Hi!
On Fri, Oct 30, 2020 at 03:19:12PM -0400, Vladimir Makarov wrote:
> The following patch fixes failures for test p9-extract-2.c on
> ppc64. The failures are a result of committing patch dealing with insn
> scratches in IRA. The pseudo corresponding the 1st scratch in the
> following insn get unexpected register class (general regs) and
> unexpected insn alternative (the 2nd one).
>
> ;; Optimize stores to use the ISA 3.0 scalar store instructions
> (define_insn_and_split "*vsx_extract_<mode>_store_p9"
> [(set (match_operand:<VS_scalar> 0 "memory_operand" "=Z,m")
> (vec_select:<VS_scalar>
> (match_operand:VSX_EXTRACT_I 1 "gpc_reg_operand"
> "<VSX_EX>,v")
> (parallel [(match_operand:QI 2 "const_int_operand"
> "n,n")])))
> (clobber (match_scratch:<VS_scalar> 3 "=<VSX_EX>,&r"))
> (clobber (match_scratch:SI 4 "=X,&r"))]
> "VECTOR_MEM_VSX_P (<MODE>mode) && TARGET_VEXTRACTUB"
>
> Actually getting the right hard register in LRA before the patch was a luck.
>
> The following patch fixes the failures by adding hints * to the constraints.
>
> Is it ok to commit?
>
> 2020-10-30 Vladimir Makarov <vmaka...@redhat.com>
>
> * config/rs6000/vsx.md (*vsx_extract_<mode>_store_p9): Add
> hints * to 1st scratch.
>
Thanks for the patch! But it has a problem:
> diff --git a/gcc/config/rs6000/vsx.md b/gcc/config/rs6000/vsx.md
> index 67e4f2fd037..78de85ccbbb 100644
> --- a/gcc/config/rs6000/vsx.md
> +++ b/gcc/config/rs6000/vsx.md
> @@ -3717,7 +3717,7 @@
> (vec_select:<VS_scalar>
> (match_operand:VSX_EXTRACT_I 1 "gpc_reg_operand" "<VSX_EX>,v")
> (parallel [(match_operand:QI 2 "const_int_operand" "n,n")])))
> - (clobber (match_scratch:<VS_scalar> 3 "=<VSX_EX>,&r"))
> + (clobber (match_scratch:<VS_scalar> 3 "=*<VSX_EX>,&*r"))
> (clobber (match_scratch:SI 4 "=X,&r"))]
> "VECTOR_MEM_VSX_P (<MODE>mode) && TARGET_VEXTRACTUB"
> "#"
You add * to both alternatives here? I would expect adding it to only
the second alternative, does it work better with both?
That also avoids a different problem: *<VSX_EX> won't work as expected.
'*' in IRA skips one constraint character, but <VSX_EX> can be "wa", a
two-letter constraint (and we do have an "a" constraint as well,
something wholly different: "wa" means a VSX register, while "a" is an
indexed address).
case '*':
/* Ignore the next letter for this pass. */
c = *++p;
break;
Segher
