From: Noah Goldstein <goldstein.w.n@gmail.com>
To: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Cc: GNU C Library <libc-alpha@sourceware.org>
Subject: Re: [PATCH v6 03/10] benchtests: Add arc4random benchtest
Date: Tue, 28 Jun 2022 10:13:10 -0700 [thread overview]
Message-ID: <CAFUsyfKzbMc_h8pO_i_yB5BKhrb08EmyK8o2_5fZansaaxPhHA@mail.gmail.com> (raw)
In-Reply-To: <20220518191424.3630729-4-adhemerval.zanella@linaro.org>
On Wed, May 18, 2022 at 12:15 PM Adhemerval Zanella via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> It shows both throughput (total bytes obtained in the test duration)
> and latecy for both arc4random and arc4random_buf with different
> sizes.
>
> Checked on x86_64-linux-gnu, aarch64-linux, and powerpc64le-linux-gnu.
> ---
> benchtests/Makefile | 5 +-
> benchtests/bench-arc4random.c | 224 ++++++++++++++++++++++++++++++++++
> stdlib/arc4random.c | 14 +--
> 3 files changed, 235 insertions(+), 8 deletions(-)
> create mode 100644 benchtests/bench-arc4random.c
>
> diff --git a/benchtests/Makefile b/benchtests/Makefile
> index de9de5cf58..76583d45a3 100644
> --- a/benchtests/Makefile
> +++ b/benchtests/Makefile
> @@ -227,7 +227,10 @@ LOCALES := \
> include ../gen-locales.mk
> endif
>
> -stdlib-benchset := strtod
> +stdlib-benchset := \
> + arc4random \
> + strtod \
> + # stdlib-benchset
>
> stdio-common-benchset := sprintf
>
> diff --git a/benchtests/bench-arc4random.c b/benchtests/bench-arc4random.c
> new file mode 100644
> index 0000000000..626f2ba48c
> --- /dev/null
> +++ b/benchtests/bench-arc4random.c
> @@ -0,0 +1,224 @@
> +/* arc4random benchmarks.
> + Copyright (C) 2022 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 "bench-timing.h"
> +#include "json-lib.h"
> +#include <array_length.h>
> +#include <intprops.h>
> +#include <signal.h>
> +#include <stdbool.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <support/support.h>
> +#include <support/timespec.h>
> +#include <support/xthread.h>
> +
> +/* Prevent compiler to optimize away call. */
> +#define DO_NOT_OPTIMIZE(__value) \
> + ({ \
> + __typeof (__value) __v = (__value); \
> + asm volatile("" : : "r,m" (__v) : "memory");\
> + })
This might be useful to have in the support header.
bench-hash-func.c has essentially the same code.
> +
> +static volatile sig_atomic_t timer_finished;
> +
> +static void timer_callback (int unused)
> +{
> + timer_finished = 1;
> +}
> +
> +static timer_t timer;
> +
> +/* Run for approximately DURATION seconds, and it does not matter who
> + receive the signal (so not need to mask it on main thread). */
> +static void
> +timer_start (void)
> +{
> + timer_finished = 0;
> + timer = support_create_timer (DURATION, 0, false, timer_callback);
> +}
> +static void
> +timer_stop (void)
> +{
> + support_delete_timer (timer);
> +}
> +
> +static const uint32_t sizes[] = { 0, 16, 32, 48, 64, 80, 96, 112, 128 };
> +
> +static double
> +bench_throughput (void)
> +{
> + uint64_t n = 0;
> +
> + struct timespec start, end;
> + clock_gettime (CLOCK_MONOTONIC, &start);
> + while (1)
> + {
> + DO_NOT_OPTIMIZE (arc4random ());
> + n++;
> +
> + if (timer_finished == 1)
> + break;
> + }
> + clock_gettime (CLOCK_MONOTONIC, &end);
> + struct timespec diff = timespec_sub (end, start);
> +
> + double total = (double) n * sizeof (uint32_t);
> + double duration = (double) diff.tv_sec
> + + (double) diff.tv_nsec / TIMESPEC_HZ;
> +
> + return total / duration;
> +}
> +
> +static double
> +bench_latency (void)
> +{
> + timing_t start, stop, cur;
> + const size_t iters = 1024;
> +
> + TIMING_NOW (start);
> + for (size_t i = 0; i < iters; i++)
> + DO_NOT_OPTIMIZE (arc4random ());
> + TIMING_NOW (stop);
> +
> + TIMING_DIFF (cur, start, stop);
> +
> + return (double) (cur) / (double) iters;
> +}
> +
> +static double
> +bench_buf_throughput (size_t len)
> +{
> + uint8_t buf[len];
> + uint64_t n = 0;
> +
> + struct timespec start, end;
> + clock_gettime (CLOCK_MONOTONIC, &start);
> + while (1)
> + {
> + arc4random_buf (buf, len);
> + n++;
> +
> + if (timer_finished == 1)
> + break;
> + }
> + clock_gettime (CLOCK_MONOTONIC, &end);
> + struct timespec diff = timespec_sub (end, start);
> +
> + double total = (double) n * len;
> + double duration = (double) diff.tv_sec
> + + (double) diff.tv_nsec / TIMESPEC_HZ;
> +
> + return total / duration;
> +}
> +
> +static double
> +bench_buf_latency (size_t len)
> +{
> + timing_t start, stop, cur;
> + const size_t iters = 1024;
> +
> + uint8_t buf[len];
> +
> + TIMING_NOW (start);
> + for (size_t i = 0; i < iters; i++)
> + arc4random_buf (buf, len);
> + TIMING_NOW (stop);
> +
> + TIMING_DIFF (cur, start, stop);
> +
> + return (double) (cur) / (double) iters;
> +}
> +
> +static void
> +bench_singlethread (json_ctx_t *json_ctx)
> +{
> + json_element_object_begin (json_ctx);
> +
> + json_array_begin (json_ctx, "throughput");
> + for (int i = 0; i < array_length (sizes); i++)
> + {
> + timer_start ();
> + double r = sizes[i] == 0
> + ? bench_throughput () : bench_buf_throughput (sizes[i]);
> + timer_stop ();
> +
> + json_element_double (json_ctx, r);
> + }
> + json_array_end (json_ctx);
> +
> + json_array_begin (json_ctx, "latency");
> + for (int i = 0; i < array_length (sizes); i++)
> + {
> + timer_start ();
> + double r = sizes[i] == 0
> + ? bench_latency () : bench_buf_latency (sizes[i]);
> + timer_stop ();
> +
> + json_element_double (json_ctx, r);
> + }
> + json_array_end (json_ctx);
> +
> + json_element_object_end (json_ctx);
> +}
> +
> +static void
> +run_bench (json_ctx_t *json_ctx, const char *name,
> + char *const*fnames, size_t fnameslen,
> + void (*bench)(json_ctx_t *ctx))
> +{
> + json_attr_object_begin (json_ctx, name);
> + json_array_begin (json_ctx, "functions");
> + for (int i = 0; i < fnameslen; i++)
> + json_element_string (json_ctx, fnames[i]);
> + json_array_end (json_ctx);
> +
> + json_array_begin (json_ctx, "results");
> + bench (json_ctx);
> + json_array_end (json_ctx);
> + json_attr_object_end (json_ctx);
> +}
> +
> +static int
> +do_test (void)
> +{
> + char *fnames[array_length (sizes)];
> + for (int i = 0; i < array_length (sizes); i++)
> + if (sizes[i] == 0)
> + fnames[i] = xasprintf ("arc4random");
> + else
> + fnames[i] = xasprintf ("arc4random_buf(%u)", sizes[i]);
> +
> + json_ctx_t json_ctx;
> + json_init (&json_ctx, 0, stdout);
> +
> + json_document_begin (&json_ctx);
> + json_attr_string (&json_ctx, "timing_type", TIMING_TYPE);
> +
> + run_bench (&json_ctx, "single-thread", fnames, array_length (fnames),
> + bench_singlethread);
> +
> + json_document_end (&json_ctx);
> +
> + for (int i = 0; i < array_length (sizes); i++)
> + free (fnames[i]);
> +
> + return 0;
> +}
> +
> +#include <support/test-driver.c>
> diff --git a/stdlib/arc4random.c b/stdlib/arc4random.c
> index f6553dfd7d..74c917b9b9 100644
> --- a/stdlib/arc4random.c
> +++ b/stdlib/arc4random.c
> @@ -25,9 +25,9 @@
> #include <sys/param.h>
> #include <sys/random.h>
>
> -/* Besides the cipher state 'ctx', it keeps two counters: 'have' is the
> - current valid bytes not yet consumed in 'buf', while 'count' is the maximum
> - number of bytes until a reseed.
> +/* arc4random keeps two counters: 'have' is the current valid bytes not yet
> + consumed in 'buf' while 'count' is the maximum number of bytes until a
> + reseed.
>
> Both the initial seed and reseed try to obtain entropy from the kernel
> and abort the process if none could be obtained.
> @@ -81,10 +81,10 @@ arc4random_getrandom_failure (void)
> __libc_fatal ("Fatal glibc error: Cannot get entropy for arc4random\n");
> }
>
> -/* Fork detection is done by checking if MADV_WIPEONFORK supported. If not
> - the fork callback will reset the state on the fork call. It does not
> - handle direct clone calls, nor vfork or _Fork (arc4random is not
> - async-signal-safe due the internal lock usage). */
> +/* If the kernel supports MADV_WIPEONFORK it is used to provide fork
> + detection. Otherwise, the state is resetied with an atfork handler. The
> + latter does not handle direct clone calls, nor vfork or _Fork (arc4random
> + is not async-signal-safe due to the internal lock usage). */
> static void
> arc4random_init (uint8_t *buf)
> {
> --
> 2.34.1
>
next prev parent reply other threads:[~2022-06-28 17:13 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-18 19:14 [PATCH v6 00/10] Add arc4random support Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 01/10] stdlib: Add arc4random, arc4random_buf, and arc4random_uniform (BZ #4417) Adhemerval Zanella
2022-06-02 9:44 ` Yann Droneaud
2022-06-10 17:26 ` Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 02/10] stdlib: Add arc4random tests Adhemerval Zanella
2022-06-28 11:59 ` Florian Weimer
2022-06-28 16:58 ` Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 03/10] benchtests: Add arc4random benchtest Adhemerval Zanella
2022-06-28 12:01 ` Florian Weimer
2022-06-28 17:06 ` Adhemerval Zanella
2022-06-28 17:13 ` Noah Goldstein [this message]
2022-06-28 17:14 ` Noah Goldstein
2022-06-28 17:22 ` Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 04/10] aarch64: Add optimized chacha20 Adhemerval Zanella
2022-06-28 12:03 ` Florian Weimer
2022-06-28 17:39 ` Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 05/10] x86: Add SSE2 " Adhemerval Zanella
2022-05-18 20:44 ` Noah Goldstein
2022-05-18 19:14 ` [PATCH v6 06/10] x86: Add AVX2 " Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 07/10] powerpc64: Add " Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 08/10] s390x: " Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 09/10] stdlib: Add TLS optimization to arc4random Adhemerval Zanella
2022-06-28 12:19 ` Florian Weimer
2022-06-28 19:45 ` Adhemerval Zanella
2022-05-18 19:14 ` [PATCH v6 10/10] manual: Add documentation for arc4random functions Adhemerval Zanella
2022-06-28 12:09 ` Florian Weimer
2022-06-28 19:15 ` Adhemerval Zanella
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=CAFUsyfKzbMc_h8pO_i_yB5BKhrb08EmyK8o2_5fZansaaxPhHA@mail.gmail.com \
--to=goldstein.w.n@gmail.com \
--cc=adhemerval.zanella@linaro.org \
--cc=libc-alpha@sourceware.org \
/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).