Re: [PATCH v8 1/7] string: Add internal memswap implementation

[Noah Goldstein <goldstein.w.n@gmail.com>]

On Fri, Oct 27, 2023 at 5:47 PM Gabriel Ravier <gabravier@gmail.com> wrote:
>
> On 10/27/23 21:23, Adhemerval Zanella Netto wrote:
> >
> > On 03/10/23 14:42, Noah Goldstein wrote:
> >> On Tue, Oct 3, 2023 at 5:22 AM Adhemerval Zanella
> >> <adhemerval.zanella@linaro.org> wrote:
> >>> The prototype is:
> >>>
> >>>    void __memswap (void *restrict p1, void *restrict p2, size_t n)
> >>>
> >>> The function swaps the content of two memory blocks P1 and P2 of
> >>> len N.  Memory overlap is NOT handled.
> >>>
> >>> It will be used on qsort optimization.
> >>>
> >>> Checked on x86_64-linux-gnu and aarch64-linux-gnu.
> >>> ---
> >>>   string/Makefile           |  12 +++
> >>>   string/test-memswap.c     | 192 ++++++++++++++++++++++++++++++++++++++
> >>>   sysdeps/generic/memswap.h |  41 ++++++++
> >>>   3 files changed, 245 insertions(+)
> >>>   create mode 100644 string/test-memswap.c
> >>>   create mode 100644 sysdeps/generic/memswap.h
> >>>
> >>> diff --git a/string/Makefile b/string/Makefile
> >>> index 8cdfd5b000..fb101db778 100644
> >>> --- a/string/Makefile
> >>> +++ b/string/Makefile
> >>> @@ -209,6 +209,18 @@ tests := \
> >>>     tst-xbzero-opt \
> >>>   # tests
> >>>
> >>> +tests-static-internal := \
> >>> +  test-memswap \
> >>> +# tests-static-internal
> >>> +
> >>> +tests-internal := \
> >>> +  $(tests-static-internal) \
> >>> +  # tests-internal
> >>> +
> >>> +tests-static := \
> >>> +  $(tests-static-internal) \
> >>> +  # tests-static
> >>> +
> >>>   # Both tests require the .mo translation files generated by msgfmt.
> >>>   tests-translation := \
> >>>     tst-strerror \
> >>> diff --git a/string/test-memswap.c b/string/test-memswap.c
> >>> new file mode 100644
> >>> index 0000000000..162beb91e3
> >>> --- /dev/null
> >>> +++ b/string/test-memswap.c
> >>> @@ -0,0 +1,192 @@
> >>> +/* Test and measure memcpy functions.
> >>> +   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
> >>> +   <https://www.gnu.org/licenses/>.  */
> >>> +
> >>> +#include <string.h>
> >>> +#include <support/check.h>
> >>> +#include <memswap.h>
> >>> +
> >>> +#define TEST_MAIN
> >>> +#define BUF1PAGES 3
> >>> +#include "test-string.h"
> >>> +
> >>> +static unsigned char *ref1;
> >>> +static unsigned char *ref2;
> >>> +
> >>> +static void
> >>> +do_one_test (unsigned char *p1, unsigned char *ref1, unsigned char *p2,
> >>> +            unsigned char *ref2, size_t len)
> >>> +{
> >>> +  __memswap (p1, p2, len);
> >>> +
> >>> +  TEST_COMPARE_BLOB (p1, len, ref2, len);
> >>> +  TEST_COMPARE_BLOB (p2, len, ref1, len);
> >>> +}
> >>> +
> >>> +static inline void
> >>> +do_test (size_t align1, size_t align2, size_t len)
> >>> +{
> >>> +  align1 &= page_size;
> >>> +  if (align1 + len >= page_size)
> >>> +    return;
> >>> +
> >>> +  align2 &= page_size;
> >>> +  if (align2 + len >= page_size)
> >>> +    return;
> >>> +
> >>> +  unsigned char *p1 = buf1 + align1;
> >>> +  unsigned char *p2 = buf2 + align2;
> >>> +  for (size_t repeats = 0; repeats < 2; ++repeats)
> >>> +    {
> >>> +      size_t i, j;
> >>> +      for (i = 0, j = 1; i < len; i++, j += 23)
> >>> +       {
> >>> +         ref1[i] = p1[i] = j;
> >>> +         ref2[i] = p2[i] = UCHAR_MAX - j;
> >>> +       }
> >>> +
> >>> +      do_one_test (p1, ref1, p2, ref2, len);
> >>> +    }
> >>> +}
> >>> +
> >>> +static void
> >>> +do_random_tests (void)
> >>> +{
> >>> +  for (size_t n = 0; n < ITERATIONS; n++)
> >>> +    {
> >>> +      size_t len, size, size1, size2, align1, align2;
> >>> +
> >>> +      if (n == 0)
> >>> +        {
> >>> +          len = getpagesize ();
> >>> +          size = len + 512;
> >>> +          size1 = size;
> >>> +          size2 = size;
> >>> +          align1 = 512;
> >>> +          align2 = 512;
> >>> +        }
> >>> +      else
> >>> +        {
> >>> +          if ((random () & 255) == 0)
> >>> +            size = 65536;
> >>> +          else
> >>> +            size = 768;
> >>> +          if (size > page_size)
> >>> +            size = page_size;
> >>> +          size1 = size;
> >>> +          size2 = size;
> >>> +          size_t i = random ();
> >>> +          if (i & 3)
> >>> +            size -= 256;
> >>> +          if (i & 1)
> >>> +            size1 -= 256;
> >>> +          if (i & 2)
> >>> +            size2 -= 256;
> >> Some tests with misaligned size seem in order.
> > I am not sure I understood what is wrong here, by 'in order' do you mean
> > exactly what?
>
> My interpretation was that this is the less commonly used meaning of "in
> order" as meaning "appropriate in the current circumstances", i.e. that
> tests that invoke memswap with unaligned operands seem like they would
> be useful.
^
>