Re: [PATCH v8 16/17] powerpc: Add string-fza.h

[Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>]

On 13/01/23 15:27, Richard Henderson wrote:
> While ppc has the more important string functions in assembly,
> there are still a few generic routines used.
> 
> Use the Power 6 CMPB insn for testing of zeros.
> 
> Checked on powerpc64le-linux-gnu.
> Message-Id: <20230111204558.2402155-17-adhemerval.zanella@linaro.org>
> ---
>  sysdeps/powerpc/powerpc32/power6/string-fza.h |  1 +
>  sysdeps/powerpc/powerpc64/power6/string-fza.h |  1 +
>  sysdeps/powerpc/string-fza.h                  | 82 +++++++++++++++++++
>  3 files changed, 84 insertions(+)
>  create mode 100644 sysdeps/powerpc/powerpc32/power6/string-fza.h
>  create mode 100644 sysdeps/powerpc/powerpc64/power6/string-fza.h
>  create mode 100644 sysdeps/powerpc/string-fza.h
> 
> diff --git a/sysdeps/powerpc/powerpc32/power6/string-fza.h b/sysdeps/powerpc/powerpc32/power6/string-fza.h
> new file mode 100644
> index 0000000000..bb00d7c4f7
> --- /dev/null
> +++ b/sysdeps/powerpc/powerpc32/power6/string-fza.h
> @@ -0,0 +1 @@
> +#include <sysdeps/powerpc/power6/string-fza.h>
> diff --git a/sysdeps/powerpc/powerpc64/power6/string-fza.h b/sysdeps/powerpc/powerpc64/power6/string-fza.h
> new file mode 100644
> index 0000000000..bb00d7c4f7
> --- /dev/null
> +++ b/sysdeps/powerpc/powerpc64/power6/string-fza.h
> @@ -0,0 +1 @@
> +#include <sysdeps/powerpc/power6/string-fza.h>

We are trying to move from this sysdeps subarch folder organization because
it over-complicates and requires extra care to see if the correct file is
selected (for instance if a folder is not correctly set in the Implies file).

That is main driver for the math refactor I did some time ago (check for
instance a1cb1888b7aaa65cfa6d4e067c7516e2e08d8942).

> diff --git a/sysdeps/powerpc/string-fza.h b/sysdeps/powerpc/string-fza.h
> new file mode 100644
> index 0000000000..6a495d14d7
> --- /dev/null
> +++ b/sysdeps/powerpc/string-fza.h
> @@ -0,0 +1,82 @@
> +/* Zero byte detection; basics.  PowerPC version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.
> +   This file is part of the GNU C Library.
> +
> +   The GNU C Library is free software; you can redistribute it and/or
> +   modify it under the terms of the GNU Lesser General Public
> +   License as published by the Free Software Foundation; either
> +   version 2.1 of the License, or (at your option) any later version.
> +
> +   The GNU C Library is distributed in the hope that it will be useful,
> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> +   Lesser General Public License for more details.
> +
> +   You should have received a copy of the GNU Lesser General Public
> +   License along with the GNU C Library; if not, see
> +   <http://www.gnu.org/licenses/>.  */
> +
> +#ifndef _POWERPC_STRING_FZA_H
> +#define _POWERPC_STRING_FZA_H 1
> +
> +/* PowerISA 2.05 (POWER6) provides cmpb instruction.  */
> +
> +#include <limits.h>
> +#include <endian.h>
> +#include <stdint.h>
> +#include <string-optype.h>
> +
> +/* The functions return a byte mask.  */
> +typedef op_t find_t;
> +
> +/* Return the mask WORD shifted based on S_INT address value, to ignore
> +   values not presented in the aligned word read.  */
> +static __always_inline find_t
> +shift_find (find_t word, uintptr_t s)
> +{
> +  if (__BYTE_ORDER == __LITTLE_ENDIAN)
> +    return word >> (CHAR_BIT * (s % sizeof (op_t)));
> +  else
> +    return word << (CHAR_BIT * (s % sizeof (op_t)));
> +}
> +
> +/* This function returns 0xff for each byte that is
> +   equal between X1 and X2.  */
> +
> +static __always_inline find_t
> +find_eq_all (op_t x1, op_t x2)
> +{
> +  return __builtin_cmpb (x1, x2);
> +}
> +
> +/* This function returns 0xff for each byte that is zero in X.  */
> +
> +static __always_inline find_t
> +find_zero_all (op_t x)
> +{
> +  return find_eq_all (x, 0);
> +}
> +
> +/* Identify zero bytes in X1 or equality between X1 and X2.  */
> +
> +static __always_inline find_t
> +find_zero_eq_all (op_t x1, op_t x2)
> +{
> +  return find_zero_all (x1) | find_eq_all (x1, x2);
> +}
> +
> +/* Identify zero bytes in X1 or inequality between X1 and X2.  */
> +
> +static __always_inline find_t
> +find_zero_ne_all (op_t x1, op_t x2)
> +{
> +  return find_zero_all (x1) | ~find_eq_all (x1, x2);
> +}
> +
> +/* Define the "inexact" versions in terms of the exact versions.  */
> +#define find_zero_low		find_zero_all
> +#define find_eq_low		find_eq_all
> +#define find_zero_eq_low	find_zero_eq_all
> +#define find_zero_ne_low	find_zero_ne_all
> +
> +#endif /* _POWERPC_STRING_FZA_H  */