code that looks a bit like this:

sc_bigint<N+1> p;    // Signed
sc_biguint<N>  q, r; // Unsigned

for(i=1; i<L; i++)
{
  ..... /* some logic */
  r = test ? p.range(i, 0) : q.range(i, 0);
}

 

However it won't compile as I get this error:

Quote

operands to ?: have different types 'sc_dt::sc_signed_subref' and 'sc_dt::sc_unsigned_subref'

 

I have to say I'm a bit disconcerted because (in my mind at least), the bit-slicing operation using the .range() operator returns a bit-pattern without any clearly defined interpretation (so I don't expect concepts such as a "signed" bit-pattern and an "unsigned" bit-pattern to even exist).

One question is of course the motivation for defining these concepts. Though I'm curious and interested to hear about it, I suppose there is a good reason for doing that and it is not the main point that I'm interested in.

The question I'm mostly interested in, is very practical: what is the clean way to get the code to compile with the desired effect ? Thanks !

 

[DavidC]

I guess I was a bit tired when asking that question... The practical solution is as trivial as a cast it seems:

  r = test ? (sc_biguint<N>)p.range(i, 0) : q.range(i, 0);

Having said that, I'm still curious to understand why this cast is necessary at all. In fact, the very same code based on sc_int / sc_uint rather than sc_bigint / sc_biguint works perfectly well without any cast.

Why does the .range() operator return signed/unsigned patterns when applied to sc_bigint, but not when applied to sc_int ?