From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-yb1-xb33.google.com (mail-yb1-xb33.google.com [IPv6:2607:f8b0:4864:20::b33]) by sourceware.org (Postfix) with ESMTPS id EA4133856DDD for ; Tue, 28 Jun 2022 17:13:21 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org EA4133856DDD Received: by mail-yb1-xb33.google.com with SMTP id o19so16781802ybg.2 for ; Tue, 28 Jun 2022 10:13:21 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=CZj2jvQ6m3CulrFB14Aj+Br0t8wE46T7L+quypaoE9A=; b=ux3o2qmdjO77tREj/KpurgGv4iQI5VPx+zL54d/OT8PJIPi9slaaFjMtB/13rwkNsC vT+f022EMys8sZzGH+Mt3PsT+nije9Xi/PVRvsEfI7/InRKnSK2+xbYsSNK22lxrVSGu j3p0/DCTjDOdOhuGMsHPpebyV4b9hACJSgmMaH7ye7ZmpF5/Qkzs4ZNR2ff9GWDtLCX1 Gy1Kw7S/B9TotZVTdt7WTQEzgA69Z75yzgTy6Sdg0cpP1RlmberDWLbTJunhF290FPrZ JeZ05nmybUjBZAENKCsgdopGHi5x09E0rPqgLFWEBLde/hAJXZ4lFZQImmIsKH89vrZ5 cVVw== X-Gm-Message-State: AJIora+/fnKs72XOGk2qpV2huFX85pffWixUjxGcjMn5Ca7RxdGow2eB QEKV1ctOsTZuczDWYb5+jPGfLInW0YCNUR2fNXuijR3wYao= X-Google-Smtp-Source: AGRyM1ugNFKA/BFrsifA2oAcLVMh6zjb/CJXGIFE0EvZMLabY6uUa9mSa56xFySG/nqHOLzrrS6eJM3CtPIyIUxWpb4= X-Received: by 2002:a5b:ecb:0:b0:66c:91e3:96b1 with SMTP id a11-20020a5b0ecb000000b0066c91e396b1mr17960207ybs.435.1656436401234; Tue, 28 Jun 2022 10:13:21 -0700 (PDT) MIME-Version: 1.0 References: <20220518191424.3630729-1-adhemerval.zanella@linaro.org> <20220518191424.3630729-4-adhemerval.zanella@linaro.org> In-Reply-To: <20220518191424.3630729-4-adhemerval.zanella@linaro.org> From: Noah Goldstein Date: Tue, 28 Jun 2022 10:13:10 -0700 Message-ID: Subject: Re: [PATCH v6 03/10] benchtests: Add arc4random benchtest To: Adhemerval Zanella Cc: GNU C Library Content-Type: text/plain; charset="UTF-8" X-Spam-Status: No, score=-8.5 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, T_SCC_BODY_TEXT_LINE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 28 Jun 2022 17:13:24 -0000 On Wed, May 18, 2022 at 12:15 PM Adhemerval Zanella via Libc-alpha 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 > + . */ > + > +#include "bench-timing.h" > +#include "json-lib.h" > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > +#include > + > +/* 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 > 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 > #include > > -/* 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 >