From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-oa1-x2d.google.com (mail-oa1-x2d.google.com [IPv6:2001:4860:4864:20::2d]) by sourceware.org (Postfix) with ESMTPS id BBBED3858D28 for ; Tue, 3 Oct 2023 17:42:59 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org BBBED3858D28 Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=gmail.com Received: by mail-oa1-x2d.google.com with SMTP id 586e51a60fabf-1dcbd8a56d6so769493fac.1 for ; Tue, 03 Oct 2023 10:42:59 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20230601; t=1696354979; x=1696959779; darn=sourceware.org; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:from:to:cc:subject:date :message-id:reply-to; bh=b9umAXq2wm5lXTrIDP11qakDtx0ezMHELN9mqL1TFSI=; b=J95yxUv03zFsEVDl5wNokmYx75+bktCx5KbqojnjmXJCeoiJe8O32BiNPtjV4Df14r hiJ032R9uKIOxKaJ4aijoSvf39ANvgobnYO1pPL3Ay8VTxW3UEmTap5URwRUXBs0va+G BXFhXn+mZkm4NLdq6JmLWoAL4w7lEIoI9efdXhpodRtRAW47I3goXOfevIbLJfSCdWAa 5j1BZrfstzUsQe37OB4ZxU39ABN6a895WlAFq/cknTaDxGmwoG0M0wOGVEWBux8nXO2A z5JvIwWvX1tjWEFfC7xGuAlb0QklWdWEXl/A/yMee0kMaHw3Sv/5YiGGUzq+GW+iP+VL uk/w== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1696354979; x=1696959779; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=b9umAXq2wm5lXTrIDP11qakDtx0ezMHELN9mqL1TFSI=; b=i1x8UZ3qbYQApoSVZ3xWm6llADJ+ahshqEHRtP+nZ8Z0pbcw4puV/g7kr8yRHHpPzg b7QUy1DvuFzuL/6tURXVsPPROCtxv6ZYDOWlyKBxIDb77mqc8liLoAyS0Z5WpZ34D31e gtbGS8fStSdXjxWb3khHNHIojodzMkNOdlcI3fACVKCIS8lR/UKg5SNDfKocn/cCSI50 lknhQBXtSbPmuQf0TIuDYDOxPM1d6NLEmvxRH9u4u2NC7aj3QHSoWl8nmZ9IiUpxuDAA 0ef5kiJGSuSKu/7AGhU46RWh0Z+mOOnXP5GKE9/vrCqEeznzzddAT1rAS2TCBC7stsM+ Cx8A== X-Gm-Message-State: AOJu0YxagCMOAtlAjPwKufjtUb1FZZfFxD7ySUpDVD3ptGgmA3Th3Vz8 wZo1aFnruFAnC/UZarY2C1jWveInnXrJlw0bwOs= X-Google-Smtp-Source: AGHT+IHorCErDe+rthPIn4/kmWXUxetox+yqlumZ2UfOt7cFgQk2GxWrPX6PUxIv8P5QNkm3zfSXm9JfJEuuP79qwQ0= X-Received: by 2002:a05:6870:1590:b0:1e1:e3a7:c414 with SMTP id j16-20020a056870159000b001e1e3a7c414mr275266oab.16.1696354977550; Tue, 03 Oct 2023 10:42:57 -0700 (PDT) MIME-Version: 1.0 References: <20231003122251.3325435-1-adhemerval.zanella@linaro.org> <20231003122251.3325435-2-adhemerval.zanella@linaro.org> In-Reply-To: <20231003122251.3325435-2-adhemerval.zanella@linaro.org> From: Noah Goldstein Date: Tue, 3 Oct 2023 10:42:45 -0700 Message-ID: Subject: Re: [PATCH v8 1/7] string: Add internal memswap implementation To: Adhemerval Zanella Cc: libc-alpha@sourceware.org, Paul Eggert , Florian Weimer Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Spam-Status: No, score=-8.9 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,FREEMAIL_FROM,GIT_PATCH_0,KAM_SHORT,RCVD_IN_DNSWL_NONE,SPF_HELO_NONE,SPF_PASS,TXREP autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org List-Id: On Tue, Oct 3, 2023 at 5:22=E2=80=AFAM Adhemerval Zanella 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 :=3D \ > tst-xbzero-opt \ > # tests > > +tests-static-internal :=3D \ > + test-memswap \ > +# tests-static-internal > + > +tests-internal :=3D \ > + $(tests-static-internal) \ > + # tests-internal > + > +tests-static :=3D \ > + $(tests-static-internal) \ > + # tests-static > + > # Both tests require the .mo translation files generated by msgfmt. > tests-translation :=3D \ > 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 > + . */ > + > +#include > +#include > +#include > + > +#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 &=3D page_size; > + if (align1 + len >=3D page_size) > + return; > + > + align2 &=3D page_size; > + if (align2 + len >=3D page_size) > + return; > + > + unsigned char *p1 =3D buf1 + align1; > + unsigned char *p2 =3D buf2 + align2; > + for (size_t repeats =3D 0; repeats < 2; ++repeats) > + { > + size_t i, j; > + for (i =3D 0, j =3D 1; i < len; i++, j +=3D 23) > + { > + ref1[i] =3D p1[i] =3D j; > + ref2[i] =3D p2[i] =3D UCHAR_MAX - j; > + } > + > + do_one_test (p1, ref1, p2, ref2, len); > + } > +} > + > +static void > +do_random_tests (void) > +{ > + for (size_t n =3D 0; n < ITERATIONS; n++) > + { > + size_t len, size, size1, size2, align1, align2; > + > + if (n =3D=3D 0) > + { > + len =3D getpagesize (); > + size =3D len + 512; > + size1 =3D size; > + size2 =3D size; > + align1 =3D 512; > + align2 =3D 512; > + } > + else > + { > + if ((random () & 255) =3D=3D 0) > + size =3D 65536; > + else > + size =3D 768; > + if (size > page_size) > + size =3D page_size; > + size1 =3D size; > + size2 =3D size; > + size_t i =3D random (); > + if (i & 3) > + size -=3D 256; > + if (i & 1) > + size1 -=3D 256; > + if (i & 2) > + size2 -=3D 256; Some tests with misaligned size seem in order. > + if (i & 4) > + { > + len =3D random () % size; > + align1 =3D size1 - len - (random () & 31); > + align2 =3D size2 - len - (random () & 31); > + if (align1 > size1) > + align1 =3D 0; > + if (align2 > size2) > + align2 =3D 0; > + } > + else > + { > + align1 =3D random () & 63; > + align2 =3D random () & 63; > + len =3D random () % size; > + if (align1 + len > size1) > + align1 =3D size1 - len; > + if (align2 + len > size2) > + align2 =3D size2 - len; > + } > + } > + unsigned char *p1 =3D buf1 + page_size - size1; > + unsigned char *p2 =3D buf2 + page_size - size2; > + size_t j =3D align1 + len + 256; > + if (j > size1) > + j =3D size1; > + for (size_t i =3D 0; i < j; ++i) > + ref1[i] =3D p1[i] =3D random () & 255; > + > + j =3D align2 + len + 256; > + if (j > size2) > + j =3D size2; > + > + for (size_t i =3D 0; i < j; ++i) > + ref2[i] =3D p2[i] =3D random () & 255; > + > + do_one_test (p1 + align1, ref1 + align1, p2 + align2, ref2 + align= 2, len); > + } > +} > + > +static int > +test_main (void) > +{ > + test_init (); > + /* Use the start of buf1 for reference buffers. */ > + ref1 =3D buf1; > + ref2 =3D buf1 + page_size; > + buf1 =3D ref2 + page_size; > + > + printf ("%23s", ""); > + printf ("\t__memswap\n"); > + > + for (size_t i =3D 0; i < 18; ++i) > + { > + do_test (0, 0, 1 << i); > + do_test (i, 0, 1 << i); > + do_test (0, i, 1 << i); > + do_test (i, i, 1 << i); > + } > + > + for (size_t i =3D 0; i < 32; ++i) > + { > + do_test (0, 0, i); > + do_test (i, 0, i); > + do_test (0, i, i); > + do_test (i, i, i); > + } > + > + for (size_t i =3D 3; i < 32; ++i) > + { > + if ((i & (i - 1)) =3D=3D 0) > + continue; > + do_test (0, 0, 16 * i); > + do_test (i, 0, 16 * i); > + do_test (0, i, 16 * i); > + do_test (i, i, 16 * i); > + } > + > + for (size_t i =3D 19; i <=3D 25; ++i) > + { > + do_test (255, 0, 1 << i); > + do_test (0, 4000, 1 << i); > + do_test (0, 255, i); > + do_test (0, 4000, i); > + } > + > + do_test (0, 0, getpagesize ()); > + > + do_random_tests (); > + > + return 0; > +} > + > +#include > diff --git a/sysdeps/generic/memswap.h b/sysdeps/generic/memswap.h > new file mode 100644 > index 0000000000..f09dae1ebb > --- /dev/null > +++ b/sysdeps/generic/memswap.h > @@ -0,0 +1,41 @@ > +/* Swap the content of two memory blocks, overlap is NOT handled. > + 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 > + . */ > + > +#include > + > +static inline void > +__memswap (void *__restrict p1, void *__restrict p2, size_t n) > +{ > + /* Use multiple small memcpys with constant size to enable inlining on= most > + targets. */ > + enum { SWAP_GENERIC_SIZE =3D 32 }; > + unsigned char tmp[SWAP_GENERIC_SIZE]; > + while (n > SWAP_GENERIC_SIZE) > + { > + memcpy (tmp, p1, SWAP_GENERIC_SIZE); > + p1 =3D __mempcpy (p1, p2, SWAP_GENERIC_SIZE); > + p2 =3D __mempcpy (p2, tmp, SWAP_GENERIC_SIZE); > + n -=3D SWAP_GENERIC_SIZE; > + } > + while (n > 0) > + { > + unsigned char t =3D ((unsigned char *)p1)[--n]; > + ((unsigned char *)p1)[n] =3D ((unsigned char *)p2)[n]; > + ((unsigned char *)p2)[n] =3D t; > + } > +} > -- > 2.34.1 >