public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: Noah Goldstein <goldstein.w.n@gmail.com>
To: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Cc: libc-alpha@sourceware.org,
	 Richard Henderson <richard.henderson@linaro.org>,
	Jeff Law <jeffreyalaw@gmail.com>,  Xi Ruoyao <xry111@xry111.site>
Subject: Re: [PATCH v11 10/29] string: Improve generic stpcpy
Date: Wed, 1 Feb 2023 13:47:57 -0600	[thread overview]
Message-ID: <CAFUsyfLtp41iKdV65QLt2LLY4dNC=HVxv5qbA__Gu5D9OHs=ww@mail.gmail.com> (raw)
In-Reply-To: <20230201170406.303978-11-adhemerval.zanella@linaro.org>

On Wed, Feb 1, 2023 at 11:04 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It follows the strategy:
>
>   - Align the destination on word boundary using byte operations.
>
>   - If source is also word aligned, read a word per time, check for
>     null (using has_zero from string-fzb.h), and write the remaining
>     bytes.
>
>   - If source is not word aligned, loop by aligning the source, and
>     merging the result of two reads.  Similar to aligned case,
>     check for null with has_zero, and write the remaining bytes if
>     null is found.
>
> Checked on x86_64-linux-gnu, i686-linux-gnu, powerpc64-linux-gnu,
> and powerpc-linux-gnu by removing the arch-specific assembly
> implementation and disabling multi-arch (it covers both LE and BE
> for 64 and 32 bits).
> ---
>  string/stpcpy.c | 92 +++++++++++++++++++++++++++++++++++++++++++++----
>  1 file changed, 86 insertions(+), 6 deletions(-)
>
> diff --git a/string/stpcpy.c b/string/stpcpy.c
> index 8df5065cfe..dd0fef12ef 100644
> --- a/string/stpcpy.c
> +++ b/string/stpcpy.c
> @@ -15,12 +15,12 @@
>     License along with the GNU C Library; if not, see
>     <https://www.gnu.org/licenses/>.  */
>
> -#ifdef HAVE_CONFIG_H
> -# include <config.h>
> -#endif
> -
>  #define NO_MEMPCPY_STPCPY_REDIRECT
>  #include <string.h>
> +#include <stdint.h>
> +#include <memcopy.h>
> +#include <string-fzb.h>
> +#include <string-misc.h>
>
>  #undef __stpcpy
>  #undef stpcpy
> @@ -29,12 +29,92 @@
>  # define STPCPY __stpcpy
>  #endif
>
> +static __always_inline char *
> +write_byte_from_word (op_t *dest, op_t word)
> +{
> +  char *d = (char *) dest;
> +  for (size_t i = 0; i < OPSIZ; i++, ++d)
> +    {
> +      char c = extractbyte (word, i);
> +      *d = c;
> +      if (c == '\0')
> +       break;
> +    }
> +  return d;
> +}
> +
> +static __always_inline char *
> +stpcpy_aligned_loop (op_t *restrict dst, const op_t *restrict src)
> +{
> +  op_t word;
> +  while (1)
> +    {
> +      word = *src++;
> +      if (has_zero (word))
> +       break;
> +      *dst++ = word;
> +    }
> +
> +  return write_byte_from_word (dst, word);
> +}
> +
> +static __always_inline char *
> +stpcpy_unaligned_loop (op_t *restrict dst, const op_t *restrict src,
> +                      uintptr_t ofs)
> +{

Is the perf cost of unaligned stores greater than merging?

> +  op_t w2a = *src++;
> +  uintptr_t sh_1 = ofs * CHAR_BIT;
> +  uintptr_t sh_2 = OPSIZ * CHAR_BIT - sh_1;
> +
> +  op_t w2 = MERGE (w2a, sh_1, (op_t)-1, sh_2);
> +  if (!has_zero (w2))
> +    {
> +      op_t w2b;
> +
> +      /* Unaligned loop.  The invariant is that W2B, which is "ahead" of W1,
> +        does not contain end-of-string.  Therefore it is safe (and necessary)
> +        to read another word from each while we do not have a difference.  */
> +      while (1)
> +       {
> +         w2b = *src++;
> +         w2 = MERGE (w2a, sh_1, w2b, sh_2);
> +         /* Check if there is zero on w2a.  */
> +         if (has_zero (w2))
> +           goto out;
> +         *dst++ = w2;
> +         if (has_zero (w2b))
> +           break;
> +         w2a = w2b;
> +       }
> +
> +      /* Align the final partial of P2.  */
> +      w2 = MERGE (w2b, sh_1, 0, sh_2);
> +    }
> +
> +out:
> +  return write_byte_from_word (dst, w2);
> +}
> +
> +
>  /* Copy SRC to DEST, returning the address of the terminating '\0' in DEST.  */
>  char *
>  STPCPY (char *dest, const char *src)
>  {
> -  size_t len = strlen (src);
> -  return memcpy (dest, src, len + 1) + len;
> +  /* Copy just a few bytes to make DEST aligned.  */
> +  size_t len = (-(uintptr_t) dest) % OPSIZ;
> +  for (; len != 0; len--, ++dest)
> +    {
> +      char c = *src++;
> +      *dest = c;
> +      if (c == '\0')
> +       return dest;
> +    }
> +
> +  /* DEST is now aligned to op_t, SRC may or may not be.  */
> +  uintptr_t ofs = (uintptr_t) src % OPSIZ;
> +  return ofs == 0 ? stpcpy_aligned_loop ((op_t*) dest, (const op_t *) src)
> +                 : stpcpy_unaligned_loop ((op_t*) dest,
> +                                          (const op_t *) (src - ofs) , ofs);
>  }
>  weak_alias (__stpcpy, stpcpy)
>  libc_hidden_def (__stpcpy)
> --
> 2.34.1
>

  parent reply	other threads:[~2023-02-01 19:48 UTC|newest]

Thread overview: 65+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-02-01 17:03 [PATCH v11 00/29] Improve generic string routines Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 01/29] Parameterize op_t from memcopy.h Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 02/29] Parameterize OP_T_THRES " Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 03/29] Add string vectorized find and detection functions Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 04/29] string: Improve generic strlen Adhemerval Zanella
2023-02-01 19:51   ` Noah Goldstein
2023-02-01 17:03 ` [PATCH v11 05/29] string: Improve generic strnlen with memchr Adhemerval Zanella
2023-02-01 19:36   ` Noah Goldstein
2023-02-01 19:57     ` Adhemerval Zanella Netto
2023-02-01 19:39   ` Noah Goldstein
2023-02-01 20:02     ` Adhemerval Zanella Netto
2023-02-01 17:03 ` [PATCH v11 06/29] string: Improve generic strchr Adhemerval Zanella
2023-02-01 19:44   ` Noah Goldstein
2023-02-01 20:03     ` Adhemerval Zanella Netto
2023-02-01 17:03 ` [PATCH v11 07/29] string: Improve generic strchrnul Adhemerval Zanella
2023-02-01 19:49   ` Noah Goldstein
2023-02-01 17:03 ` [PATCH v11 08/29] string: Improve generic strcmp Adhemerval Zanella
2023-02-01 17:34   ` Richard Henderson
2023-02-02 11:57     ` Adhemerval Zanella Netto
2023-02-01 17:03 ` [PATCH v11 09/29] string: Improve generic strncmp Adhemerval Zanella
2023-02-01 19:42   ` Noah Goldstein
2023-02-02 12:01     ` Adhemerval Zanella Netto
2023-02-01 17:03 ` [PATCH v11 10/29] string: Improve generic stpcpy Adhemerval Zanella
2023-02-01 17:29   ` Xi Ruoyao
2023-02-02 13:32     ` Adhemerval Zanella Netto
2023-02-01 17:37   ` Richard Henderson
2023-02-01 19:47   ` Noah Goldstein [this message]
2023-02-01 20:26     ` Richard Henderson
2023-02-01 17:03 ` [PATCH v11 11/29] string: Improve generic strcpy Adhemerval Zanella
2023-02-01 17:38   ` Richard Henderson
2023-02-01 17:03 ` [PATCH v11 12/29] string: Improve generic memchr Adhemerval Zanella
2023-02-01 19:49   ` Noah Goldstein
2023-02-01 17:03 ` [PATCH v11 13/29] string: Improve generic memrchr Adhemerval Zanella
2023-02-01 19:50   ` Noah Goldstein
2023-02-01 17:03 ` [PATCH v11 14/29] hppa: Add memcopy.h Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 15/29] hppa: Add string-fza.h, string-fzc.h, and string-fzi.h Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 16/29] alpha: Add string-fza, string-fzb.h, string-fzi.h, and string-shift.h Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 17/29] arm: Add string-fza.h Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 18/29] powerpc: " Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 19/29] sh: Add string-fzb.h Adhemerval Zanella
2023-02-01 17:03 ` [PATCH v11 20/29] riscv: Add string-fza.h and string-fzi.h Adhemerval Zanella
2023-02-01 17:53   ` Richard Henderson
2023-02-02 12:30     ` Adhemerval Zanella Netto
2023-02-02 16:24       ` Richard Henderson
2023-02-04 16:31       ` Jeff Law
2023-02-05 17:33         ` Richard Henderson
2023-02-01 18:08   ` Noah Goldstein
2023-02-01 17:03 ` [PATCH v11 21/29] string: Hook up the default implementation on test-strlen Adhemerval Zanella
2023-02-01 17:54   ` Richard Henderson
2023-02-01 17:03 ` [PATCH v11 22/29] string: Hook up the default implementation on test-strnlen Adhemerval Zanella
2023-02-01 17:54   ` Richard Henderson
2023-02-01 17:04 ` [PATCH v11 23/29] string: Hook up the default implementation on test-strchr Adhemerval Zanella
2023-02-01 17:55   ` Richard Henderson
2023-02-01 17:04 ` [PATCH v11 24/29] string: Hook up the default implementation on test-strcmp Adhemerval Zanella
2023-02-01 17:55   ` Richard Henderson
2023-02-01 17:04 ` [PATCH v11 25/29] string: Hook up the default implementation on test-strncmp Adhemerval Zanella
2023-02-01 17:56   ` Richard Henderson
2023-02-01 17:04 ` [PATCH v11 26/29] string: Hook up the default implementation on test-stpcpy Adhemerval Zanella
2023-02-01 17:56   ` Richard Henderson
2023-02-01 17:04 ` [PATCH v11 27/29] string: Hook up the default implementation on test-strcpy Adhemerval Zanella
2023-02-01 17:56   ` Richard Henderson
2023-02-01 17:04 ` [PATCH v11 28/29] string: Hook up the default implementation on test-memchr Adhemerval Zanella
2023-02-01 17:57   ` Richard Henderson
2023-02-01 17:04 ` [PATCH v11 29/29] string: Hook up the default implementation on test-memrchr Adhemerval Zanella
2023-02-01 17:57   ` Richard Henderson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAFUsyfLtp41iKdV65QLt2LLY4dNC=HVxv5qbA__Gu5D9OHs=ww@mail.gmail.com' \
    --to=goldstein.w.n@gmail.com \
    --cc=adhemerval.zanella@linaro.org \
    --cc=jeffreyalaw@gmail.com \
    --cc=libc-alpha@sourceware.org \
    --cc=richard.henderson@linaro.org \
    --cc=xry111@xry111.site \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).