Convert Unsigned_32 to Ieee_Float_32

Convert Unsigned_32 to Ieee_Float_32

[Richard Charts]

Well I ran into a fun error today.

Apparently on Green Hills,
Interfaces.Unsigned_32( [Some Item of Type Interfaces.Ieee_Float_32])
causes a compiler error.
C:\DOCUME~1\...\LOCALS~1\Temp\gh_lk1.s 2212: lwz f29,112(sp)
C:\DOCUME~1\...\LOCALS~1\Temp\gh_lk1.s 2212: ----^-
C:\DOCUME~1\...\LOCALS~1\Temp\gh_lk1.s 2212: expected register

So just to get my project compiling, I substituted:
Interfaces.Unsigned_32'Value(Interfaces.Ieee_Float_32'Image(Real_Obj.Resolution))

But that's a rather expensive convertion.
Can anyone recommend a better way until, I find a way around the above
error?
Thanks.

Re: Convert Unsigned_32 to Ieee_Float_32

[Ian Caldwell]

Just an idea, I don't have Green Hills so can't test it.

Go via an intermediate signed integer, if all you values will fit in a signed integer.

Re: Convert Unsigned_32 to Ieee_Float_32

[Keith Thompson]

"Richard Charts" <richard.charts@gmail.com> writes:
> Well I ran into a fun error today.
>
> Apparently on Green Hills,
> Interfaces.Unsigned_32( [Some Item of Type Interfaces.Ieee_Float_32])
> causes a compiler error.
> C:\DOCUME~1\...\LOCALS~1\Temp\gh_lk1.s 2212: lwz f29,112(sp)
> C:\DOCUME~1\...\LOCALS~1\Temp\gh_lk1.s 2212: ----^-
> C:\DOCUME~1\...\LOCALS~1\Temp\gh_lk1.s 2212: expected register
>
> So just to get my project compiling, I substituted:
> Interfaces.Unsigned_32'Value(Interfaces.Ieee_Float_32'Image(Real_Obj.Resolution))
>
> But that's a rather expensive convertion.
> Can anyone recommend a better way until, I find a way around the above
> error?

That's not going to work anyway (though, as you say, it will compile).
The string returned by Ieee_Float_32'Image isn't going to be a valid
integer literal, so Unsigned_32'Value will raise Constraint_Error.

The first thing I'd try is to do the conversion via some other
intermediate type; maybe the bug doesn't happen to affect Float_64,
for example.  If that fails, you could write your own brute-force
conversion function.  If you just want to get it working quickly,
without worrying about efficiency, you could instantiate Float_IO for
tpe IEEE_Float_32, then use Put with Fore => 0, Aft => 0, Exp => 0 to
write it to a string, delete or otherwise ignore the trailing ".0" (or
".1", or whatever), and apply Unsigned_32'Value to that.

-- 
Keith Thompson (The_Other_Keith) kst-u@mib.org  <http://www.ghoti.net/~kst>
San Diego Supercomputer Center             <*>  <http://users.sdsc.edu/~kst>
We must do something.  This is something.  Therefore, we must do this.