Re: [PATCH, rs6000] Optimization for vec_xl_sext

[Bill Schmidt <wschmidt@linux.ibm.com>]

Hi Hao Chen,

I don't understand.  This patch was already approved and you committed it. :-)  I know
because I needed to make corresponding adjustments to the new builtins code.

Thanks,
Bill

On 11/15/21 8:16 PM, HAO CHEN GUI wrote:
> Hi,
>
>    The patch optimizes the code generation for vec_xl_sext builtin. Now all the sign extensions are done on VSX registers directly.
>
>    Bootstrapped and tested on powerpc64le-linux with no regressions. Is this okay for trunk? Any recommendations? Thanks a lot.
>
> ChangeLog
>
> 2021-11-16 Haochen Gui <guihaoc@linux.ibm.com>
>
> gcc/
>         * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin): Modify
>         the expansion for sign extension. All extensions are done on VSX
>         registers.
>
> gcc/testsuite/
>         * gcc.target/powerpc/p10_vec_xl_sext.c: New test.
>
> patch.diff
>
> diff --git a/gcc/config/rs6000/rs6000-call.c b/gcc/config/rs6000/rs6000-call.c
> index b4e13af4dc6..587e9fa2a2a 100644
> --- a/gcc/config/rs6000/rs6000-call.c
> +++ b/gcc/config/rs6000/rs6000-call.c
> @@ -9779,7 +9779,7 @@ altivec_expand_lxvr_builtin (enum insn_code icode, tree exp, rtx target, bool bl
>
>    if (sign_extend)
>      {
> -      rtx discratch = gen_reg_rtx (DImode);
> +      rtx discratch = gen_reg_rtx (V2DImode);
>        rtx tiscratch = gen_reg_rtx (TImode);
>
>        /* Emit the lxvr*x insn.  */
> @@ -9788,20 +9788,31 @@ altivec_expand_lxvr_builtin (enum insn_code icode, tree exp, rtx target, bool bl
>         return 0;
>        emit_insn (pat);
>
> -      /* Emit a sign extension from QI,HI,WI to double (DI).  */
> -      rtx scratch = gen_lowpart (smode, tiscratch);
> +      /* Emit a sign extension from V16QI,V8HI,V4SI to V2DI.  */
> +      rtx temp1, temp2;
>        if (icode == CODE_FOR_vsx_lxvrbx)
> -       emit_insn (gen_extendqidi2 (discratch, scratch));
> +       {
> +         temp1  = simplify_gen_subreg (V16QImode, tiscratch, TImode, 0);
> +         emit_insn (gen_vsx_sign_extend_qi_v2di (discratch, temp1));
> +       }
>        else if (icode == CODE_FOR_vsx_lxvrhx)
> -       emit_insn (gen_extendhidi2 (discratch, scratch));
> +       {
> +         temp1  = simplify_gen_subreg (V8HImode, tiscratch, TImode, 0);
> +         emit_insn (gen_vsx_sign_extend_hi_v2di (discratch, temp1));
> +       }
>        else if (icode == CODE_FOR_vsx_lxvrwx)
> -       emit_insn (gen_extendsidi2 (discratch, scratch));
> -      /*  Assign discratch directly if scratch is already DI.  */
> -      if (icode == CODE_FOR_vsx_lxvrdx)
> -       discratch = scratch;
> +       {
> +         temp1  = simplify_gen_subreg (V4SImode, tiscratch, TImode, 0);
> +         emit_insn (gen_vsx_sign_extend_si_v2di (discratch, temp1));
> +       }
> +      else if (icode == CODE_FOR_vsx_lxvrdx)
> +       discratch = simplify_gen_subreg (V2DImode, tiscratch, TImode, 0);
> +      else
> +       gcc_unreachable ();
>
> -      /* Emit the sign extension from DI (double) to TI (quad).  */
> -      emit_insn (gen_extendditi2 (target, discratch));
> +      /* Emit the sign extension from V2DI (double) to TI (quad).  */
> +      temp2 = simplify_gen_subreg (TImode, discratch, V2DImode, 0);
> +      emit_insn (gen_extendditi2_vector (target, temp2));
>
>        return target;
>      }
> diff --git a/gcc/testsuite/gcc.target/powerpc/p10_vec_xl_sext.c b/gcc/testsuite/gcc.target/powerpc/p10_vec_xl_sext.c
> new file mode 100644
> index 00000000000..78e72ac5425
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/powerpc/p10_vec_xl_sext.c
> @@ -0,0 +1,35 @@
> +/* { dg-do compile } */
> +/* { dg-require-effective-target int128 } */
> +/* { dg-require-effective-target power10_ok } */
> +/* { dg-options "-mdejagnu-cpu=power10 -O2" } */
> +
> +#include <altivec.h>
> +
> +vector signed __int128
> +foo1 (signed long a, signed char *b)
> +{
> +  return vec_xl_sext (a, b);
> +}
> +
> +vector signed __int128
> +foo2 (signed long a, signed short *b)
> +{
> +  return vec_xl_sext (a, b);
> +}
> +
> +vector signed __int128
> +foo3 (signed long a, signed int *b)
> +{
> +  return vec_xl_sext (a, b);
> +}
> +
> +vector signed __int128
> +foo4 (signed long a, signed long *b)
> +{
> +  return vec_xl_sext (a, b);
> +}
> +
> +/* { dg-final { scan-assembler-times {\mvextsd2q\M} 4 } } */
> +/* { dg-final { scan-assembler-times {\mvextsb2d\M} 1 } } */
> +/* { dg-final { scan-assembler-times {\mvextsh2d\M} 1 } } */
> +/* { dg-final { scan-assembler-times {\mvextsw2d\M} 1 } } */
>