[PATCH] emit-rtl.c: Allow vector subreg of float vectors

Richard Sandiford richard.sandiford@arm.com
Mon Aug 10 16:22:58 GMT 2020


Andrew Stubbs <ams@codesourcery.com> writes:
> On 06/08/2020 04:54, Richard Sandiford wrote:
>>> diff --git a/gcc/emit-rtl.c b/gcc/emit-rtl.c
>>> index f9b0e9714d9..d7067989ad7 100644
>>> --- a/gcc/emit-rtl.c
>>> +++ b/gcc/emit-rtl.c
>>> @@ -947,6 +947,11 @@ validate_subreg (machine_mode omode, machine_mode imode,
>>>     else if (VECTOR_MODE_P (omode)
>>>   	   && GET_MODE_INNER (omode) == GET_MODE_INNER (imode))
>>>       ;
>>> +  /* Allow sections of vectors, both smaller and paradoxical.  (This just
>>> +     works for integers, but needs to be explicitly allowed for floats.)  */
>>> +  else if (VECTOR_MODE_P (omode) && VECTOR_MODE_P (imode)
>>> +	   && GET_MODE_INNER (omode) == GET_MODE_INNER (imode))
>>> +    ;
>> 
>> Might be missing something, but isn't this a subcondition of the
>> previous “else if”?  It looks like that ought to catch every case
>> that the new one does.
>
> Apparently, Hongtao and Uroš fixed my problem while I was working on this.
>
> Yes, my patch does the same (although I would question whether it's safe 
> to use "GET_MODE_INNER (imode)" without having first confirmed "imode" 
> is a vector mode).

FWIW, skipping the VECTOR_MODE_P test means that we also allow vector
paradoxical subregs of scalars, which can be useful if you're trying
to make a vector in which only element 0 is initialised.  (Only works
for element 0 though.)  IIRC this is what some of the x86 patterns do.

Guess it means we also allow vector subregs of complex values,
but that kind-of seems OK/useful too.

Thanks,
Richard


More information about the Gcc-patches mailing list