[OpenRISC] [PATCH v2 07/13] or1k: math soft float support

[Stafford Horne <shorne@gmail.com>]

On Thu, Nov 18, 2021 at 10:28:03PM +0000, Joseph Myers wrote:
> On Sat, 13 Nov 2021, Stafford Horne via Libc-alpha wrote:
> 
> > diff --git a/sysdeps/or1k/bits/fenv.h b/sysdeps/or1k/bits/fenv.h
> > new file mode 100644
> > index 0000000000..49194ad851
> > --- /dev/null
> > +++ b/sysdeps/or1k/bits/fenv.h
> > @@ -0,0 +1,68 @@
> > +/* Copyright (C) 2021 Free Software Foundation, Inc.
> 
> Missing one-line description at start of file.

OK Added for v3.
 
> The constants you're defining here seem to be specific to hard float, but 
> they're defined unconditionally.  Is this because you use the same ABI for 
> hard float and soft float (which is generally the case when such hard 
> float constants might be defined for soft float as well - a soft float 
> compilation could be using a hard float libc that supports those 
> constants)?

Yes, it's the same ABI.  I have patches for hard float but am not submitting
until the Linux side is sorted out as mentioned in the comment.  It seems
delcairing these conditionally was normal, hence I didn't mention anything about
it.

I didn't think as far as allowing for shared libc usage between soft/float
float, but I think that would work.

> > +/* Define bits representing exceptions in the FPCSR status word.  */
> > +enum
> > +  {
> > +    FE_OVERFLOW =
> > +#define FE_OVERFLOW	1 << 3
> 
> Macros need to be properly surrounded by parentheses so they always group 
> as a single operand in any expression.
> 
> > +#define FP_EX_OVERFLOW		1 << 3
> > +#define FP_EX_UNDERFLOW		1 << 4
> > +#define FP_EX_INEXACT		1 << 8
> > +#define FP_EX_INVALID		1 << 9
> > +#define FP_EX_DIVZERO		1 << 11
> 
> Likewise in sfp-machine.h.

OK, I have cleaned those up.

> -- 
> Joseph S. Myers
> joseph at codesourcery.com