This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
4.5: half-precision floating point for ARM Cortex A9
- From: Sandra Loosemore <sandra at codesourcery dot com>
- To: GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Wed, 21 Jan 2009 13:16:59 -0500
- Subject: 4.5: half-precision floating point for ARM Cortex A9
I just want to give folks a heads-up on this, to avoid possible duplication of
effort:
I will be submitting patches to implement half-precision floating point support
for the ARM Cortex A9 at some point after trunk re-opens for new features.
Half-precision (16-bit) floating point is a new feature in the IEEE 754-2008
standard and some of the pieces I've added for C and C++ language support are
not ARM-specific. However, since ARM had an existing implementation in their
own RealView compiler already, we've pretty much copied what they did in terms
of syntax and semantics. The main peculiarity is that this is a storage format
only, not an arithmetic format, so all __fp16 values are promoted to float in
expressions much the same way that char and short are promoted to int.
I originally made these changes on a 4.3-based toolchain and am just now
starting to update the patches and clean them up for submission, now that it
looks like we are getting close to forking 4.4. Since this is currently a
background task for me, it'll likely be a week or two before I have anything
ready to post.
-Sandra