Re: [PATCH, ARM] Fix Neon floating-point constant encoding & support floating-point syntax

[Nick Clifton <nickc at redhat dot com>]

Hi Julian,

> Thanks! I've got a slightly-improved version of this patch thanks to 
> offline discussion with Paul Brook though, which uses a generic function 
> to help with the printing of float immediates. It relies on this 
> (applied) fix:
>
> http://gcc.gnu.org/ml/gcc-patches/2006-04/msg00925.html
>
> The only differences are in arm-dis.c, and a small cleanup in 
> is_quarter_float. Sorry for the hassle (the difference is probably small 
> enough to count as obvious), but is this still OK to apply on mainline? 
> (And also on the CSL branch?).

OK for mainline.

Check with CodeSourcery for permission to check into the CSL branch, but 
I expect that since you have been talking to Paul that they will have no 
objection.

Cheers
  Nick

> ChangeLog (gas):
>
>     * gas/config/tc-arm.c (neon_is_quarter_float): Move, and rename
>     to...
>     (is_quarter_float): Rename from above. Simplify slightly.
>     (parse_qfloat_immediate): Parse a "quarter precision" floating-point
>     number.
>     (parse_neon_mov): Parse floating-point constants.
>     (neon_qfloat_bits): Fix encoding.
>     (neon_cmode_for_move_imm): Tweak to use floating-point encoding in
>     preference to integer encoding when using the F32 type.
>
> ChangeLog (gas/testsuite):
>
>     * gas/testsuite/gas/arm/neon-const.s: New testcase. Neon
>     floating-point constants.
>     * gas/testsuite/gas/arm/neon-const.d: Expected output of above.
>     * gas/testsuite/gas/arm/neon-cov.d: Expect floating-point
>     disassembly for VMOV.F32.
>
> ChangeLog (opcodes):
>
>     * arm-dis.c (print_insn_neon): Disassemble floating-point constant
>     VMOV.


PS. Usually you would either delete these files or else put them into 
your global .cvsignore before performing the diff:

> ? bfd/doc/bfd.info
> ? binutils/doc/binutils.info
> ? gas/doc/as.info
> ? ld/ld.info