Re: [PATCH 6/6] [RS6000] inline plt call sequences

[Segher Boessenkool <segher@kernel.crashing.org>]

On Wed, Nov 28, 2018 at 12:04:01PM +1030, Alan Modra wrote:
> On Tue, Nov 27, 2018 at 11:17:48AM -0600, Segher Boessenkool wrote:
> > That all sounds great :-)  Has this been tested with older binutils, too?
> 
> Yes, lots of times.  Sometimes by accident when I forgot to install
> a new binutils.  :-)
> 
> > > +(define_insn "*sibcall_indirect_nonlocal_sysv<mode>"
> > 
> > > +   (set (attr "length")
> > > +	(cond [(and (and (match_test "!rs6000_speculate_indirect_jumps")
> > > +			 (match_test "which_alternative != 1"))
> > > +		    (match_test "(INTVAL (operands[2]) & (CALL_V4_SET_FP_ARGS | CALL_V4_CLEAR_FP_ARGS))"))
> > > +		  (const_string "12")
> > > +	       (ior (and (match_test "!rs6000_speculate_indirect_jumps")
> > > +			 (match_test "which_alternative != 1"))
> > > +		   (match_test "(INTVAL (operands[2]) & (CALL_V4_SET_FP_ARGS | CALL_V4_CLEAR_FP_ARGS))"))
> > > +		  (const_string "8")]
> > > +	      (const_string "4")))])
> > 
> > Please use set_attr_alternative for these.
> 
> Are you sure about that?  The expression blows out to something like
> (completely untested):

But you don't need "cond", something with if_then_else is simpler.  And
(INTVAL (operands[2]) & (CALL_V4_SET_FP_ARGS | CALL_V4_CLEAR_FP_ARGS))
could be put into a macro or inline or even a predicate.

Those apply to both with set_attr_alternative and without, of course;
please use what is easiest to read and understand.

(Just committing without changes is fine as well of course).

Thanks,


Segher