On 2/11/19 8:11 AM, Segher Boessenkool wrote:
> On Mon, Feb 11, 2019 at 07:17:16AM -0600, Bill Schmidt wrote:
>> At -O0 (if I hand-inline everything myself to avoid errors), we scalarize
>> the modulo/masking operation into a rldicl for each doubleword. I really
>> don't see any reason to change the code.
> So what does this look like at expand (at -O0)? Is it something that
> is done at gimple level, is it expand itself, is it some target thing?
It's already a mask at expand, even at -O0. vregs dump excerpt:
(insn 13 12 14 2 (set (reg:DI 129 [ _16 ])
(vec_select:DI (reg:V2DI 126 [ _9 ])
(parallel [
(const_int 0 [0])
]))) "vec-srad-modulo.c":40:10 1223 {vsx_extract_v2di}
(nil))
(insn 14 13 15 2 (set (reg:DI 130 [ _17 ])
(and:DI (reg:DI 129 [ _16 ])
(const_int 63 [0x3f]))) "vec-srad-modulo.c":40:10 195 {anddi3_mask}
(nil))
(insn 15 14 16 2 (set (reg:DI 131 [ _18 ])
(vec_select:DI (reg:V2DI 126 [ _9 ])
(parallel [
(const_int 1 [0x1])
]))) "vec-srad-modulo.c":40:10 1223 {vsx_extract_v2di}
(nil))
(insn 16 15 17 2 (set (reg:DI 132 [ _19 ])
(and:DI (reg:DI 131 [ _18 ])
(const_int 63 [0x3f]))) "vec-srad-modulo.c":40:10 195 {anddi3_mask}
(nil))
>
>>>> For -mcpu=power9, we get close, but have some bad register allocation and
>>>> an unnecessary extend:
>>>>
>>>> xxspltib 0,4 <- why not just xxspltib 32,4?
>>>> xxlor 32,0,0 <- wasted copy
>>> Yeah, huh. Where does that come from... I blame splitters after reload.
>> This only happens at -O2 and up, FWIW. At -O1 we allocate the registers
>> reasonably.
> Heh.
>
>>>> Weird. I just tried adding -mvsx
>>> Does it _need_ VSX anyway? Are these builtins defined without it, too?
>> Yes (vector long long / V2DImode requires VSX).
> So something like
>
> /* { dg-do run } */
> /* { dg-require-effective-target vsx_hw } */
> /* { dg-options "-mvsx -O2" } */
>
> then?
What I have now:
/* { dg-do run { target { vsx_hw } } } */
/* { dg-options "-O2 -mvsx" } */
>
>> I pointed to the bugzilla in another reply -- which was "resolved" with a
>> hack.
>> I consider it still broken this way...
> Reopen that PR?
Already done. ;-)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88920
Bill
>
>> I tested a revised version of the patch overnight and will submit shortly.
> Thanks.
>
>
> Segher
>
