[Bug libc/6411] PowerPC: Extend fpu fenv operations to operate on 64-bit FPSCR

["rsa at us dot ibm dot com" <sourceware-bugzilla at sourceware dot org>]

------- Additional Comments From rsa at us dot ibm dot com  2008-09-04 15:41 -------
Created an attachment (id=2938)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=2938&action=view)
enable 64-bit FPSCR but don't touch fpu_control.h

Hi Ulrich,

After talking with Steve Munroe we're concerned with how previous users may be
using the control word vs. how they should be.	They should only be relying on
bits that have masks provided by fpu_control.h but our feeling is that they're
probably relying on a 1:1 correspondence with the bits in the 32-bit FPSCR.

With this in mind I think we should probably leave the _FPU_[GS]ETCW macros as
they currently are and direct users to get the high 32-bits of the 64-bit FPSCR
in another way.  This way we don't break any existing applications which
stupidly rely on 1:1 mapping of FPSCR to fpu_control_t.

The rest of the patch still stands. I've respun the patch to remove
modification of fpu_control.h which means I can eliminate the new tests as
well.

In my standalone DFP library I'll create a wrapper around fpu_control.h that
does an include_next after the following check:

#ifdef #ifdef __STDC_WANT_DEC_FP__
# warning "Please use the fe_dec_getround() and fe_dec_setround() functions to
query the decimal rounding mode from the FPSCR."
#endif

This would then cover the bases for soft-decimal rounding mode as well.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
Attachment #2918 is|0                           |1
           obsolete|                            |


http://sourceware.org/bugzilla/show_bug.cgi?id=6411

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.