This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Reload generate invalid instruction on ppc64


I found the root cause.

In function rs6000_preferred_reload_class, it specifically check the
case that reload 0 into a VSX register, then the target reload class
is VSX register. VSX instructions can't load a constant into VSX
registers directly, I guess the author wanted to use a SUB or XOR
instruction to get a 0 value.
Then function gen_reload calls gen_move_insn to generate the reload
instruction, which actually generates a movdi_internal64 insn, and it
doesn't contain a constraint to handle the 0->VSX register case, and
causes ICE.

So it can be fixed by adding a new constraint to movdi_internal64 to
load 0 to VSX register.

thanks
Guozhi Wei

On Tue, Aug 5, 2014 at 1:38 PM, Segher Boessenkool
<segher@kernel.crashing.org> wrote:
> On Tue, Aug 05, 2014 at 01:32:00PM +0930, Alan Modra wrote:
>> On Mon, Aug 04, 2014 at 05:54:04PM -0700, Carrot Wei wrote:
>> > Another problem is in the definition of insn pattern "*movdi_internal64".
>> >
>> > (define_insn "*movdi_internal64"
>> >   [(set (match_operand:DI 0 "nonimmediate_operand"
>> > "=Y,r,r,r,r,r,?m,?*d,?*d,r,*h,*h,r,?*wg,r,?*wm")
>> >         (match_operand:DI 1 "input_operand"
>> > "r,Y,r,I,L,nF,d,m,d,*h,r,0,*wg,r,*wm,r"))]
>> >   "TARGET_POWERPC64
>> >    && (gpc_reg_operand (operands[0], DImode)
>> >        || gpc_reg_operand (operands[1], DImode))"
>> >
>> > The predicates of this insn pattern allow the moving of an integer to
>> > VSX register, but there is no constraint allow this case. Can this
>> > cause problem in reload?
>>
>> Probably, just as you found with fprs.  The underlying issue is that
>> the operand predicates don't match the operand constraints.  What's
>> more, you can't make them match without breaking up the insn, or
>> adding a whole lot of extra alternatives.
>
> Can you disallow this in the condition of the pattern, just like it
> already requires at least one of the operands to be gpc_reg_operand?
>
>
> Segher


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]